UNIVERSITY OF CALIFORNIA
Santa Barbara
A Psychological Calculus for Welfare Tradeoffs
A dissertation submitted in partial satisfaction of the
requirements for the degree Doctor of Philosophy
in Psychology
by
Andrew Wyatt Delton
Committee in charge:
Professor Leda Cosmides, Chair
Professor John Tooby
Professor Daphne B. Bugental
Professor Tamsin C. German
Professor Nancy L. Collins
September 2010
UMI Number: 3427833
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
UMI 3427833
Copyright 2010 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106-1346
The dissertation of Andrew Wyatt Delton is approved.
_____________________________________________
John Tooby
_____________________________________________
Daphne B. Bugental
_____________________________________________
Tamsin C. German
_____________________________________________
Nancy L. Collins
_____________________________________________
Leda Cosmides, Committee Chair
August 2010
A Psychological Calculus for Welfare Tradeoffs
Copyright © 2010
by
Andrew Wyatt Delton
iii
Acknowledgements
I dedicate this to my parents, Jim and Sharon Delton, for their years of love
and support. Whether they know it or not, this is all their fault.
I would like to acknowledge my collaborators on much of this research,
Daniel Sznycer and Julian Lim, whose contributions greatly improved the research
and whose jokes kept me greatly entertained. Thanks also go to the members of the
Center for Evolutionary Psychology for their valuable feedback throughout the years.
Daphne Bugental, Tamsin German, and Nancy Collins all graciously agreed to
be committee members and, perhaps more importantly, generously shared of their
time and expertise during my entire graduate career. I would especially like to thank
Daphne for her wonderful grant-writing advice that made the impossible possible.
Scott Grafton was instrumental in helping us to design the pilot fMRI
experiment presented in the appendices. Many thanks to him for his collaboration on
this project.
It’s difficult to express my appreciation for my advisors, Leda Cosmides and
John Tooby, particularly given the circumstances under which this research was
conducted. Despite numerous acts of god that threatened their health, home, and
sanity, they remained insightful, supportive, and somehow—through a supreme act of
biblical defiance—good-humored. Most people know their scientific heroes through
works of history. I was luck enough to be trained by mine.
Finally, I would like to thank my lifetime collaborator and the love of my life,
Tess Robertson, for making happiness all the rage.
iv
Vita of Andrew Wyatt Delton
August 2010
Research Interests
Evolutionary psychological approaches to cooperation and altruism, evolutionary
game theory, social neuroscience, and neuroimaging. Supervised by Professors Leda
Cosmides and John Tooby.
Education
BS in Psychology (2003), honors college graduate, with highest distinction (Summa
Cum Laude), Arizona State University.
Thesis: Memory for faces and evolved motivational states.
Thesis Advisor: Dr. Douglas Kenrick
MA in Psychology (2006), University of California, Santa Barbara.
Thesis: What counts as free riding? Using intentions—but not contribution
level—to identify free riders.
Thesis Advisor: Dr. Leda Cosmides
PhD in Evolutionary and Developmental Psychology (2010), with an emphasis in
Quantitative Methods for the Social Sciences, an emphasis in Cognitive
Sciences, and an emphasis in Interdisciplinary Human Development,
University of California, Santa Barbara.
Dissertation Advisors: Dr. Leda Cosmides and Dr. John Tooby.
Publications
Cimino, A. & Delton, A. W. (2010). On the perception of newcomers: Toward an
evolved psychology of intergenerational coalitions. Human Nature, 21, 186202.
Delton, A. W. & Cimino, A. (2010). Exploring the NEWCOMER concept: Experimental
tests of a cognitive model. Evolutionary Psychology, 8, 317-335.
Delton, A. W., Krasnow, M. M., Cosmides, L., & Tooby, J. (2010). Evolution of
Fairness: Rereading the data. Science, 329 (5990), 389.
Delton, A. W., Robertson, T. E., & Kenrick, D. T. (2006). The mating game isn’t
over: A reply to Buller’s critique of the evolutionary psychology of mating.
Evolutionary Psychology, 4, 262-273.
v
Griskevicius, V., Delton, A. W., & Robertson, T. E., Tybur, J. M. (in press).
Environmental contingency in life-history strategies: Influence of mortality
and socioeconomic status on reproductive timing. Journal of Personality and
Social Psychology.
Kenrick, D. T., Delton, A. W., Robertson, T. E., Becker, D. V., & Neuberg, S. L.
(2007). How the mind warps. In J. P. Forgas, M. G. Haselton, & W. Von
Hippel (Eds.), The Evolution of the social mind: Evolutionary psychology and
social cognition (pp. 49-68). New York: Psychology Press.
Klein, S. B., Robertson, T. E., & Delton, A. W. (2010). Facing the future: Memory as
an evolved system for planning future acts. Memory and Cognition, 38, 13-22.
Klein, S. B., Robertson, T. E., & Delton, A. W. (in press). The future-orientation of
memory: Planning as a key component mediating the high levels of recall
found with survival processing. Memory.
Maner, J. K., Kenrick, D. T., Becker, D. V., Delton, A. W., Hofer, B., Wilbur, C. J.,
& Neuberg, S. L. (2003). Sexually selective cognition: Beauty captures the
mind of the beholder. Journal of Personality and Social Psychology, 85,
1107-1120.
Maner, J. K., Kenrick, D. T., Becker, D. V., Robertson, T. E., Hofer, B., Neuberg, S.
L., Delton, A. W., Butner, J., & Schaller, M. (2005). Functional projection:
How fundamental social motives can bias interpersonal perception. Journal of
Personality and Social Psychology, 88, 63-78.
Grants
Co-PI on The Hidden Correlates of Social Exclusion (NSF funded; $400,000). Grant
investigates evolved, domain-specific responses to social exclusion and their
hormonal correlates. With Co-PIs Theresa Robertson, Leda Cosmides, & John
Tooby.
Invited Talks
Delton, A. W. (2009, March). How the mind makes welfare tradeoffs. Invited paper
presentation at the 3UC Conference, San Luis Obispo, CA.
Delton, A. W. & Krasnow, M. M. (2008, February). A cue-theoretic approach to
cooperation. Invited paper presentation at the Evolution and Sociality of Mind
Conference, Santa Barbara, CA.
vi
Published Conference Presentations
Petersen, M. B., Delton, A. W., Robertson, T. E., Tooby, J. and Cosmides, L. (2008,
April) Politics in the Evolved Mind: Political Parties and Coalitional
Reasoning. Paper presented at the Midwest Political Science Association 66th
Annual National Conference, Chicago, IL. PDF available at
http://www.allacademic.com/meta/p268315_index.html.
Professional Employment
Graduate Researcher (2006-2007, 2008-2010) Supported on project entitled: “NIH
Director’s Pioneer Award for a Computational Approach to Motivation,” UC
Santa Barbara, Principal Investigators: Leda Cosmides & John Tooby. This
research breaks down problems of motivation into their elementary
computational components using theories from evolutionary biology.
Student Researcher (2003) Supported on projected entitled: “Social Motives and
Cognition Grant,” Arizona State University, Principal Investigators: Douglas
T. Kenrick, Steven L. Neuberg, & Mark Schaller. This research examined
how fundamental motivational states relevant to social relationships influence
basic social cognition processes.
Assistant Textbook Researcher (2003) for a revision of Social Psychology:
Unraveling the mystery by Douglas T. Kenrick, Steven L. Neuberg, and
Robert B. Cialdini.
Duties: Assisted in review of current social psychological literature.
Instructional Activities
Lecturer (Instructor of Record)
Developmental Psychology (2008)
Developmental Psychology (2006)
Laboratory Instructor
Advanced Research Methods (2005)
Discussion Leader
Introductory Statistics (Spring 2005, Summer 2005)
Introductory Psychology (2004)
Teaching Assistant
Developmental Psychology (2004)
Social Psychology (2004)
vii
Fellowships, Scholarships, and Awards
New Investigator Award from the Human Behavior and Evolution Society (2009)
Fellowship for the Summer Institute in Cognitive Neuroscience: Social Neuroscience
& Neuroeconomics, UC Santa Barbara (2007)
National Science Foundation Graduate Fellowship Honorable Mention (2005)
Winner, Special Competition of Time-sharing Experiments in the Social Sciences
(TESS; an NSF funded research platform) for “Motivational Systems in
Political Collective Action: Managing Free-Riders, Exiters, and Recruits.”
Co-authored with Leda Cosmides (2005)
National Science Foundation Graduate Fellowship Honorable Mention (2004)
UCSB Chancellor’s Funds for Professional Activities (2004)
UCSB Chancellor’s Fellowship (2003)
Outstanding Undergraduate Research Paper in Psychology, ASU (2003)
National Dean’s List, ASU (2003)
ASU Dean’s Council Award for Psychology (2002)
National Merit Finalist Tuition Waiver (1999-2003)
National Merit Finalist Resident University Award (1999-2003)
National Merit Finalist University Award (1999-2003)
ASU President’s Scholarship (1999-2003)
ASU Dean’s List (1999-2003)
viii
Abstract
A Psychological Calculus for Welfare Tradeoffs
by
Andrew Wyatt Delton
Members of social species routinely make decisions that involve welfare
allocations—decisions that impact the welfare of two or more parties. These decisions
often involve welfare tradeoffs such that increasing one organism’s welfare comes at
the expense of another organisms’ welfare. The overarching goal of this dissertation
is to understand how the human mind makes these tradeoffs. First, I review a number
of theories from evolutionary biology that predict how and why welfare tradeoffs
should be made. Collectively, these theories predict that organisms must make
estimates of a number of features about themselves and others (for instance, degree of
relatedness and quality as a reciprocation partner). To make effective welfare
tradeoffs, however, these estimates must combined into a summary variable—a
welfare tradeoff ratio—that is used to regulate social decision-making. By consulting
this variable, an organism can determine when it is and is not appropriate to cede
personal welfare on behalf of another. Second, I provide experimental, cross-national
evidence that welfare tradeoff ratios exist and regulate social decision-making in
ix
surprisingly precise ways. Demonstrating this requires developing a new quantitative
measurement device and methods to analyze the resulting data. Third, in further
experiments I show that the mind can also estimate welfare tradeoff ratios stored in
the minds of others. Finally, I review the connections between welfare tradeoff theory
and other approaches to social decision-making. The discovery of this
neurocomputational element helps to explain how complex behaviors—such as
cooperation, generosity, and aggression—can arise from a physical device such as the
human brain.
x
Table of Contents
Chapter 1: Theoretical Overview of Welfare Tradeoff Theory ............................. 1
Computational Theories......................................................................................... 2
Evolutionary Biological Theories of Valuation and Welfare Tradeoffs ............ 5
Helping Through Recent Common Descent ..................................................... 6
Direct Reciprocity ............................................................................................... 9
Reputation & Indirect Reciprocity.................................................................. 13
Graph Theory.................................................................................................... 15
The Asymmetric War of Attrition................................................................... 16
Multi-Level Selection Theory........................................................................... 18
Mutualism, Coordination, Pseudo-reciprocity, Markets, and Externalities 19
Summary............................................................................................................ 22
Computational Mechanisms that Compute and Instantiate Valuation ........... 23
A Kinship Index ................................................................................................ 24
Indices Relevant to Reciprocity ....................................................................... 26
A Formidability Index ...................................................................................... 28
Sexual Value Indices ......................................................................................... 28
The Problem of Integration and the Necessity of Welfare Tradeoff Ratios 30
Standards of Evidence for the Existence of Welfare Tradeoff Ratios ......... 31
The Problem of Estimation .............................................................................. 34
Standards of Evidence that the Mind Estimates Welfare Tradeoff Ratios . 35
Summary of Predictions from Welfare Tradeoff Theory ................................. 36
Chapter 2: The Existence of Welfare Tradeoff Ratios .......................................... 38
The Welfare Tradeoff Task.................................................................................. 38
Theoretical Overview........................................................................................ 38
The Task ............................................................................................................ 46
Methods.................................................................................................................. 53
General Design .................................................................................................. 53
Subjects and Specific Design............................................................................ 53
Dependent Measures......................................................................................... 58
Results .................................................................................................................... 63
Result 1: A Single Switch Point Predicts Decisions ....................................... 64
Result 2: Consistency is Not Due to the Artificial Nature of the Task......... 66
Result 3: Subjects’ Decisions Represent Actual Valuation of Others; They
Are Not “Cheap Talk”...................................................................................... 68
Result 4: WTRs Are Person-Specific; They Carry More Information Than a
Simple Other-Regarding Preference............................................................... 68
Result 5: Individualized WTRs Are Correlated with Decisions Involving
Only Third-Parties............................................................................................ 73
Results 6 & 7: Welfare tradeoff ratios behave as other magnitudes, but are
algorithmically separable from other magnitudes......................................... 77
Correlations of the WTR Measure with Other Measures............................. 88
Summary of Chapter 2: The Existence of Welfare Tradeoff Ratios................ 90
xi
Chapter 3: The Estimation of Welfare Tradeoff Ratios........................................ 92
The Logic of Welfare Tradeoff Ratio Estimation .............................................. 93
Study 6: Variations in Cost Incurred.................................................................. 95
Methods.............................................................................................................. 95
Results ................................................................................................................ 98
Study 7: Varying Costs with Constant Intentions ........................................... 103
Method ............................................................................................................. 103
Results .............................................................................................................. 104
Study 8: Manipulating the Target of Costs Incurred ...................................... 104
Method ............................................................................................................. 105
Results .............................................................................................................. 105
Discussion......................................................................................................... 106
Studies 9 & 10: Varying Benefits Provided ...................................................... 107
Methods............................................................................................................ 108
Results & Discussion....................................................................................... 109
Summary of Chapter 3: The Estimation of Welfare Tradeoff Ratios ........... 110
Chapter 4: Discussion and Conclusions................................................................ 111
Extensions and Future Directions ..................................................................... 111
Ongoing Work on Welfare Tradeoff Theory ............................................... 111
Studying Welfare Tradeoffs Without Numbers........................................... 113
Broadening the Breadth of Decisions............................................................ 114
Comparisons to Other Research Programs...................................................... 116
Affective Heuristic Approaches ..................................................................... 116
The Elementary Forms of Human Interaction Approach .......................... 118
Other-Regarding Preferences Approaches................................................... 121
Interdependence Theory ................................................................................ 123
Social Contract Theory................................................................................... 125
Parting Thoughts: Existentialism Meets Evolutionary Psychology ............... 128
Appendix 1: Example Materials for Study 1........................................................ 129
Appendix 2: Screen Shots of Computerized Tasks.............................................. 138
Appendix 3: Materials for Study 3 ........................................................................ 139
Appendix 4: Material for Study 4.......................................................................... 142
Appendix 5: Materials for Study 5 ........................................................................ 146
Appendix 6: Probing the Sensitivity of Welfare Tradeoff Computations ......... 151
Appendix 7: Dissociating Welfare Tradeoff Computation from Other Intuitive
Economic Computation: A Pilot Functional Magnetic Resonance Imaging Study
................................................................................................................................... 159
Appendix 8: Materials for Chapter 3.................................................................... 177
References................................................................................................................ 188
xii
Chapter 1: Theoretical Overview of Welfare Tradeoff Theory
Chad sits alone in his dorm room with an important decision to make: Does he
help tutor a friend who desperately needs to pass a calculus exam or does he go on a
blind date as originally planned?
Paola also has an important decision to make: She’s currently safe from the
guerilla fighters as she hides in the forest just outside her small village, but her
younger brother has yet to make it out. Does she go back to help him or does she
remain where she is?
Whether considering mundane daily life or more serious circumstances,
decisions involving welfare allocations—that is, decisions impacting the welfare of
two or more parties—are a ubiquitous part of life for members of a social species.
Often, welfare allocations involve welfare tradeoffs, as in the above examples. Such
decisions involve a tradeoff between personal welfare and the welfare of other
(Tooby & Cosmides, 1990; Tooby, Cosmides, Sell, Lieberman, & Sznycer, 2008)
The first goal of this dissertation is to elucidate the engineering complexities of
making welfare allocations and tradeoffs—that is, to describe at least some of
representations and algorithms necessary for a machine like the human mind to
successfully make welfare tradeoffs and to describe what would count as success. The
second goal is to provide preliminary evidence for the existence of a critical
component of this machinery, the welfare tradeoff ratio (WTR), a variable computed
by the mind that specifies how much of one’s own welfare should be traded off to
1
increase or decrease the welfare of another individual. The third goal is to provide
preliminary evidence that the mind has the ability to estimate the welfare tradeoff
ratios in the minds of others. Finally, I will compare welfare tradeoff ratio theory to
relevant alternative approaches.
Computational Theories
A few pounds of protinaceous fibers strung along a series of semi-rigid rods,
all interconnected by miles of wire and plumbing, with a stretchable sack to hold it
together. How does one get a machine made of meat to make welfare tradeoffs? I
approach the problem from a computational perspective (Marr, 1982). Such an
approach focuses on formally characterizing the information to be processed (i.e. the
representations the system operates on and the primitives from which they are built)
and the transformations performed on these representations (i.e. the rules of the
system). This style of analysis has led to numerous insights in linguistics (Jackendoff,
2002), vision (Marr, 1982), and in other domains of thought, including social thought
(Jackendoff, 2006; Marr, 1982; Pinker, 2007).
As an example, consider that basic arithmetic is a simple computational
system (see Marr, 1982, for an extended discussion). It has a set of primitives (the
numerals “0” through “9”, “+”, “–”, and “=”); real world entities can be represented
within the system (e.g., three socks can be represented by the primitive “3”); and
larger representations can be built up (e.g., “3 + 4”) and transformed into new
representations (e.g., via the rules of arithmetic “3 + 4” is transformed into “7”). As is
always the case in computational systems, not all possible properties of the to-be-
2
represented entity are actually represented (e.g., given a set of three socks, many
things besides the size of the set could be represented, such as their softness); which
features are represented is determined by the problem that is being solved and the
representational system activated (e.g., the number of socks—not their shape, color,
or texture—is the only feature relevant for arithmetical inference). Finally, it is the
logic that a notational system embodies—not its particular orthography—that matters
(e.g., instead of Arabic numerals, we could use “I”, “V”, “X”, “L”, “C”, “D”, and
“M”; although the mechanics of arithmetic are more difficult using Roman numerals,
the underlying logic is identical.)
But how do we extrapolate from computational theories concerned with
seemingly objective relationships, such as arithmetical inference, to devise theories of
seemingly subjective phenomena, such as making “appropriate” welfare tradeoffs?
Indeed, as a rough heuristic it is useful to draw a distinction between computation that
generates representations about states of the world (loosely, what exists?) and
computation that generates representations about valuation and motivation (loosely,
what should be done?) because different approaches apply to each. Although one
cannot go directly from facts of the world to statements about what one should do in
the world, it is still possible to scientifically investigate the computations underlying
motivation.
A familiar example of value computation with a long history in psychology is
null-hypothesis significance testing. This approach divides the task of statistics into
two. First, generate a statistical description of the world: the computation of means,
3
standard deviations, t- or F-statistics, p values and so on. Second, generate a
statistical inference by determining a significance threshold (e.g., p must be less than
.05 to conclude or act as if there is a difference between groups). This latter task is
ultimately a value judgment—as anyone familiar with the ease of calculating a mean
versus the difficulty of deciding what result is truly “significant” can attest—but it
can, in principle, be arrived at through means that quantify the likely costs of making
various kinds of decisions (e.g., the cost of rejecting true hypotheses, the cost of
allowing false hypotheses to be believed true). Of course, these computations were
devised by human scientists and statisticians over the course of a few centuries. We
are, however, concerned with computational devices crafted by natural selection over
deep time, the more common focus of computational approaches.
Marr’s (1982) pioneering work on applying a computational approach was
targeted at the problem of vision. The assumption underlying this approach was that
there really is a three dimensional world out there, with surfaces, color, etc. This
implies that the laws of physics, such as the laws of optics and the laws of motion,
could be used to deduce what computational procedures the mind/brain would need to
successfully create useful descriptions of the physical world (Leslie, 1994; Shepard,
1984). But how do we develop theories of computation for valuation? Although
physics may not be of much direct help here, we can instead turn to evolutionary
biology. Over the past five decades, evolutionary biologists have developed a variety
of theories for predicting when and why organisms should tradeoff off their own
welfare in favor of another organism.
4
Evolutionary Biological Theories of Valuation and Welfare Tradeoffs
Evolutionary biologists study how the process of natural selection crafts
organisms over deep time. A central problem for early generations of biologists was
the problem of altruism and cooperation. Basically, why ever be nice? It would seem
that genes that created generous organisms would be at a disadvantage relative to
genes that did not. All else equal, organisms—and hence the genes inside them—
replicate faster when they have more resources. Why ever give any of these resources
away?
Despite this, we regularly see animals behaving generously to conspecifics.
To understand the process by which natural selection creates generous behavior I
review in some detail two theories for the evolution of generosity and review several
others more synoptically. The analyses discussed below are generally presented in the
framework of evolutionary game theory as developed by Maynard Smith (1982),
among others. Under this framework, one specifies (1) different heritable strategies
that organisms can play (such as giving to others at a personal cost or aggressing
against intruders in one’s territory), (2) the fitness consequences for those strategies
when they meet themselves or other strategies (what are the fitness impacts for a
strategy for giving when meeting another strategy for giving? when it meets a
strategy that does not give?), and (3) the starting frequencies of various strategies
(some strategies may outcompete other strategies only if already at high frequencies,
but not if there are only a few copies in the population). One can then determine the
long run outcome (does a strategy for giving outcompete and replace other strategies
5
over evolutionarily-relevant timescales?). The ultimate goal of reviewing these
theories is to illustrate how diverse causal processes all lead to the same underlying
logic.
Helping Through Recent Common Descent
Early work by Williams & Williams (1957) and Hamilton (1964) developed
one solution to this problem. Although indiscriminate generosity cannot evolve, one
route for the evolution of discriminate generosity is to selectively give to those who
are likely to share the same gene for giving by recent common descent. This is often
summarized in a mathematical statement that called Hamilton’s Rule, motivated
below.
Hamilton’s rule describes the logic under which an organism should, at a cost
to the self, provide benefits to other organisms as a function of genetic relatedness by
recent common descent. Indeed, Hamilton’s rule was one of the first formal
treatments in a rigorous, modern evolutionary context that describes when and why
organisms should give benefits to another. To understand the logic of Hamilton’s
Rule, assume that organisms have the possibility to give one another a benefit b at
personal cost c. Assume also that benefits are greater than costs, b > c.
Next, imagine that among the entire population a rare mutant has a novel
design that causes it to actually perform this behavior of giving at personal cost.
However, this rare mutant—call it ALLG for Always Give—gives at a personal cost
to any organism it encounters. This, of course, contrasts with the majority of the
population which plays a strategy of never giving—NEVERG. How do ALLG and
6
NEVERG fare against each other? The vast majority of encounters ALLG has will be
with NEVERG. ALLG’s fitness after these encounters will be reduced by c, whereas
NEVERG’s will be increased by b. (When NEVERG encounters itself, it doesn’t
give, so its fitness doesn’t change.) Clearly, ALLG is doing worse. NEVERG will
reproduce at a faster rate and ALLG is likely to disappear from the population.
Do things get better if we assume that, somehow, ALLG has managed to
become the dominant type in the population—that most of the population is ALLG
with a few scattered NEVERG mutants? This does not solve the problem. Most of
NEVERG’s encounters will be with ALLG. As before, in these encounters, NEVERG
does better than ALLG, earning b compared to ALLG’s –c. Most of ALLG’s
encounters will be with other copies of itself, earning it a net of b – c. Notice that
NEVERG is usually earning b and ALLG is usually earning b – c. Obviously b > b –
c. Thus, NEVERG is still doing better than ALLG. Thus, even if ALLG starts out at
the predominant strategy in population, it will eventually be overtaken by NEVERG.
Hamilton’s insight was that strategies for providing benefits to others at a
personal cost can evolve in a population if they find ways to selectively transfer
benefits only to copies of themselves. One way to do this is for the strategy to
selectively transfer benefits to organisms who have copies of it due to recent common
descent. This is usually considered in the context of diploid organisms (e.g.,
mammals, birds, many insects) where each organism has two copies of each
chromosome, one copy inherited maternally and one paternally.
7
Although the logic does not depend on this, it is easiest to see the logic of
recent common descent if by imagining that you have a rare mutation passed on to
you by one of your parents (ignoring for now what that mutation does). This mutation
is so rare that only one of your parents has it (i.e., only one of the chromosomes you
inherited has this rare mutation and it could only have come from one of your
parents). Ignoring some complications (such as genetic imprinting, Haig, 1998) and
assuming simple Mendelian inheritance, you have no way to know which parent you
inherited this rare gene from, so there is an even chance it comes from either parent.
Now, if you have a sibling, what is the chance that they also have this rare
gene? If your mother had the rare gene (50% chance of this), there is a 50% chance
your sibling will inherit it from her. A 50% chance of your mother having it
multiplied by a 50% chance of passing it on to a sibling leads to there being a 25% (=
50% * 50%) chance that your sibling will inherit through your mother the same rare
gene you have. Similar considerations apply when considering inheritance through
your father, meaning there is also a 25% chance that your sibling inherits the gene
from your father. Adding these two probabilities reveals a 50% chance that your
sibling will inherit the same rare gene that you have. Hamilton termed this probability
the coefficient of relatedness, r. For a full sibling, r = 0.5.
What Hamilton showed formally was that if this rare gene caused its carrier to
help individuals who are likely to have the same gene due to recent common descent,
the gene would spread through the population if its help only occurred when r * b > c.
That is, genes that cause organisms to give to others at a personal cost can evolve, but
8
only when it is sufficiently likely that the help will be targeted at other organisms
who also have the gene for helping. Another way of stating this, useful for later
discussions, is that the benefits that will be shared must first be discounted by a factor
that tracks the selectivity of help (here, the coefficient of relatedness); only then can
the (now discounted) benefits be compared to the costs. Yet another way of saying
this, also useful for later, is that the focal organism must translate the fitness benefits
that the other will receive into fitness benefits that the self will receive. Given that
there is uncertainty about whether a full sibling would actually share this rare gene,
the gene that causes an organism to give at a personal cost is not guaranteed to recoup
what it gives; there is only a probability r that the gene will recoup what it gives. Fifty
years later, this rule still holds: Although in modern formalizations the interpretations
of the various symbols in the equation may be somewhat more complicated, the
functional form of r * b > c is still there (Gardner, West, & Barton, 2007). Because
this approach specifies how aid toward kin should evolve it is often called kin
selection theory.
Direct Reciprocity
However, many animals, including humans, provide costly benefits to
individuals who are not closely related to them. What explains this type of
cooperation? Trivers (1971) formalized one solution to this problem, now usually
called direct reciprocity (Trivers preferred reciprocal altruism). Although Trivers’
original description was largely verbal, I will frame this discussion based on the
mathematical version developed by Axelrod & Hamilton (1981).
9
Imagine that pairs of organisms have the opportunity to deliver benefits b to
each other at a personal cost c, where benefits and costs are measured in fitness terms.
If I deliver you a benefit but you do not give me one, then my fitness has been
decreased by c and yours increased by b. If we both give each other benefits, then our
net fitness change is b – c. If neither of us give, then our fitness does not change.
Formally, this arrangement of costs and benefits constitutes a prisoners’ dilemma
(PD). If a PD is played only one time, then the fitness maximizing choice is to not
give—to defect. This is because regardless of what you do, I am always better off not
giving: If you give and I do not, then I earn b compared to the smaller quantity if I do
give of b – c. If you don’t give and I don’t give, I earn 0 compared to the negative
quantity if I do give of –c. In fact, it turns out that if we can chain multiple PDs
together in row, it is still fitness maximizing to always defect—so long as the number
of times we play the PD (i.e., the number of rounds) is fixed and finite. Although one
could imagine some sort of system of threats that could keep organisms “honest”—
strategies that specify giving so along as the other behaves in a certain way—these
ultimately unravel. Consider that no matter how many round are played, there is
always the last round, round t. Given that there will be no more rounds, this becomes
a one-shot PD and the fitness maximizing choice is to defect no matter what. (The
ability to modify behavior based on entering the last round would not require an
explicit representation that round t is the final round. For instance, if an organism
could count the number of rounds, then it could evolve decision rules that use the
round number as input.) Now, given that in round t both organisms are definitely
10
defecting, what happens in round t – 1? This round becomes effectively the last round
of interaction and therefore effectively a one-shot PD. Hence, the fitness maximizing
choice here is to also defect. This logic continues on round after round, implying that
the fitness maximizing strategy when the number of rounds is fixed and finite is to
always defect—whether the number of rounds is 5 or 5 million. (Although this logic
is reminiscent of what economists call backward induction, it is only fitness
accounting, not the rational calculation implied by economic backward induction. Of
course, the two processes would be much more similar if backward induction is taken
to refer to an end-state produced by an unspecified process; in this case, backward
induction would only metaphorically describe the logic leading to the endpoint, not
the equilibration process itself.)
One insight of Axelrod & Hamilton (1981) was that if interactions had
indefinite endpoints—an unknown number of rounds—then there is the possibility for
mutual giving to evolve in and remain in the population. Given that humans are longlived and have mobility, residence patterns, and life spans affected by a number of
stochastic factors, this seems to be a fair characterization of human interactions at
least.
They considered how the strategy Tit-for-Tat (TFT) would perform against a
strategy that Always Defects (ALLD) when interactions have indefinite endpoints.
TFT cooperates on the first round of interactions and then copies whatever its partner
did on the previous round. ALLD, as its name implies, defects no matter what. When
playing a copy of itself, TFT will give at personal cost and receive a benefit each
11
round until the interaction ends. When playing ALLD, TFT will pay a one time cost
of c on the first round, ALLD will receive a one-time benefit of b on the first round,
and thereafter neither will give and thus neither’s fitness will change. When playing
itself, ALLD’s fitness does not change.
By selectively choosing who they interact with across multiple rounds, TFT
earns large fitness benefits when it interacts with copies of itself and only pays a
small cost when interacting with ALLD. Thus, although a specific organism playing
TFT has lower fitness than its partner if its partner plays ALLD, across an entire
population playing these strategies—where TFT will sometimes be playing itself—
TFT will on average earn greater fitness than ALLD, allowing TFT to dominate the
populate.
Axelrod & Hamilton show mathematically that TFT dominates the population
when w * b > c, where w represents the probability that an interaction will continue
from one round to the next. In words, this mean that so long as the length of
interactions is sufficiently long (i.e., w is sufficiently large) and the net within-round
benefits are sufficiently large, then giving at a personal cost is evolutionarily stable.
(Axelrod & Hamilton modeled this using the simplifying assumption that the
population was literally infinite. Under this assumption, rare mutant organisms
playing TFT cannot invade a population otherwise composed of ALLD without there
being some sort of positive assortment—that is, unless TFT is more likely to play
itself than random pairings would predict. However, Nowak, 2006, reviews analytic
work showing that in finite populations with the possibility of genetic drift, even a
12
single TFT mutant, with no possibility of assortment, can eventually lead to TFT
invading the population.)
Although it took a lot of work to get here, the upshot of all this is that
Hamilton’s Rule, r * b > c, and Axelrod and Hamilton’s condition for the evolution of
direct reciprocity, w * b > c, are surprisingly similar. I next review more briefly
several other theories that lead to other similar rules for making welfare tradeoffs.
Reputation & Indirect Reciprocity
Formalizations of direct reciprocity typically assume that agents learn about
their partner’s strategy solely from personal interaction with their partner. However, a
pre-theoretical observation, about humans at least, is that we learn a great deal about
other’s cooperative tendencies by observing them in interactions and from hearing
what others say about them (gossip, reputation, etc., Dunbar, 2004). This can form
another basis for making welfare tradeoffs.
How does reputation-based cooperation evolve? One simple way to formally
model this would be to extend Axelrod & Hamilton’s model. Their model assumes
that an organism interacts with one and only one agent and then dies immediately
after reproducing according to its fitness in that one interaction. One could imagine,
though, that organisms have multiple, serial interaction partners in their lifetime. If
TFT begins its second interaction with knowledge of what its partner’s first round
behavior was in its partner’s previous interaction, then TFT can avoid paying even the
first round cost of providing benefits to ALLD by defecting immediately. After all, if
the strategies under consideration are TFT and ALLD, then all organisms defecting
13
on the first round of their first interaction will be ALLD. Thus, a variant of TFT that
conditions its current behavior on its partner’s past behavior would do better than a
variant of TFT that does not. Even if we make knowledge imperfect—sometimes
TFT knows it partner’s past behavior, sometimes not—it’s hard to see how this could
make cooperation less profitable than in the austere case considered by Axelrod &
Hamilton.
However, reputation is often modeled in a different way, through a paradigm
known as indirect reciprocity (Nowak & Sigmund, 2005; Ohtsuki & Iwasa, 2006;
Panchanathan & Boyd, 2003). In indirect reciprocity, organisms are randomly paired
with each other and only one member of the pair has an opportunity to give at a
personal cost. After this one opportunity, these organisms can never interact again.
Moreover, they can never end up in long distance loops; the possibility of A giving to
B, B giving to C, and C giving to A is ruled out a priori (see Boyd & Richerson,
1989, for a model of network reciprocity like this).
Given these constraints, organisms are paired with multiple partners across a
number of rounds. Except for the first round, organisms who have the opportunity to
give at a personal cost can do so contingent on whether their partner gave in a
previous round. (Note that this requires organisms switching roles across rounds—
sometimes giver, sometimes receiver.) Strategies that cause organisms to give to
other organisms that also give can evolve when q * b > c, where q is the probability
that the recipient’s past behavior is known (Nowak, 2006b; Nowak & Sigmund,
2005). Although there a number of qualifications to this work (Leimar &
14
Hammerstein, 2001; Ohtsuki & Iwasa, 2006; Panchanathan & Boyd, 2003), it’s
striking how this model with its very different—and somewhat esoteric—assumptions
leads to an equation with the same structural form as Hamilton’s rule and direct
reciprocity.
Graph Theory
A more recent tool for modeling social evolution (and its attendant welfare
tradeoffs) is graph theory (Nowak, 2006b). Most modeling approaches assume a
population with no structure: Although there may be some type of assortment, agents
are randomly mixed together with fixed probabilities. Graph Theory goes beyond this
assumption, allowing populations to have a “physical” geometry. This work imagines
that agents are vertices on a lattice. If two vertices are connected by an edge, then
those two organisms can interact; those organisms are neighbors. (Edges also
determine where offspring end up.) Consider a population containing cooperators
who give to each neighbor a benefit b, with each act of giving incurring a personal
cost c, and defectors who do not provide any help. Note that cooperators do not have
a conditional strategy; they always cooperate with their neighbors no matter what.
Given a number of simplifying assumptions, it can be shown that cooperators
outcompete defectors when k-1 * b > c, where k represents the number of neighbors
each organism has; that this equation involves the reciprocal of k indicates that more
neighbors makes it more difficult for cooperation to evolve. The intuition behind this
result is that cooperators flourish if they are in clusters; the smaller the neighborhood,
15
the less likely that cooperator–cooperator interactions will be swamped by
cooperator–defector interactions.
Despite a very different analysis, graph theory leads to a Hamilton’s rule-style
equation for making welfare tradeoffs. Although in the abstract this rule doesn’t seem
to imply any information processing, taking a larger view we can envision an
organism whose social ecology includes many different “graphs,” where some graphs
involve few neighbors and others involve many. Organisms in such an environment
would need some way to discriminate different graphs with different neighbors.
The Asymmetric War of Attrition
The previous models investigated cases where there is some potential for the
alignment of interests. In the PD, both players are better off if both give than if
neither give. However, animals routinely face situations of conflicting interests, such
as situations of zero-sum resource divisions. Moreover, the previous models
considered situations where organisms could deliver benefits to others. Here I
consider a model of conflict where organisms can inflict costs on each other. Despite
these differences, the functional form of the equation describing the selection pressure
is surprisingly similar to those described above.
Hammerstein & Parker (1982) studied this under the rubric of the asymmetric
war of attrition. For exposition, imagine two animals who come across an indivisible
packet of food. The animals fight until one gives up and walks away, leaving the
other with the food. What determines who gives up first? What determines this will
be how the costs of fighting compare to the value of the food. So long as the food
16
brings the animal more fitness than it loses through its cumulative time spent fighting,
it should continue fighting. Once the cumulative costs of fighting are greater than the
benefits provided by the food, the animal should walk away. Thus the war of attrition:
One doesn’t “defeat” one’s opponents so much as wear them down.
One component of the model is the value of the food, which may differ
between animals. The first animal, for instance, may have just eaten and thus their
internal valuation of the food may be relatively low, whereas the second animal may
have not eaten for a number of days, and thus highly value the food. All else equal,
the second animal in this example will fight longer for the food.
The second component of the model is the relative abilities of the animals to
inflict costs on each other and to bear costs, a feature of the animals called
formidability. The first animal, for instance, may be physically stronger than the
second animal. All else equal, the first animal in this example will fight longer for the
food. Mathematically, an animal should walk away first when vself < f * vother, where vi
represents the value of the resources to animal i and f represents the relative
formidability such that f is larger the greater the opponent’s formidability relative to
one’s own. As before, the “benefits” received by another are discounted by a factor
and compared to a “cost” to the self. (If animals can reliably judge their relative
formidability, they can potentially avoid paying any costs by walking away before
fighting begins; see Maynard Smith & Price, 1973.)
As with the other models, this recovers an equation that looks strikingly
similar to the form of Hamilton’s Rule.
17
Multi-Level Selection Theory
The functional forms of the equations described above seem to be a general
principle of social evolution: other-benefiting, personally costly acts can evolve so
long as the benefits given are greater than the costs—after the benefits have been
discounted by a factor that (with the exception of the asymmetric war of attrition)
represents positive assortment—the degree to which generous strategies are paired
with other generous strategies.
This can be seen by looking at the equations derived from multi-level
selection theory (McElreath & Boyd, 2007; Price, 1970, 1972), a very general
accounting system for tracking allele changes over time. To understand this theory, it
is easiest to imagine (a) alleles contained within individual organisms and selection
occurring between individuals (called selection within groups) while simultaneously
viewing (b) these individual organisms as “contained” within groups and considering
selection occurring between groups (called selection between groups).
Will strategies that engage in beneficial, but personally costly behaviors
spread? Yes, so long as σ * βgroup > βpersonal, where βgroup represents the degree of
positive effect that this strategy has on the group, βpersonal represents the degree of
negative effect that this strategy has on the individual, and σ represents the amount of
variation that exists between groups, that is, the extent to which generous organisms
are more likely than chance to be in groups with other generous organisms. (Cases of
mutualism would have βpersonal be not a cost but a benefit. Under most realistic
18
assumptions, for most species, σ is effectively zero so there is no possibility for
between group selection.)
Although the mathematics of multi-level selection are more challenging than
simpler game theoretic approaches designed to deal with specific cases, multi-level
selection in fact subsumes all these other approaches, such as kin selection and direct
reciprocity (Gardner, et al., 2007; Henrich, 2004; D. S. Wilson & Sober, 1994). In
other words, models of social evolution are generally all the same equation just
expressed in different mathematical languages. In an insightful review of these issues,
Henrich (2004) notes that “[a]ll genetic evolutionary explanations to the altruism
dilemma are successful to the degree that they allow natural selection to operate on
statically reliable patterns or regularities in the environment” (p. 4) and that
understanding what these regularities are in any particular case requires knowledge of
“[t]he details of an organism’s social structure, physiology, genome, cognitive
abilities, migration patterns or imitative abilities” (p. 6). Given that all explanations
for generous but costly acts ultimately reduce to the same equation, the real scientific
challenge will be to unpack the statistical regularities that allow the evolution of
generosity and the psychology that operates over these regularities.
Mutualism, Coordination, Pseudo-reciprocity, Markets, and Externalities
Not all cases of non-kin generosity and cooperation represent cases where
there is a possibility of cheating—cases where non-cooperators can even in principle
do better than cooperators. Consider a style of cooperation often called coordination
or mutualism (Clutton-Brock, 2002, 2009). In situations of this structure, benefits are
19
only available to those individuals who choose to cooperate; non-contributors receive
no benefits. Only if we both coordinate our actions can we successfully reap any
benefits: One cannot defect; one can only make an error.
Because there is no selectionist puzzle here, there has not been much
mathematical modeling of this case (but see Pacheco, Santos, Souza, & Skyrms,
2009). Logically, however, there should be a general equation with a similar
functional form, something along the lines of p * b > o, where p represents the
probability that cooperation will be effective (e.g., as a function of each party’s skill
at the task at hand), b represents the benefit obtained by mutual cooperation, and o
represents the “outside option,” the benefit available by not trying to coordinate one’s
actions with others.
Other cases of non-kin cooperation and beneficence may be driven by the
availability of positive externalities (Tooby & Cosmides, 1996); this style of
cooperation has been sometimes been termed pseudo-reciprocity (Connor, 1986).
Externalities represent by-products of one’s actions that impact on others;
importantly, they are not designed consequences, but are incidental outcomes. For
instance, as many plants grow they provide increasing shade. The shade created is not
a design feature of the plant; it is a by-product of increasing leaf size and number.
Humans and other animals routinely take advantage of these externalities to avoid
direct sunlight.
Pseudo-reciprocity occurs when an organism is designed to increase the rate
or quality of externalities another organism delivers. For instance, one could imagine
20
an organism that depends on trees for shade and has adaptations to fertilize trees so as
to create more shade. To the extent that the tree has no way to distinguish situations
where other organisms “purposefully” fertilize it from situations where it was simply
lucky enough to grow up in rich soil, the tree can do nothing to respond to the other
organism’s actions as such. However, it will still grow bigger and provide greater
shade.
Note that there is no possibility for the organism doing the fertilizing or the
organism being fertilized to cheat each other in this situation. The tree cannot take
fertilizer yet refuse to grow bigger; why would it want to? Although the other
organism pays the cost of providing fertilizer, it necessarily gets more shade (and
presumably finds the benefits of shade to outweigh the costs of fertilizer). Thus, an
organism can engage in acts that are both other-benefiting and, ultimately, selfbenefiting without the recipient of their largesse acting contingently. (It is a separate
issue whether other organisms who could enjoy the shade could exploit the organism
doing the fertilizing.)
One up-shot of these theories is that there becomes a market for interaction
partners (Noe & Hammerstein, 1994, 1995). Even if cheating is not possible, there are
still transaction costs. Imagine that an organism provides you with positive
externalities. It takes you some time and energy to reap these externalities and to
increase the rate at which they are delivered. Although you are getting a net fitness
benefit by interacting with this organism, there may be other organisms who would
provide you with even greater fitness benefits. Thus, your current interaction
21
represents an opportunity cost. This suggests that there may be a number of
dimensions that organisms evaluate each other along, such as foraging efficiency or
sexual attractiveness, to choose interaction partners who can deliver the greatest
benefits—even if this benefit-delivery provides no possibility of cheating.
Summary
Fifty years of theory have investigated the selection pressures that can cause
organisms to trade off personal welfare to benefit others. Some of these selection
pressures, such as kin selection or externality-driven altruism, lead to unilateral
helping: The recipient need not respond or even know that aid was given—altruism is
evolutionarily favored regardless. Other selection pressures, such as direct
reciprocity, require contingency: Aid that is given must be eventually returned.
Although such a design might go into temporary fitness “debt,” it must recoup its
losses—and make a profitable return—by the end of its life. Still other selection
pressures, such as the asymmetric war of attrition, explore zero-sum situations where
benefits are extracted through threat of force.
Despite these larges differences in the situations they apply to, all of these
selection pressures lead to equations with remarkably similar functional forms: d * b
> c; the benefits given must outweigh the costs incurred, after the benefits have been
discounted. This discount factor generally represents the relationship between the
donor and the receiver, such as how likely they are to be of the same type or how
their relative formidabilities compare. What consequences do these theories of
selection have for theories of computation and psychology?
22
Computational Mechanisms that Compute and Instantiate Valuation
The previous section discussed a number of selection pressures that lead one
organism to trade off personal welfare in favor of another organism, each of which
generally has the form d * b > c. However, selection pressures describe the abstract
logic by which genes can increase or decrease in frequency in the population; they do
not describe the causal routes through which information relevant to these selection
pressures is detected and acted upon. Consider a hypothetical species of wasp where
females lay their eggs inside figs. Females mate with only one male and only lay eggs
in figs that no other wasps have already used. In this case, a newborn wasp will meet
only other full siblings inside the fig; it does not need any psychological mechanisms
that allow it to distinguish kin from non-kin. Thus, even if kin selection was acting on
this species to affect its behavior toward sibs inside a wasp’s natal fig, there would be
no need for any machinery that could estimate kinship or act differentially toward kin
and non-kin; the world, not the organism’s behavior, creates a statistical regularity
that selection can act on (Tooby & Cosmides, 1989). This is not the case for
humans—long-lived, mobile organisms who interact repeatedly with many
individuals, some kin, some non-kin. If kin selection has crafted mechanisms in the
human mind, one component would be the ability to estimate the kinship status of
others vis-à-vis the self. The goal of this section is to review evidence for
psychological adaptations in the human mind that instantiate the logic of kin selection
and other theories. I review these proposals some detail because (a) they illustrate
23
how valuation can be studied computationally and (b) fleshing out these proposals
reveals a more general problem that they collectively create.
A Kinship Index
The theory of inclusive fitness predicts that organisms should aid individuals
who are likely related by recent common decent. This follows because recent
common descent is likely to cause the strategy for aiding close kin to reside in both
the body of the donor and the body of the recipient. Although humans preferentially
shunt resources toward close-kin over more distant relatives and non-kin (as one
example among many, see Kaplan & Hill, 1985), seeming to support inclusive fitness
theory, what are the psychological mechanisms that allow humans to detect kin, that
is, to estimate r in r * b > c (Kurland & Gaulin, 2005)?
Work spanning over one hundred years suggests that one available
environmental cue that allows for sibling detection in particular is the length of coresidence during childhood: The longer you co-reside with someone, the more likely
they are to be a close sibling. Westermarck (1891) originally hypothesized this cue in
the context of incest avoidance. Given, for instance, that close relatives often have the
same deleterious recessives (rare alleles that cause serious developmental
abnormalities or death when an organism has two copies), mating with close relatives
has a much greater potential to create unfit offspring, creating opportunity costs for
the parents. Thus, Westermarck proposed that to avoid this problem, humans are
designed to develop strong sexual aversions to those who co-resided with us during
childhood. (More recent research has quantitatively matched length of co-residence to
24
the degree of disgust at sibling incest (Fessler & Navarrete, 2004; Lieberman, Tooby,
& Cosmides, 2003).)
Recent theory and data (Lieberman, Tooby, & Cosmides, 2007) suggests,
however, that human kin detection may be more complex. Human foraging societies
have a fission-fusion structure: Although nuclear families often aggregate into large
groupings, nuclear families regularly migrate independently of one another. Thus,
regular co-residence would be a valid cue of siblinghood. However, there may be a
better cue available. First, this theory assumes that humans can know with great
certainty who their biological mothers are, a reasonable assumption given the close
and frequent interactions that human infants have with their mothers. Building on
this, older siblings can observe a cue that Lieberman and colleagues call maternal
perinatal association—the close interaction of one’s mother with a newborn infant.
Just as knowledge of one’s mother’s identity is gained through close interaction with
her during infancy, knowledge of one’s younger sibling’s identity is gained through
observations of close interactions with one’s mother during a sib’s infancy.
For a variety of reasons, maternal perinatal association is likely to be a
stronger cue than co-residence (e.g., a cousin might be co-resident if orphaned). Thus,
Lieberman and colleagues predict and find evidence for the following: First, older
siblings, with respect to their younger siblings, use maternal perinatal association
when it is available as a cue and neglect co-residence in regulating altruism and
incest. Second, because younger siblings do not have access to maternal perinatal
association, they use co-residence to regulate altruism and incest. This evidence
25
reveals mechanisms in the human mind for estimating r; Lieberman and colleagues
call this estimate a psychological kinship index.
Indices Relevant to Reciprocity
The theory of direct reciprocity predicts that organisms should make shortterm fitness sacrifices to aid another organism when there is a sufficient probability
that there will be a long-term net benefit to making those sacrifices. One way to cause
this is for interactions to be sufficiently long. Thus, human reciprocity should be
conditioned on cues to the length of the remaining interaction. Consistent with this, a
great deal of behavioral economic evidence reveals that reciprocity diminishes when
people are informed that their experimental interactions will be short or ending soon
(Camerer, 2003): As the final rounds of interaction approaches, defection becomes
more and more common. Interestingly, this is not an effect of rational updating in
these games. First, although a group of experimental subjects interacting in a
prisoners’ dilemma-style laboratory game will lower their levels of cooperation as the
game approaches the final round, if they start a new game, even with the same
partners, their cooperation will start out high again, only diminishing as they
approach the final round again. Second, it is not the length of the games per se that
leads to a drop in cooperation: Whether a game lasts 10 round or 100 rounds, it is
only within approximately the last 5 to 10 rounds that levels of cooperation begin to
drop.
A related line of research has investigated the effects of anonymity—the
greater the anonymity the less likely that the interaction can effectively continue into
26
the future. Here, too, greater anonymity leads to lower levels of cooperation
(Hoffman, McCabe, & Smith, 1996). Still other evidence, consistent with direct
reciprocity and reputation-based theories, shows that the mind conditions its own
cooperative behavior toward a partner on that partner’s behavior toward others
(Milinski, Semmann, Bakker, & Krambeck, 2001).
Ultimately, however, long interactions can only sustain direct reciprocity
because they allow contingently cooperative strategies to outcompete cheaters: A
cheater might receive a short-term benefit at the expense of a cooperator, but the
contingent cooperator will quickly stop cooperating. When cooperator meets
cooperator, they engage in cycles of mutual cooperation. The length of these cycles is
important: Only if they are sufficiently long can the benefits gained by a contingently
cooperative strategy, when averaged across multiple bodies, outweigh the average
fitness gains made by a cheater strategy that exploits cooperators.
Thus, the theory of direct reciprocity also requires that behavior be
conditioned not just on the shadow of the future but on the behavior of one’s partner.
Cosmides (1989; Cosmides & Tooby, 2005) presents evidence that humans have a
complex psychology that allows for the contingent, mutual exchange of rationed
benefits. This research shows that the mind contains mechanisms that embody a logic
of social exchange that is not reducible to the content-free logics developed by
logicians. One component of this psychology is an ability to detect cheaters on
exchanges—a necessity if one is going to behave contingently in social exchanges.
27
Thus, research also reveals evidence that the mind computes estimates of w and
related factors, what might be called reciprocity indices.
A Formidability Index
The theory of the asymmetric war of attrition predicts that organism will cede
a resource to another organism as a function of their relative formidability. This
implies that the mind has mechanisms for estimating the formidability of others—for
estimating their ability to inflict costs. One component of men’s ability to inflict costs
is strength, particularly upper body strength. Sell, Cosmides, and colleagues (2009)
present evidence that humans can reliably estimate the strength of men from both
men’s bodies and faces. Moreover, people’s judgments of men’s fighting abilities are
essentially identical to their judgments of men’s strength. In other research, Sell,
Tooby, & Cosmides (2009) show that men’s strength regulates how entitled they feel
and their self-reported ability to prevail in conflicts. Aside from personal physical
strength, another potential component of formidability is coalitional strength: the
number and physical strength of one’s allies. Sell, Tooby, & Cosmides (2009) show
that, among a society of hunter-horticulturalists, one’s number of allies regulates
emotions related to entitlement. Taken together, this research suggests that the mind
computes one or more formidability indices that are used to regulate who gets what
during a resource conflict.
Sexual Value Indices
Although kinship, reciprocity, and threat of force are the most commonly
studied routes through which organisms can provide benefits to one another, there are
28
a number of other routes as well (see, e.g., above section on Mutualism). As an
example of this, I develop the possibility of sexual value indices.
Like other mammals, human females energetically invest a great deal in their
offspring (e.g., through gestation, nursing). Moreover, human males, unlike many
other mammals, also invest a great deal of energy provisioning their offspring (e.g.
through hunting of calorically dense meat) (Kaplan, Hill, Lancaster, & Hurtado,
2000). Given that so much is invested in the products of a sexual union, it should not
be surprising that there is a complex psychology that determines sexual valuation.
The complex psychology of human sexual valuation has been one of the most
intensely studied topics within evolutionary psychology. Researchers in this domain
have identified a number of cues that play a pivotal role in determining attractiveness.
These include visual cues of youth and a relatively low waist-to-hip ratio (for men
judging women), cues of status and ambition (for women judging men), and cues of
kindness (especially towards oneself) and of shared beliefs and worldviews (true for
either sex judging the other) (reviewed in Buss, Shackelford, Kirkpatrick, & Larsen,
2001; Sugiyama, 2005). Moreover, cues of close kinship can radically lower the
sexual value of what might otherwise be an attractive partner (Fessler & Navarrete,
2004; Lieberman, et al., 2003, 2007).
Excepting certain cases (e.g., potential sib–sib matings), those with greater
sexual value are generally preferred as sexual partners. Those with high sexual value,
moreover, are valued outside of explicit mateships (Sugiyama, 2005) and receive
better treatment from others in a variety of ways (e.g., Budesheim & DePoala, 1994;
29
Downs & Lyons, 1991; Stewart, 1985). Moreover, attractive women’s minds embody
a recognition of their greater value to others—and hence get angrier more easily and
feel more entitled (Sell, Tooby, et al., 2009). Although the lack of game theoretical
challenges means there is little math modeling regarding sexual value (outside of
models of arbitrary signals (Grafen, 1990)), there is a rich psychology that determines
the sexual value of others and uses it to determine behavior.
The Problem of Integration and the Necessity of Welfare Tradeoff Ratios
The previous section reviewed evidence that the mind has a number of
specific dimensions that it evaluates others on and that it uses these dimensions of
valuation to make decisions that affect both the self and the other. But this presents an
important information processing problem: How are these various dimensions to be
used to make any given decision? Although research often proceeds best when
examining a single factor in isolation (e.g., examining only how variations in
attractiveness affect treatment), things are not so simple in the real world.
As an example, consider an interaction between Joe and Ken. Ken has no cues
of being related to Joe, Ken sometimes returns favors (but sometimes does not), and
Ken is very strong. If Joe has an opportunity to aid Ken, should he? According to
theories of kinship, Joe should not: his kinship index toward Ken is vanishingly
small. According to theories of reciprocity, Joe should be uncertain: his reciprocity
index toward Ken is on a knife-edge that separates cooperation from non-cooperation.
According to theories of conflict, Joe should help Ken: he has a high formidability
index toward Ken. Joe, being a macroscopic physical object, cannot easily be in a
30
superposition of states—he cannot simultaneously help Ken and not help Ken.
Logically, the mind must have some way of integrating together all these various
indices into a single summary index that can be used to make specific decisions in the
here and now.
We call this summary index a welfare tradeoff ratio (WTR). The crux of this
dissertation is that WTRs enter into decisions through a process that is a
generalization of the theories of valuation reviewed above: An organism should give
to a specific other when WTR * b > c; that is, when benefits to the other, after being
discounted by the self’s WTR toward the other, are greater than the costs the self pays
to deliver the benefits. By hypothesis, welfare tradeoff ratios are computed by a
welfare tradeoff function, a function that combines the various indices discussed
above, other indices that remain to be discovered, as well as other situational factors.
The goal of this dissertation is to provide preliminary evidence for the existence of
welfare tradeoff ratios. In the next section, I outline a number of design features that
WTRs are predicted to have.
Standards of Evidence for the Existence of Welfare Tradeoff Ratios
What would it take to show that such an internal regulatory variable exists,
that the mind computes welfare tradeoff ratios? Here I provide several standards of
evidence. First, if such a variable exists, then the precision of its operation should be
evident in the ways that people make welfare tradeoffs regarding a specific other
person. That is, if we observe a person making a number of welfare tradeoffs
regarding a specific other, that person’s responses should be summarizable by a
31
single number—the hypothesized WTR. (This assumes we hold constant situational
factors, such as audience and type or amount of resource at stake.) If a person’s
responses cannot be so summarized that would provide evidence against the view
advanced here. I call this the criterion of consistency.
Such consistency, while expected by welfare tradeoff theory, would be
particularly surprising in light of the many inconsistencies and irrationalities shown in
human judgment and decision making. For example, research shows that people do
not have consistent revealed beliefs (Ellsberg, 1961), that independence of
preferences is often violated (e.g., Oliver, 2003), that preferences can be reversed in
the presence of irrelevant alternatives (e.g., Ohtsubo & Watanabe, 2003), that
preferences are not always transitive (e.g., Tversky, 1969), and that moral preferences
are changed by seemingly irrelevant factors (e.g., the side-effect effect, Leslie,
Knobe, & Cohen, 2006).
The studies described below use controlled, but artificial laboratory
experiments to test whether people’s responses can be summarized by a single
number. Even if there is substantial consistency in people’s responses, what does this
imply? Just as human cognition often appears blinkered when analyzed under
inappropriate assumptions or when tested with inappropriate methods in laboratory
settings (see Gigerenzer, Hell, & Blank, 1988), could the artificiality of the lab lead to
levels of consistency that would not be observed out in the real world? I call this the
criterion of non-artificiality.
32
Related to laboratory artificiality is the issue of “cheap talk.” The studies
presented below use welfare tradeoffs decisions that generally involve hypothetical
sums of money. When real money is not at stake, it is always possible that subjects’
decisions to do not represent their real valuation of the options, that they are engaging
in cheap talk in the service of (e.g.) self-presentation. Thus, it is important to test
whether the responses we assess in welfare tradeoff tasks involving hypotheticals do
not differ from tasks involving real stakes. I call this the criterion of consequentiality.
Welfare tradeoffs ratios are hypothesized to be computed as target-specific
magnitudes. Thus, the mind should assign different WTRs to different social agents in
one’s environment. This follows because different people represent different levels of
formidability, different levels of sexual value, different levels of kinship, etc. I call
this the criterion of target-specificity. It is important to be clear that WTRs do not
represent the sum total of the mind’s knowledge of a specific other. Thus, it is
entirely possible that the mind could distinguish between two different social agents
and still assign them approximately the same welfare tradeoff ratio. Nonetheless,
although it is logically possible that the various people in one’s social network all
ultimately get assigned the same number, in practice the multitude of factors that
should contribute to setting WTRs make this seem unlikely. The theory predicts that
at least some pairs of people are assigned different WTRs.
The first-order problem solved by the existence of welfare tradeoff ratios is
that of making tradeoffs involving the self and another. But a commonly faced
second-order problem is making tradeoffs involving two other individuals. One’s
33
WTR to person A should be a function of a number of upstream factors characterizing
person A, and similarly for person B. How one makes tradeoffs between the welfare
of A and B should be, in part, a function of the same upstream factors. Thus, one’s
WTRs for A and B should predict how one makes tradeoffs between A and B. I call
this the criterion of common effects. Tests that falsify this would provide serious
obstacles to the view advanced here.
WTRs are hypothesized to be internal magnitudes that are used to regulate
behavior. They should therefore behave like other psychological magnitudes. I call
this the criterion of magnitude isofunctionality. This can be testing by comparing the
consistency of decisions that involve welfare tradeoffs to the consistency of decisions
that necessarily involve the psychological manipulation of other types of magnitudes.
However, WTRs are hypothesized to be magnitudes computed by specialized
algorithms (where the specialization of the algorithms derives from the specialized
tasks they are designed to accomplish). Thus, there should be ways in which
computation involving WTRs differs from the use of other magnitudes. I call this the
criterion of algorithmic separability.
The Problem of Estimation
I have outlined a number of reasons to think that the mind computes welfare
tradeoff ratios that regulate one’s social decision making: The mind computes a
number of specific indices, all of which follow the same abstract logic, but the mind
must also integrate these specific indices into a downstream index to actually regulate
behavior. If this is true, it implies that not only will the mind assess multiple factors
34
about others and regulate behavior toward them as a function of these factors, but also
that others will be doing the same. Thus, the WTRs that others compute will have
consequences for both the self directly and, more generally, for the self’s social
world.
This implies that mind may also contain mechanisms that can estimate the
magnitude of WTRs of in others’ minds. Such an ability would be beneficial because
it would allow one to know who has a high WTR toward the self, a person worth
allocating time and energy toward (Tooby & Cosmides, 1996), and who has a low
WTR toward the self, a person who might profitably be ignored or even avoided.
Moreover, knowing the WTRs of others can allow one to manipulate those WTRs.
For instance, Sell, Tooby, & Cosmides (2009) provide evidence that the emotion of
anger is designed to raise other’s WTRs when they are lower than what the mind
calculates they should be. This necessarily involves an ability to estimate the welfare
tradeoff ratios set by others toward the self.
Standards of Evidence that the Mind Estimates Welfare Tradeoff Ratios
The mind should allocate in favor of another when the benefit the other
receives is greater than the cost the self incurs, but only after the benefit to the other
has been discounted by the welfare tradeoff ratio toward the other. Mathematically,
allocate to the other when WTR * b > c.
Although welfare tradeoff ratios are assumed to be purely internal
magnitudes—there are no external physical objects or actions that directly correspond
to a WTR—I assume that the delivery of benefits and the imposition of costs are
35
(often) observable. That is, I assume that the mind contains procedures that can
translate objects, actions, and effects in the real world (e.g., a bite of an apple, a
punch in the solar plexus) into an internal numerical scale (Tooby, Cosmides, &
Barrett, 2005). Such an ability would be necessary for even a solitary forager who
must decide whether this or that patch is more profitable to forage in. (However,
because it is beyond the scope of this investigation, I am ignoring the fact that what
counts as an object, an action, etc. is not a feature of the objective world, but is a
feature of the organism’s psychology.)
Given this, the mind can infer (boundaries) on another’s WTR by observing
the benefits it delivers and consumes and the costs it incurs and inflicts. By
rearranging, we see that the mind should allocate to another when WTR > c / b. Thus,
if a benefit is given at a personal cost, one can infer that the WTR of the agent
delivering the benefit is at least as great as the cost–benefit ratio. Holding benefits
constant, greater costs incurred imply greater WTRs. Thus the mind should infer
differences in the lower bounds of welfare tradeoff ratios as a function of differences
in costs incurred. This prediction is tested in Chapter 3. That chapter also conducts
further experiments to distinguish predictions from welfare tradeoff theory from
rational choice and payoff-based theories.
Summary of Predictions from Welfare Tradeoff Theory
According to welfare tradeoff theory, the mind needs a number of
computational abilities. One is the ability to compute a downstream variable—a
welfare tradeoff ratio—that determines how personal welfare is traded off against the
36
welfare of another. I have outlined multiple standards of evidence for demonstrating
the existence of this variable and test for its existence in Chapter 2. A second ability
is to estimate the welfare tradeoff ratios that exist in the minds of others. I have also
outlined standards of evidence for this ability and test for it in Chapter 3. The two
chapters use very different methods to provide convergent validity for the general
concept of welfare tradeoff ratios. The confirmation of predictions in both chapters
would provide strong support for welfare tradeoff theory.
37
Chapter 2: The Existence of Welfare Tradeoff Ratios1
The Welfare Tradeoff Task
Theoretical Overview
The goal of this chapter is to provide evidence consistent with the hypothesis
that the mind computes a magnitude used to regulate welfare allocations—a welfare
tradeoff ratio. WTRs are predicted to be fairly precise magnitudes. Thus, to assess
this prediction, we need an appropriate measurement device. This rules out many
standard measurement instruments, such as arbitrary numbers attached to ad hoc
rating scales. It’s difficult to see, moreover, how the basic hypothesis about the
existence of welfare tradeoff ratios can be tested using standard social cognition
methods such as reaction times based measures (e.g., the implicit associations test).
Instead, this research borrows methods developed by cognitive, behavioral, and
economic psychology to study other internal regulatory variables. Considering this
research program in some detail will help draw out the logic and assumptions of the
approach.
Considerable effort in cognitive psychology has been devoted to
understanding how the mind makes intertemporal tradeoffs (Green & Myerson,
2004). Typically, this research asks the question of whether a smaller, earlier benefit
would be preferred over a larger, later benefit. (Often, the smaller benefit would be
available immediately if chosen.) An assumption of this work is that, all else equal,
1
Theresa Robertson, Daniel Sznycer, and Julian Lim all collaborated on the research described in this
chapter.
38
the longer a benefit is delayed into the future, the less subjective value it will have
(e.g., $5 today is worth more than $5 in ten years). This would happen, for example,
if the mind embodies the fact that at greater and greater delays, the future is less and
less likely to be reached (Williams, 1957; M. Wilson & Daly, 2004). Thus there is a
tradeoff: Although the delayed benefit is absolutely larger than the immediate benefit,
its delay lowers its value, potentially making it less valuable than the absolutely
smaller, but immediate, benefit. A second assumption is that animal minds cannot
determine which benefit to choose unless it can put them on a common scale, a scale
that incorporates both the amount of benefit available and the degree of delay. To do
this, the mind must discount the value available in the future so that it can be stated in
terms of present, immediately available benefits. This is accomplished by plugging
future benefits into a discount function, which research often shows to have the form:
computed present value = nominal value / (1 + k * delay), where k is an individual
difference parameter that determines how much the future is discounted. (There is not
yet widespread agreement for why future discounting takes this form, called a
hyperbolic function, but see Green & Myerson (2004) and Killeen (2009) for possible
rationalizations.)
Consider the choice between $10 available immediately and $20 available in
two weeks. One could imagine a mind that uses one or the other factor alone to make
this choice. Whatever your preference, however, prior research indicates that it would
change if either one of the amounts or the delay were sufficiently modified. Thus, the
mind does not consider the absolute amounts or the delay alone; the decision must be
39
made with some sort of process that combines both. One could imagine a heuristic
process that evaluates one attribute first, only moving on to the other if the first
evaluation fails (see Brandstatter, Gigerenzer, & Hertwig (2006) for such a proposal
in the related domain of probability discounting). For instance, a simple heuristic
might be “If one benefit is at least an order of magnitude larger than the other, choose
that benefit. If not, then choose the benefit that has a shorter delay, so long as the
delay is at least one month shorter. If this cannot be satisfied, then choose randomly.”
Such a heuristic does not explicitly recognize that there is tradeoff to be made
between an earlier, small benefit and a later, large benefit: The same basic heuristic
(with different content) could equally well be applied to a situation where (a) a set of
cues are all positively correlated with some outcome and each other, (b) some cues
have stronger predictive ability such that if you have these cues available there is
nothing further to be gained by looking at additional cues, and (c) there is imprecision
in estimates of all cues such that very close values cannot reasonably be distinguished
from each other. In such a situation, the mind should start with the most predictive
cue, discard it if it yields an ambiguous determination, move on to the next most
predictive cue, and so on.
Could a simple heuristic such as this be applied to temporal discounting?
Probably not: the complexity and nuance of people’s intuitions regarding temporal
discounting speaks against such an account (Green & Myerson, 2004), as does recent
neuroscience evidence (Kable & Glimcher, 2007; Preuschoff, Bossaerts, & Quartz,
2006). Moreover, despite recent proposals, simple heuristics such as this do not seem
40
to sufficiently account for people’s decisions on gambling-like choice tasks (tasks
involving risk; see commentators on Brandstatter, Gigerenzer, & Hertwig (2006):
(Birnbaum, 2008a; Johnson, Schulte-Mecklenbeck, & Willemsen, 2008; Rieger &
Wang, 2008)). Thus, to make the above tradeoff, the mind must first convert $20
available in two weeks to an equivalent amount available now. It must then compare
this computed amount to $10 available now and choose the larger. Whether the
computed amount is greater or less than $10 will depend on how strongly the mind
discounts the future.
This style of analysis can also be applied to welfare allocations. First, is there
a potential for welfare allocations to involve tradeoffs? One factor determining
welfare allocations is the targets of the allocation: If all else is held equal (and
notwithstanding a few important exceptions), $5 in my hands is more valuable to me
than $5 in your hands. But another factor is the relative size of the allocations:
Diminishing marginal returns ensures that many situations exist in which a large
benefit can be delivered to another at only a small cost to oneself (e.g., a successful
hunter who has eaten his fill could share leftover meat with a hunter who has not been
successful). This means that decision-makers will sometimes find themselves facing a
choice between allocating a smaller benefit to themselves or allocating a larger
benefit to another. (In many cases, the “benefit” allocated to the self is avoiding the
cost that would have been incurred to allocate a benefit to another.) Because the two
factors point in opposite directions, there is a tradeoff—e.g., do I allocate $10 to
myself or $20 to a friend?
41
Second, is there reason to think that a mathematical function (as opposed to
some sort of simpler heuristic) is necessary to solve this problem? Suggestive
evidence comes from an investigation on how (a) the potential cost of helping another
and (b) the relationship of the potential helper to the recipient combine to determine
the help given (Stewart-Williams, 2007). In this study, the benefits of helping were
perfectly correlated with costs of helping; more help means more cost. For kin
(siblings and cousins), subjects were increasingly likely to help as the benefits of help
increases. For non-kin (friends and acquaintances), subjects were decreasingly likely
to help as the benefits of help increased. This pattern could only occur if both kinship
and cost of helping had simultaneous, functional consequences during at least some
steps of the decision making process. Although the evidence here is consistent with
the necessity of a mathematical function, the dependent measures were ad hoc rating
scales asking about difficult-to-quantify behaviors.
More direct evidence comes from research using methods from the
discounting literature in cognitive psychology (Jones & Rachlin, 2006, 2009; Rachlin
& Jones, 2008a, 2008b). These researchers analogized from the existence of temporal
discounting to hypothesize that there exists social discounting. Just as one can use
knowledge of temporal delay and a person’s temporal discounting parameter (k in the
equation above) to compute the subjective value delayed benefits would have if they
were instead available immediately, so, too, can one use knowledge of social distance
(the analogue of temporal delay) and a person’s social discounting parameter (the
analogue of k) to compute the subjective value that benefits accruing to another
42
person would have if they instead accrued to the self. Thus, these researchers asked
subjects to “imagine that you have made list of the 100 people closest to you in the
world ranging from your dearest friend or relative at position #1 to a mere
acquaintance at #100 (Jones & Rachlin, 2006). The person at number one would be
someone you know well and is your closest friend or relative. The person at #100
might be someone you recognize and encounter but perhaps you may not even know
their name. You do not have to physically create the list—just imagine that you have
done so” (p. 284). The position of another person on this list represents social
distance. The researchers next asked subjects to make a series of monetary decisions
involving tradeoffs. In these decisions, subjects chose between Allocation A—
allocating a large sum of money to themselves and nothing to a specific other—or
Allocation B—allocating a smaller sum to themselves and a sum equal to this smaller
sum to the specific other. All subjects answered regarding multiple specific others:
the people occupying the positions 1, 2, 5, 10, 20, 50, & 100 on their lists. Note,
however, that subjects did not actually generate any such list and were never asked to
specify the specific people occupying even the few positions they were asked about.
Thus, subjects were asked about “specific others” only in the sense of there being
specific numbers, not in the sense of there being specific, identifiable people.
The results showed a quite surprising effect: Subjects had no difficulty doing
this task and the mathematical function best describing how they made interpersonal
tradeoffs was the same form as the equation given above for intertemporal tradeoffs.
This form, moreover, also describes how people make probabilistic tradeoffs (e.g., do
43
you want $10 for sure or a 60% chance at earning $20?). (See Green & Myerson,
2004, on how intertemporal and probabilistic tradeoffs are instantiated by different
mechanisms.) Despite this similarity, all three types of tradeoffs involve functions
that are not economically rational—market forces should create exponential
discounting functions, yet all three of these functions are hyperbolic discounting
functions (Green & Myerson, 2004; Killeen, 2009). This helps rule out the possibility
that all of this is somehow the product of general rationality, instead of specialized
computational systems. Altogether, these results are consistent with the need to model
welfare allocations—and the resulting tradeoffs—as being the output of a
psychological mechanism that mathematically integrates the amounts being allocated
and the nature of the targets of the allocation.
Although I will be using a task similar to the social discounting task of Jones
& Rachlin (2006), it is important to examine the theoretical and methodological
differences between our approaches. First, Jones & Rachlin suggest that there is
unitary dimension that determines choices—social distance—along which people can
be arrayed in one’s mind. Although they do not commit themselves to a definition of
social distance, they suggest that social distance may be a product of common or
shared interests (p. 285). This is reflected in the instructions they use (“the 100 people
closest to you”).
Although welfare tradeoff theory also proposes that there is a single
downstream variable (the welfare tradeoff ratio) that ultimately determines people’s
choices, there are multiple, principled determinants of welfare tradeoff ratios. A
44
common interest is not the only, or necessarily the primary, determinant: Some
relationships may be regulated almost entirely by relative formidability, with a less
formidable person assigning a more formidable person a very high WTR despite a
complete lack of psychological closeness. Other WTRs may be mostly set by the
ability to confer benefits, such as setting a high WTR toward a person who is
attractive or has high status—in what sense is valuing someone due to their
attractiveness or status synonymous with common interests or closeness? Indeed, on
welfare tradeoff theory it is difficult to create a linear ordering of targets because the
welfare tradeoff function integrates multiple factors—which can “contradict” one
another—and these factors are dynamically updated. (Even estimates of kinship can
be updated into adulthood: E.g., if a man learns his wife is currently engaging in
extra-pair mating, his mind may lower its estimate of his paternity for previous
children she has borne.) Thus, welfare tradeoff theory provides a more nuanced view
of the determinants of social discounting.
Second, Jones & Rachlin’s assumption of a unitary dimension of social
distance motivated their empirical strategy of asking about unspecified others defined
by numbers on an imaginary list. Welfare tradeoff theory focuses on the personspecific determinants of welfare tradeoff ratios. Without specifying a specific other, it
is unclear exactly what is being tapped by these decision questions—are subjects
spontaneously filling in specific others, resorting to “default” settings, or something
else? Indeed, Stewart-Williams’ (2007) data showing that friends are less likely to
receive help as it becomes more costly, but kin more so, shows how important it is to
45
take into account person-specific factors, instead of a general dimension of closeness.
Nonetheless, Jones & Rachlin’s data provides converging evidence that the mind
computes relatively precise internal magnitudes that regulate welfare allocations.
The Task
The welfare tradeoff task is designed to measure how much personal welfare
the mind will tradeoff to enhance the welfare of a specific other. To do this, the task
involves money. Using money has several benefits: First, it allows for easy and
precise numerical quantification, allowing a very stringent test of the hypothesis that
these tradeoffs are regulated by decision rules that consult a variable that has a
magnitude. Second, it allows for goods available to the self and for goods available to
others to be in the same currency. Third, previous works shows that monetary tasks
provide orderly and replicable results for domains, such as temporal discounting (a
domain which replicates with other species and currencies, such as pigeons and food).
There are also drawbacks: Money is evolutionarily novel. Ancestrally, many acts of
help and generosity—such as childcare or food sharing—involve currencies that are
not fungible. Money, on the other hand, is completely fungible. (In Chapter 4 I
outline one way to investigate welfare tradeoffs without directly using money.)
Inspired both by temporal and social discounting tasks (Green & Myerson,
2004; Jones & Rachlin, 2006), the task I created requires subjects to decide whether
(a) the other would receive a sum of money and the subject would receive nothing or
(b) the subject would receive a different, usually smaller sum of money and the
specific other nothing. A simply put example: Will you allocate $5 to yourself or will
46
you allocate $10 to your friend? The key to estimating WTRs lies in having subjects
complete multiple decisions regarding a specific other where the ratio of the values is
manipulated. Considering our simple example, if you prefer to allocate to your friend,
that implies that $5 < WTR * $10. That is, even after discounting the $10 due to the
fact your friend receives it instead of you, it is still a better outcome to you than
directly receiving $5. Simple algebraic rearrangement implies that your WTR toward
your friend is at least 0.5. What if we give you another decision, this time $7 for you
or $10 for your friend, and you now choose to allocate to yourself? This implies that
your WTR is no more than 0.7. By varying the ratios, we now have an upper and
lower bound for your WTR toward your friend; simple arithmetic averaging estimates
your WTR toward your friend as 0.6. The actual instrument uses this strategy, but
with a larger range of possible ratios, as well as multiple decisions with the same
ratio.
In the experiments presented below, for each specific other, a subject
completes six anchored sets, each set with ten separate decisions. Thus, for a given
target, subjects make sixty separate decisions. I call the sets “anchored” because
within a set the amount that the other could be allocated is fixed—it provides an
anchor for the variable sums that could be allocated to the self. This constant
anchoring amount for the other was 19, 23, 37, 46, 68, or 75. The amount for the
subject was determined by multiplying the anchor by different, evenly-spaced
ratios—-0.35, -0.15, 0.05, 0.25, 0.45, 0.65, 0.85, 1.05, 1.25, and 1.45—and then
rounding to the nearest integer. In principle, each ratio represents the WTR of a
47
subject who is perfectly indifferent between the two choices; the ratios are
“indifference WTRs.” Figure 1 shows some of the decisions created by this
algorithm. When making these decisions, subjects are asked to assume that each
choice is made independently and to let other choices affects their decisions. They are
also asked to assume that the money cannot be shared between themselves and the
targets of their choices (e.g., if they allocate to a friend, that friend cannot then give
them a portion of that allocation).
As an example, consider the decision between $75 for the other and $0 for the
self or $0 for the other and $49 for the self; the ratio of these values is 49 / 75 = 0.65.
If a subject were perfectly indifferent between these two options, then the subject’s
WTR for the other is simply the ratio of the two sums, .65. People, however, are not
indifferent and usually prefer one over the other. In this example, a subject may prefer
to receive 49 instead of allocating 75 to the other. But the same subject may prefer to
allocate 75 to the other instead of receiving 34 (where 34 = 0.45 * 75). Thus, one can
infer that the subject’s WTR with respect to the other is less than 0.65 but greater than
0.45. A numerical estimate can be obtained by averaging these two bounds together:
(0.65 + 0.45)/2 = 0.55. One way of putting this is that by giving subjects a series of
decisions one can look for a switch point. These switch points represent subjects’
WTRs. By giving subjects multiple anchored sets, the reliability of the measure is
increased.
However, subjects are not always perfectly consistent such that there is one
and only one switch point (or no switch points) in a given question. Therefore, to
48
estimate WTRs in the absence of perfect consistency, I employ a method developed
by Kirby & Maraković (1996). In this method, each possible WTR is computed by
averaging together each successive pair of the ratios listed above (-.35 to 1.45). This
gives possible WTRs between -.25 and 1.35 inclusive, with increments of .2. For
simplicity, I assume that if subjects always allocate to themselves in an anchored set
then that reflects a WTR of -.45 (the average of -0.35 and the value -0.55 that would
be present if the instrument continued downward in similar increments), and if
subjects always allocate to the other, then that reflects a WTR of 1.55 (the average of
1.45 and 1.65 if the instrument continued upward in similar increments). This gives a
total of 11 possible WTRs that could be assigned to a single anchored set.
Within an anchored set, for each possible WTR, the number of choices
consistent with that possible WTR is calculated. The possible WTR that has the
greatest consistency is then assigned as the actual WTR for that anchored set. (It is
the best-fitting WTR.) If two possible WTRs have equal consistency, then they are
averaged together. Finally, the WTRs assigned to each anchored set are averaged
together to give an estimated WTR for a participant.
49
You Other
A
Option 1 54
0
Option 2
0
37
You Other
B
Option 1 15
0
Option 2
0
23
You Other
Option 1 46
0
Option 2
0
37
You Other
Option 1
4
0
Option 2
0
75
You Other
Option 1 39
0
Option 2
0
37
You Other
Option 1 -3
0
Option 2
0
19
You Other
Option 1 31
0
Option 2
0
37
You Other
Option 1 71
0
Option 2
0
68
You Other
Option 1 24
0
Option 2
0
37
You Other
Option 1 31
0
Option 2
0
37
Option 1
Option 2
You Other
17
0
0
37
Option 1
Option 2
You Other
21
0
0
46
Option 1
Option 2
You Other
9
0
0
37
Option 1
Option 2
You Other
54
0
0
37
You Other
Option 1
2
0
Option 2
0
37
You Other
Option 1 44
0
Option 2
0
68
You Other
Option 1 -6
0
Option 2
0
37
You Other
Option 1 -6
0
Option 2
0
37
Figure 1. Examples of the choice task. The left pane (A) shows an example of the
Ordered condition and the right pane (B) shows an example of the Random condition
in Studies 1 & 2.
50
I calculate two additional dependent measures from these data. The first is a
perfect consistency score. For each anchored set, I determine whether there is one and
only one switch point (or no switch points at all for the lowest and highest possible
WTR values). Thus, for a specific target, subjects can have between 0 and 6 perfectly
consistent anchored sets. The final perfect consistency score is this count divided by 6
to create a percentage (e.g., if a subject had 5 perfectly consistent anchored sets for a
given target, her perfect consistency score would be 5/6 or 83.3%). Obviously,
because the perfect consistency score requires perfect consistency, it is susceptible to
even minor lapses in attention. The second additional measure, a less strict
consistency measure, is a consistency maximization score. For a given anchored set,
this is the percentage of decisions consistent with the WTR that is most consistent
with the decisions in that set. Thus, if the best-fitting WTR has only 7 decisions
consistent with it, then the consistency maximization score for that anchored set is 7 /
10 = 70%.
It is important to note that WTRs, perfect consistency scores, and consistency
maximization scores can be all calculated separately for each anchored set. Thus,
each of these dependent measures is computed six times for each target and, within a
dependent measure, they can then be aggregated together in various ways depending
on the purpose of the analysis.
Does human performance conform to the standards evidence described above?
Specifically, does performance conform to the criteria of:
51
1. Consistency: Are welfare tradeoffs made consistently? Are decisions
summarizable by a single number?
2. Non-artificiality: If welfare tradeoffs are made consistently is this due to the
artificiality of the lab task or to real abilities of the mind?
3. Consequentiality: Are the revealed welfare tradeoff ratios due to cheap talk or
real valuation?
4. Target-specificity: Are welfare tradeoff ratios target-specific?
5. Common effects: Do WTRs toward two others—which regulate how the self’s
welfare is traded off to benefit those others—correlate with how tradeoffs
involving only those two others are made?
6. Magnitude Isofunctionality: Do WTRs behave as other psychological
magnitudes?
7. Algorithmic Separability: Are welfare tradeoff computations performed by
specialized algorithms?
Because many of these criteria were tested across multiple studies, I first provide
details of the methods of each study and then discuss the results organized by the
hypotheses. In addition to the above criteria, which were designed specifically to test
for the existence of an internal magnitude, I also test whether welfare tradeoff ratios
correlate with standard self-report measures of valuation. This provides convergent
evidence that the welfare tradeoff task is tapping what it is theoretically predicted to
tap.
52
Methods
General Design
The general design of all studies in this chapter was relatively straightforward:
With respect to one or more targets, subjects completed the 60 decisions that
comprised the WTR task. For instance, in Studies 1 and 2, subjects completed the 60
decisions about a same-sex friend and also completed the 60 decisions about a samesex acquaintance. After making these decisions, in most studies subjects next
completed additional questions. In the remainder of this section I give the details of
the specific studies.
Subjects and Specific Design
Study 1: Measuring WTRs in the United States. One hundred sixty-seven
subjects (90 female) at the University of California, Santa Barbara, completed the
welfare tradeoff ratio task as detailed above, once each for a close friend and for an
acquaintance. The task was in pencil-and-paper format. The order of answering about
the friend and acquaintance was randomized across subjects. Subject were also
randomly assigned to one of two between-subjects conditions. In the Ordered
condition, the decisions were presented as in Figure 1’s left-hand side such that all
decisions with the same anchored value for the other were grouped as a set and within
each of these sets the value for the self was titrated up or down. In the Random
condition, the same sixty decisions were presented in a random order (which was held
constant across subjects). After making all of the welfare tradeoff decisions regarding
both their friend and acquaintance, subjects then answered several additional
53
dependent measures: the social value orientation, the communal strength scale, the
inclusion of other in self scale, and a relationship quality index (the later scale was an
ad hoc scale developed specifically for this research). Appendix 1 presents the
welfare tradeoff component of Study 1 in full, including all instructions. It also
contains the relationship quality index.
Study 2: Measuring WTRs in Argentina. The sample for Study 2 consisted
of 125 students (78 female) from the University of Buenos Aires in Argentina. The
details were otherwise identical to Study 1 except for the following: In this study, the
instruments were translated into Spanish by a native Spanish speaker, one of the
collaborators on this project (Daniel Sznycer). We did not employ any backtranslation or verification techniques.
Study 3: Measuring WTRs when real dollars are at stake. One hundred
forty-seven subjects (93 female) from UCSB completed a computerized version of
the WTR task as Study 3. The same set of decisions used in Studies 1 and 2 were
employed. Also as in Studies 1 and 2, subjects responded about both a friend and an
acquaintance; which came first was randomized. Because of the computerized format
of Study 3, however, subjects only saw a single decision at a time and could not
revisit or review a decision once it was made. Although the randomized version in
Studies 1 & 2 was a single randomization used across all subjects, in Study 3 each
subject experienced a unique randomization of all decisions within a target.
Moreover, the computer displayed the friend’s or the acquaintance’s initials in place
54
of “Other” as shown in the examples of the task in Figure 1. Screen shots of the
computerized tasks used in Studies 3-5 are shown in Appendix 2.
In Studies 1 and 2, decisions were purely hypothetical. In this study, however,
half of the subjects could potentially earn real money for themselves or allocate real
money to a friend or acquaintance based on their decisions on the WTR task. To do
this, this study used an incentive compatible payoff mechanism. Subjects in this
condition were told that after they completed the task they would have the
opportunity to roll two dice. If they rolled double sixes, then two things would
happen. First, the subject would receive an additional $30. Second, they could
randomly select one of their decisions (i.e. blind to the values or their choice on that
decision) and that decision would be made real. If that decision gave money to the
subject, they would receive a check in the mail for that sum plus $30. If that decision
gave money to the friend or acquaintance, the subject would receive a $30 check in
the mail and the friend or acquaintance would receive a check in mail for the sum that
the decision dictated be allocated to the friend or acquaintance. The initial sum of $30
was chosen so that no subject would pay out-of-pocket to participate in the
experiment (there are decisions that could cause the subject to lose up to $27). After
learning about the payoff mechanism and being thoroughly quizzed to ensure
understanding, subjects then completed the WTR task for both a friend and an
acquaintance. To avoid violations of the privacy of the friends and acquaintances,
subjects were asked to select same-sex friends and acquaintances with the constraint
that these people appeared in a public phone book that was provided. In practice, only
55
1 subject in this condition rolled double sixes and the choice dictated giving money to
the subject. After completing their decisions, but before finishing with the payoff
mechanism, subjects completed the same additional dependent measures as in Studies
1 & 2.
In the other half of Study 3’s sample, subjects did not have the opportunity for
their decisions to become real. However, they did have the opportunity to roll two
dice at the end of the experiment. If they rolled double sixes, they would receive $30
in the mail. Subjects were also quizzed prior to engaging in the task to ensure they
understood the payoff mechanism. All other aspects were identical between the two
conditions.
In both conditions, subjects chose their friend and acquaintance prior to
learning the payoff mechanism; thus, whether money was a stake could not affect the
targets subjects chose. In both conditions, all subjects received $10 in cash as a showup fee.
The instructions given verbally by the experimenter to subjects are shown in
Appendix 3. Once subjects began the computerized task, instructions were as in
Studies 1 and 2.
Study 4: Decisions involving two other individuals. Study 4 consisted of 82
subjects (48 female) from UCSB, who were paid for their participation. Subjects
again completed a computerized version as in Study 3. The program asked subjects to
choose six individuals with whom they had a range of relationships. See Appendix 4
for complete instructions. The program then randomly selected two of these
56
individuals. This was done to get a broader range of targets. At the end of the survey,
subjects were asked to choose which of the two they felt closer to. To simplify
exposition, in Study 4 I call the closer person “friend” and the more distant person
“acquaintance.” When subjects were filling out the WTR tasks, these individuals
were specified by their initials.
Subjects completed the WTR task with four different types of content: (1)
Tradeoffs involving themselves and their friend. (2) Tradeoffs involving themselves
and their acquaintance. (3) Tradeoffs between their friend and acquaintance with
acquaintance in the “anchored” role. (4) Tradeoffs between their friend and
acquaintance with friend in the “anchored” role. The order that subjects completed
these four formats in was randomized. (1) & (2) provide estimates of standard WTRs.
(3) & (4) provide numerical estimates of the tradeoff between third-parties; these
third-party tradeoffs should be a function of standard WTRs. (Subjects did not
complete any additional dependent measures in this study.)
Study 5: Comparing logically identical tasks. Study 5 consisted of 112
subjects (78 female) from UCSB who were paid for their participation. Subjects in
this sample completed the WTR task for a friend or an acquaintance, not both. They
also completed two versions of a task that asked them to compare sums of money
expressed as dollars and Euros and determine which was larger. These tasks were
computerized and randomized as in Studies 3 and 4.
Instead of using the actual exchange rate, subjects were given a hypothetical
exchange rate that allowed them to convert Euros to dollars. These hypothetical
57
exchange rates were drawn from the empirically determined WTRs in Study 1. Thus,
the exchange rate task had the same logical structure as the WTR task: compare two
numbers after discounting one by a specific number. Each subject completed one
version of the exchange task with the exchange rate presented as a decimal and
another with a different exchange rate presented as a fraction. Whether the first
exchange task presented the rate as a decimal or fraction was randomized across
subjects. Full instructions are given in Appendix 5. Subjects did not complete the
same additional dependent measures as in Studies 1-3, but did complete a different,
smaller set of additional measures described in the next section.
Dependent Measures
Welfare tradeoff task. In all studies, subjects completed the welfare tradeoff
task described above and illustrated in Figure 1. The full ordered version of the task
as employed in Study 1 can be seen in Appendix 1.
Exchange rate tasks and supplemental questions. Each subject in Study 5
completed two tasks that were mathematically identical to the WTR task. Both tasks
asked subjects to pick which of two sums of currency represented a larger value. One
sum was expressed as US dollars, the other as Euros. Each task required subjects to
make sixty decisions. The numbers used in these decisions were identical to the
numbers used in the WTR task; the sums expressed in Euros corresponded to the
anchored values in the anchored WTR sets.
Unlike the WTR task, the exchange rate tasks required us to give subjects a
tradeoff value to use in their decisions—an exchange rate that allowed them to
58
convert Euros into dollars. One version of the task gave subjects the exchange rate as
a decimal; the other as a fraction. As exchange rates, I used the WTRs assigned to
subjects in the unordered condition of Study 1. (Although subjects in Study 1 did not
complete the computerized version of the task, the procedure was otherwise the most
similar to Sample 5.) I used WTRs from a previous sample for two reasons. First,
although subjects in Sample 5 will not be using self-generated numbers in the
exchange rate task, they will still be using numbers that other subjects were able to
use successfully in making decisions. Second, by using WTRs as exchange rates I can
further reduce any differences between the exchange rates tasks and the WTR task.
There were two exceptions to using Sample 1’s WTRs as exchange rates:
First, I excluded negative WTRs because an exchange rate cannot be negative.
Second, if the WTR from Sample 1 was between -.05 and 0 inclusive, I recoded the
WTR as +.01. Given our scoring algorithm, -.05 is one possible WTR that could be
assigned to an anchored set and, for consistency, we assigned that WTR to subjects
with the appropriate switch point. However, it is more realistic to expect that subjects
with this switch point simply put zero weight on the other person’s welfare. Because
my goal with Sample 5 was to use estimated WTRs as exchange rates, I elected to
convert these to slightly positive numbers.
After making these exclusions and conversions, and rounding each revealed
WTR to two decimal places, I then created two multinomial distributions, one
composed of WTRs for friends and one of WTRs for acquaintances. The probability
associated with each WTR in these distributions was the proportion of subjects
59
having that WTR in Sample 1. For each subject in Sample 5, the computer randomly
drew two numbers from the friend distribution or two numbers from the acquaintance
distribution, consistent with the subject’s condition. One of these numbers was
expressed as a two digit decimal numeral; the other was expressed as a fraction.
Because the same numbers were used in this task as in the WTR task and the logic of
the tasks is identical, I can compute a “revealed” exchange rate for the exchange rate
tasks in the same manner as an estimated WTR is computed in the WTR task. Thus, I
compute a revealed exchange rate as if I had not given subjects an exchange rate and
was attempting to determine what exchange rate they were in fact using. The
consistency measures presented below are based on this revealed exchange rate. This
avoids biasing the results such that WTR performance appears more consistent—after
all, I do not have access to subjects’ actual WTRs, only their performance.
Subjects’ responses and reactions times for each decision were recorded in
Study 5. After completing all three tasks, subjects were also asked to rate the
difficulty of each task on separate 7-point scales from 1 = “not at all difficult” to 7 =
“very difficult.” Due to a computer error, this measure was not recorded for eleven
subjects.
Additional measures. To assess whether there were predictable relationships
between our assessment of subjects’ WTRs and psychological responses that should
covary with or are instantiated by WTRs, we included additional measures in Studies
1, 2, & 3.
60
Social Value Orientation. The Social Value Orientation (SVO) measure is
designed to assess general preferences for resource allocation and has been shown to
be an ecologically valid measure of actual behavior (Van Lange, Otten, DeBruin, &
Joireman, 1997). This measure consists of nine items. Each item has three choices.
Each choice allocates a number of “points” to the subject and a number of “points” to
the friend/acquaintance. One choice is the prosocial choice. Subjects choosing this
option reveal a preference that both they and the other receive equal amounts.
Another choice is the individualistic choice. Subjects choosing this prefer to receive
the most possible (relative to their payoffs under the other two choices), regardless of
the other’s outcome. (In this case, the other receives less than the subject and less
than in the prosocial choice.) The remaining choice is competitive. Subjects preferring
this choice desire to maximize the difference between their allocation and the other’s
allocation even though this makes the other worse off than in the prosocial and
individualistic choices and makes the subject worse off than in the individualistic
choice but identical to the prosocial choice.
According to the standard scoring procedure (Van Lange, et al., 1997), a
subject is categorized as prosocial, individualistic, or competitive if 6 out the 9
choices are consistent with that orientation. Subjects who do not have 6 choices
consistent with any of the orientations do not appear in any analyses involving SVO.
In practice, very few subjects were classified as competitive. To simplify the analyses
I combine the individualist and competitive categories into a single category (as is
standard in this literature, Van Lange (1999)). If the WTR measure is actually
61
measuring an internal magnitude that regulates welfare tradeoffs, then there should be
a positive correlation between subjects’ SVO (prosocial dummy coded as 1;
individualistic/competitive dummy coded as 0) and WTRs: Individuals with higher
WTRs toward their friends/acquaintance should act in a more prosocial way toward
them.
Communal Strength. The Communal Strength Scale (CSS) is designed to
measure “the degree of responsibility a person feels for a particular communal
partner’s welfare” (Mills, Clark, Ford, & Johnson, 2004, p. 213). It is a valid and
reliable measure that predicts actual behavior (Mills, et al., 2004). Sample items
include “How large a cost would you incur to meet a need of your
friend/acquaintance?” and “How much would you be willing to give up to benefit
your friend/acquaintance?” Responses are made on 10-point scales with “not at all”
and “extremely” as anchors. In many ways, the CSS is a face-valid rating-scale
measure of the WTR construct (especially as applied to relationships based on
“intrinsic” valuation). Reliability measures (Cronbach’s α) were high, ranging from
.75 to 88. The final measure is created by averaging the items. CSS scores should
positively correlate with WTR scores.
Inclusion of Other in Self. The Inclusion of Other in Self (IOS) scale is
designed to measure the phenomenology of psychological closeness (Aron, Aron, &
Smollan, 1992). The scale is a single, graphical item. It depicts seven sets of two
circles that are labeled “self” and “other.” The pairs of the circles are depicted as
having progressively more overlap. Subjects are asked to choose which degree of
62
overlap best represents their relationship with the other. More overlap represents
greater inclusion of the other in the self. The scale is coded 1 to 7, with higher values
representing greater closeness. IOS scores should positively correlate with WTR
scores.
Relationship Quality Items. I created six additional items to measure the
quality of subjects’ relationships with their friend and acquaintance. Items were rated
on a 7-point scale with anchors of “not at all,” “somewhat,” and “very” at 1, 4, and 7.
The items were (1) how much the subjects liked their friend/acquaintance, (2) how
close they felt to their friend/acquaintance, (3) how much they could rely on their
friend/acquaintance, (4) how much their friend/acquaintance could rely on them, (5)
how willing they would be to do their friend/acquaintance a favor, and (6) how
willing their friend/acquaintance would be to do them a favor. Reliability measures
(Cronbach’s α) were high, ranging from .88 to .92. The final measure is created by
averaging the items. The relationship quality items should be positively correlated
with WTRs.
Results
To recap, for each subject I estimated (i) an average WTR across the six
anchored sets, (ii) the percentage of anchored sets that are perfectly consistent (i.e.
have one switch point), and (iii) the average percentage of choices within an anchored
set that are consistent with the WTR assigned to that set by consistency
maximization. Welfare tradeoff decisions were completed by subjects once for a close
63
friend and once for an acquaintance (in Studies 1-4) or for only one of these
categories (Study 5).
All significance tests use two-tailed p-values. Standard Pearson correlations
(r) are used as a measure of effect size. Unless otherwise noted, I test hypotheses
using focused contrasts (Rosenthal, Rosnow, & Rubin, 2000). I examined every WTR
measure across all five studies for sex differences, applying a two-tailed α of .005 to
each comparison. (Given that many of the dependent measures are not independent, it
is difficult to do standard corrections, such as the Bonferroni. An α of .005 applied to
each comparison acts as if there were only two independent tests per study and a total
α of .05, obviously a very liberal threshold). No differences passed this threshold. Sex
differences do not appear to qualify my results.
Result 1: A Single Switch Point Predicts Decisions
Do subjects’ responses meet the criterion of consistency? Are their decisions
summarizable by a single number? Yes. In the US, 83-85% of anchored sets were
perfectly consistent with a single switch point; in Argentina, the percent was lower,
but still remarkably high at 70% (see Table 1, leftmost data column). If subjects were
responding randomly, the expected value is only 1.1% (relative to this null, all ps <
10-43, effect sizes (r) range from .91 to .97).
64
Table 1
Means (Standard Deviations) of the Consistency Measures
Perfect
Consistency
Sample 1
Ordered
Unordered
Collapsed
Sample 2
Ordered
Unordered
Collapsed
Sample 3
(Unordered)
Sample 4
(Unordered)
Consistency
Maximization
% Sets Consistent
Using Target
Identity
% Sets Consistent As
General Social
Preference
Test of the
Difference
r
p
88% (23%)
83% (24%)
85% (23%)
—
—
25% (32%)
—
—
.88
—
—
10-54
74% (32%)
65% (29%)
70% (31%)
—
—
16% (27%)
—
—
.86
—
—
10-39
85% (22%)
32% (33%)
.86
10-44
83% (25%)
32% (35%)
.84
10-22
% Decisions
Consistent Using
Target Identity
% Decisions
Consistent As
General Social
Preference
Test of the
Difference
r
p
Sample 1
Ordered
98% (6%)
—
—
—
Unordered
97% (5%)
—
—
—
Collapsed
97% (5%)
90% (8%)
.76
10-32
Sample 2
Ordered
95% (8%)
—
—
—
Unordered
95% (7%)
—
—
—
Collapsed
95% (8%)
88% (9%)
.82
10-31
Sample 3
98% (4%)
93% (6%)
.71
10-23
(Unordered)
Sample 4
97% (5%)
91% (10%)
.64
10-10
(Unordered)
Note. Results in the columns labeled “Using Target Identity” average across
consistency scores assigned to tasks involving friends and acquaintances. “Collapsed”
averages across data in the “Ordered” and “Unordered” rows.
65
As its name implies, the perfect consistency score requires perfect
consistency. But inconsistencies due to lapses of attention or error are expected. If
even one decision among 10 is not consistent with the switch point used for the other
9 decisions in an anchored set, that set will not count as perfectly consistent. The
consistency maximization measure quantifies consistency without requiring
perfection. By this alternative measure, 95-98% of choices were consistent with the
welfare tradeoff ratio ultimately assigned to that ordered set; this is higher than the
expected value of 70.84% if answers were random (relative to this null, all ps < 10-59,
effect sizes (r) range from .95 to .99). By either measure, subjects’ decisions were
highly consistent with a single switch point, suggesting that decision-making
procedures are consulting a precise, internal magnitude for each social other.
Result 2: Consistency is Not Due to the Artificial Nature of the Task
Are subjects’ consistent responses due to the artificial nature of the task? To
see how this could be the case, observe that, in Fig. 1A, the tradeoffs within an
anchored set are ordered: the amount for the other—the anchor—stays the same, and
the amount for the self decreases with each successive decision: for example, $39 for
self or $37 for friend; $31 self or $37 friend; $24 self or $37 friend, and so on.
Ordered sets invite consistency, especially when all tradeoffs are presented
simultaneously on the same paper or screen: a person who is willing to forgo $31 (but
not $39) to deliver $37 to a friend, can see, with no further computation, that all
subsequent tradeoffs in the set allow them to deliver the same benefit to their friend
($37) with even less self-sacrifice (forgoing $24, $17, $9, etc). Consider, in contrast,
66
Fig. 1B, where the order of presentation is randomized, not just within an anchored
set but between them: $31 self or $37 friend could be flanked by $21 self or $46
friend, $71 self or $68 friend or indeed, by any tradeoff in the instrument. There is no
simple way to extract a single switch point from this unordered task format; any
attempt to achieve consistency through deliberation—by remembering and comparing
one’s responses across tradeoffs—will impose a large load on working memory,
especially when tradeoffs are presented one at a time on a computer screen.
If the consistency observed results from deliberative processes that construct a
switch point on the fly—that is, in response to the task instead of through
computations based on stored magnitudes—then consistency will suffer when
subjects are given the unordered, randomized format. But if decisions are made by a
system that consults internal magnitudes, the welfare tradeoff ratios, then consistency
should be just as high for unordered as for ordered sets. Because each decision is
standalone, it can be forgotten as soon as it is made with no loss in consistency. On
this view, consistency is achieved by consulting an internal magnitude computed
from stored, upstream values, not by deliberative processes that remember and make
comparisons across decisions.
As Table 1 shows, consistency is high regardless of task format: there were no
significant differences in consistency between the ordered and unordered versions of
the task in Samples 1 and 2, whether the standard was perfect consistency or
consistency maximization (.08 ≤ ps ≤ .92, .01 ≤ rs ≤ .16). Indeed, consistency was
just as high for unordered tasks of Samples 3 and 4, where tradeoffs were presented
67
one at a time on a computer screen—a method that prevents subjects from revisiting
any previous decision. High levels of consistency that are insensitive to large
variations in task formats is what one would expect if responses are made by
consulting welfare tradeoff ratios rather than being computed online.
Result 3: Subjects’ Decisions Represent Actual Valuation of Others; They Are
Not “Cheap Talk”
As seen in Table 2, subjects’ average welfare tradeoff ratios are well above
zero. Is this apparent generosity a product of “cheap talk?” No: Hypothetical
decisions and decisions with real monetary consequences produced similar estimated
welfare tradeoff ratios and consistencies (Sample 3, Table 2). For WTRs for friends,
the size of the difference between the two conditions was r = .14, p = .099. For WTRs
for acquaintances, the size of the difference was even smaller, r = .05, p = .51. As
shown in Fig. 2, the frequencies of estimated welfare tradeoff ratios for friends and
acquaintances, while not identical, are extremely similar across the two conditions.
Regardless of the stakes, subjects’ responses appear to be based on an internal
variable representing real valuation.
Result 4: WTRs Are Person-Specific; They Carry More Information Than a
Simple Other-Regarding Preference
Result 4A: The mind assigns different WTRs to different individuals. Do
subjects’ responses reveal target-specificity in their decision making? Or are friends
and acquaintances treated identically, as if subjects had a single other-regarding
preference? If the mind computes person-specific welfare tradeoff ratios as part of a
68
system for making adaptive welfare allocations, then different individuals should be
valued differently. To test this prediction in Studies 1-4, subjects completed the task
twice, once for a friend and once for an acquaintance. As Table 2 shows, subjects’
revealed WTRs were higher for friends than for acquaintances.
Result 4B: Decisions regarding two individuals cannot be summarized by
a single WTR. Result 1 showed that subjects’ responses for a given target could be
summarized by a single switch point. But does that finding reflect the existence of a
target-specific WTR or is it due to a general—but consistent—social preference that
is applied across targets? I recomputed consistency scores for each subject, collapsing
across the 10 decisions they made for a friend and the 10 they made for an
acquaintance in a given anchored set. If a subject’s decisions reflect nothing more
than a general social preference, a single switch point should be found; that is,
consistency scores should be as high as those found when the computation is done
separately for friends and acquaintances.
An examination of Table 1 shows that the identity of the target cannot be
ignored. Whether looking at perfect consistency or consistency maximization,
subjects’ responses are much less consistent if summarized by a single switch point.
Effect sizes (r) when comparing target-specific consistency to general preference
consistency were large, ranging from .84 to .88 for perfect consistency and .64 to .82
for consistency maximization.
69
Table 2
Welfare Tradeoff Ratio Means (Standard Deviations)
Friend-Acq
Difference
Origin &
Sample #
Friend Welfare
Tradeoff Ratio
Acquaintance
Welfare
Tradeoff Ratio
r
p
United States
.62 (.35)
.34 (.36)
.67
10-22
Sample 1
Argentina
.75 (.37)
.50 (.37)
.66
10-17
Sample 2
United States
Sample 3
Pay
.43 (.40)
.30 (.41)
.56
10-6
Hypothetical
.54 (.38)
.34 (.39)
.60
10-8
Collapsed
.48 (.39)
.32 (.40)
.58
10-13
United States
.44 (.45)
.27 (.40)
.41
10-4
Sample 4
Note. Acq = acquaintance. “Collapsed” averages across data in the “Pay” and
“Hypothetical” rows.
70
WTR for Friend
Number of Subjects
25
Hypothetical
Pay
20
15
10
5
1.
45
25
1.
05
1.
0.
85
65
0.
45
0.
25
0.
05
0.
15
-0
.
-0
.
35
0
WTR Value
WTR for Acquaintance
Number of Subjects
20
Hypothetical
Pay
15
10
5
.1
5
0.
05
0.
25
0.
45
0.
65
0.
85
1.
05
1.
25
1.
45
-0
-0
.3
5
0
WTR Value
Figure 2. Frequency distributions of welfare tradeoff ratios for friends and
acquaintances when choices do and do not have monetary consequences. Although
the size of the WTRs is reduced in the condition with monetary consequences, the
majority of subjects in that condition still have greater-than-zero WTRs.
71
However, at least for the perfect consistency measure, there is a potential
confound. For an anchored set to count as perfectly consistent when treating targets
individually, a subject need only make ten decisions without error. However, to count
a set as perfectly consistent when treating WTRs as a general social preference, a
subject must make twenty decisions without error (i.e., the ten about a friend and the
ten about an acquaintance for the same anchored value). Because there are more
possible ways to make an error, this could be artificially lowering the apparent
consistency of subjects when their responses are evaluated as a general social
preference. (Note that this objection cannot apply to the consistency maximization
results and so cannot be used as a counter-hypothesis for this dependent variable. This
is because the consistency of decisions regarding, e.g., a friend cannot affect the
consistency of decisions regarding, e.g., an acquaintance for the consistency
maximization measure—the consistent effects are purely additive. However, for the
perfect consistency measure, if the decisions regarding a friend are not made perfectly
consistently then, when analyzing as a general social preference, it will not matter
whether the decisions involving an acquaintance were made perfectly consistently—
regardless, the combined set will be counted as not perfectly consistent.)
To make sure this was not the case I did the following. The probability of
making no errors in a ten item ordered set—call this P(correct|10 decisions)—can be
estimated by computing the average percentage of perfectly consistent anchored sets
that a subject completes. Note that completing two anchored sets perfectly is required
of subjects when consistency is evaluated as if WTRs were a general social
72
preference. So, if subjects did have a general social preference, then the probability of
completing the same anchored set for a friend and an acquaintance in a perfectly
consistent fashion is the square of P(correct|10 decisions). For example, using Sample
1’s collapsed mean from Table 1, P(correct|10 decisions) = 85%. Squaring this value
returns 72% as the predicted probability of making a perfectly consistent set of 20
decisions. However, attempting to summarize the 10 decisions regarding a friend and
the 10 decisions regarding an acquaintance with only a single general social
preference results in only 25% of sets being perfectly consistent, far below 72%. I
computed this value for each subject and tested it against the percentage of consistent
sets when acting as if WTRs were a general social preference. Relative to the effect
sizes in Table 1, the effects sizes in these comparisons are barely reduced, .77 < r <
.84, all ps < 10-17. This analysis reaffirms the conclusion that subjects’ decisions
regarding multiple others cannot be reduced to a general social preference.
Result 5: Individualized WTRs Are Correlated with Decisions Involving Only
Third-Parties
Welfare tradeoff ratios are hypothesized to be part of an evolved welfare
allocation system. If this is true, then aside from allowing direct tradeoffs between
one’s own and another’s welfare, decision-making procedures will also consult
WTRs, or the variables that determine WTRs, when making tradeoffs between two
other individuals (the criterion of common effects). Consider, for example, decisions
in which a parent must decide which of two children gets the last piece of food.
73
Making such a decision requires integrating the mind’s separate valuations of each
child.
To test this, subjects in Study 4 completed the standard self–other tasks for a
friend and an acquaintance (Fig 1B), and also a version in which they made tradeoff
decisions between the same friend and acquaintance. Just as the standard version
allows the estimation of how much a subject will tradeoff personal welfare to benefit
another, this new version allows the estimation of how much a subject will tradeoff a
friend’s welfare to benefit an acquaintance.
If target-specific WTRs (or the factors that set target specific WTRs) are used
in making tradeoffs between two third-parties, then these two values—WTRfriend and
WTRacquaintance—should predict friend-acquaintance tradeoffs in a multiple regression.
The more one values a friend—that is, the greater one’s WTR for the friend—the less
one will trade off that friend’s welfare to benefit an acquaintance. I first test this using
as the dependent variable the version of the task involving trading off a friend’s
welfare to benefit an acquaintance (i.e., where the acquaintance is in the “anchored”
position). Results conformed to predictions: When WTRfriend and WTRacquaintance were
entering simultaneously in a multiple regression analysis, the unstandardized
regression weight for WTRfriend as a predictor of friend-acquaintance tradeoffs was
negative, b = -.32, p = .008. Similarly, the more one values an acquaintance—that is,
the greater one’s WTR for the acquaintance—the more one will trade off a friend’s
welfare to benefit that acquaintance. This too was supported: in the same regression
analysis, the unstandardized regression weight for WTRacquaintance was positive, b =
74
.65, p = 10-6 . These results show that decisions involving two other individuals are
partially a function of the target-specific welfare tradeoff ratios (or the variables that
are ultimately used to compute WTRs), as predicted by the hypothesis that WTRs are
important internal regulatory variables within an integrated social allocation system.
Like the self–other instrument, the friend–acquaintance instrument used above
puts the individual likely to be less valued—the acquaintance—in the anchor position.
Although I also gave subjects a friend–acquaintance instrument in which the friend
was in the anchor position, there are interpretational issues when the more highly
valued person is in the anchor position (e.g., for the self-other instrument, putting self
in the anchor role rather than where it is in Figure 1; or for the friend-acquaintance
instrument, putting the friend into the anchor role). Consider a case where my
WTRfriend is.5 and my WTRacquaintance is .1. Hypothetically, I should give to the
acquaintance if the following inequality is satisfied: WTRfriend * Benefitfriend <
WTRacquaintance * Benefitacquaintance. Rearranging, WTRfriend/WTRacquaintance <
Benefitacquaintance/Benefitfriend. Substituting in .5/.1 = 5 < Benefitacquaintance/Benefitfriend.
So Benefitacquaintance would need to be more than five times Benefitfriend for me to even
begin allocating to my acquaintance. That is, if my friend is the anchor/“other” role in
Fig. 1 and my acquaintance is in the “you” role, then I might not choose the option
benefiting my acquaintance until the amount was >5 times higher than the amount
available for my friend. Yet the instrument never had differences this large; the value
for the acquaintance did not go above 1.55 times the amount for the friend. This
means that no subject who values their friend > 1.55 times as much as their
75
acquaintance will make a decision favoring their acquaintance over their friend. As a
result, these people will all be scored as having the highest friend-favoring ratio (i.e.,
the highest WTRfriend/WTRacquaintance ratio) that the instrument allows (1.55), even
though their actual friend-favoring ratio can have any value exceeding 1.55
(reciprocally, they will have the lowest acquaintance-favoring WTR allowed by the
instrument, 1/1.55 producing a floor effect). This ceiling effect dramatically curtails
the variance in measured WTRs, restricting its ability to correlate with other
measures. A ceiling effect was clearly present in the data: when friend was in the
anchor position, the distribution of measured WTRs was strongly skewed with the
modal value being the highest possible score. (In contrast, WTRs were normally
distributed when acquaintance was in the anchor position, just as they were when
other was the anchor in the self-other instruments. The distribution of choice values
were chosen on the assumption that the lower valued person would be in the anchor
position).
Although this effect was anticipated, I decided to keep the task identical to the
other samples to allow direct comparability. Even though there was the expected
ceiling effect when friend was in the anchor position, multiple regression showed that
this measure of tradeoff between acquaintance and friend was positively related to
WTRfriend (b = .38, p = .016), and negatively (though not significantly) related to
WTRacq (b = -.24, p = .16). That is, you are more willing to sacrifice the welfare of
your acquaintance to help your friend the more highly you value your friend relative
to yourself; and you are less willing to sacrifice the welfare of your acquaintance to
76
help your friend when you value that acquaintance highly relative to yourself. The
weaker effect sizes here are expected given the range restriction artificially imposed
by my measure, but the direction of the effect sizes was preserved.
Results 6 & 7: Welfare tradeoff ratios behave as other magnitudes, but are
algorithmically separable from other magnitudes
Welfare tradeoff theory predicts that the mind contains an evolved internal
variable designed to regulate welfare tradeoffs and that an instantiation of this
variable is created for each social actor with whom the self interacts. Further, the
existence of this variable would explain subjects’ ability to achieve high levels of
consistency on the WTR task. However, one may reasonably ask to what extent these
effects are due to general rationality or mathematical abilities.
To begin answering this question, note that the relevance of the consistency
data to the hypothesis that there is an internal regulatory variable for making welfare
tradeoffs is due to the switch point being self-generated. Subjects appear to create
their own decision threshold and use that threshold to make an internally consistent
series of decisions. Moreover, this decision threshold correlates with theoretically
relevant measures (see Table 3 below) and is largely unaffected by whether decisions
have monetary consequences (Figure 2 and Table 2). If the WTR was not a stable
feature of the mind, then manipulations that make it less likely for subjects to create a
single, coherent switch point—such as a random presentation of the choices—should
have a considerable impact on subjects’ consistency. Yet they do not (compare the
ordered and random conditions in Result 2).
77
The hypothesis is not that people—particularly college-educated subjects—
would have difficulty making consistent tradeoffs if they were given a specified
tradeoff ratio to use in their calculations. There is a great deal of evidence that human
mind contains evolved specializations for manipulating analog magnitudes and that
numerals can be assigned to these magnitudes (e.g., Spelke, 2000). Moreover,
members of industrialized societies have “culturally bootstrapped” systems of exact
mathematics (e.g. conscious arithmetic skills) (e.g., Dehaene, Izard, Spelke, & Pica,
2008). These abilities would allow subjects to easily complete the task—if a tradeoff
value were specified for them.
The ability of subjects to complete the WTR task using a self-generated
tradeoff value (i.e. the hypothesized WTR) is most likely partially dependent on the
ability to represent and manipulate analog representations of quantity. To choose
between two sums for self and other expressed with numerals, the numerals would
need to be mapped to internal magnitudes. The novel theoretical point advanced here
is that the internal representations of the two external sums would then be input for a
domain-specific algorithm that arrives at an adaptive decision using the WTR. Two
potential ways this algorithm may work are by (a) discounting the internal value for
the other by the WTR and then comparing it to the internal value for the self or (b)
computing the ratio of the value for the self to the value for the other and then
comparing this to the WTR.
Because a general mathematical ability account could not, in and of itself,
explain the origin of welfare tradeoff ratios—why subjects spontaneously generate
78
these numbers or any numbers at all—it is difficult to see how such an account could
parsimoniously explain the fundamental result—the self-generation of a single,
precise, target-specific WTR. Nonetheless, I would like to go further and examine the
ease with which subjects completed our task. If there is domain-specific machinery
for making welfare tradeoffs, subjects should find this an easy and intuitive task.
To examine this, I conducted Study 5 in which subjects again completed the
WTR task (for one person only) and also completed two tasks that were
mathematically identical to the WTR task. Both tasks asked subjects to decide
whether X dollars or Y Euros was worth more. Subjects were given a hypothetical
exchange rate that allowed them to convert Euros to dollars. One version of the task
gave subjects a hypothetical exchange rate expressed as a decimal. The second
version gave subjects a different hypothetical exchange rate expressed as a fraction.
Subjects were not required to memorize these exchange rates; the computer displayed
them throughout the task.
Given the structure of this task, it can also answer questions about whether our
results could plausibly be due to the operation of domain-general rationality. Welfare
tradeoff theory proposes that the mind contains specialized information-processing
machinery designed for making adaptive welfare tradeoffs and allocations. One might
propose as an alternative hypothesis that the mind has a general mechanism for
computing preferences for a wide variety of stimuli, social and non-social, and that
domain-general rationality operates over the resulting preference function. Under this
view, preferences would be assigned to the X dollars for self and the Y dollars for the
79
other, or, just as easily, to hot dogs, investment rates into a personal retirement
account, potential spouses, and so on. Then, (for mutually exclusive choices) the
various preferences would be compared and the most preferred would be chosen.
Given that different targets elicit different revealed WTRs, the function that
maps objective dollars for the other person into a preference must necessarily include
a person-specific component. I have been assuming that, at a minimum, it involves
multiplying dollars for the other by a person-specific constant. Even if one considers
this particular assumption questionable, the more general claim that decision-making
must involve multiplication is not: all psychological theories of choices over
subjective utilities necessitate that the mind be able to multiply (Birnbaum, 2008b).
Thus, a logically prior ability for making choices over preferences is the ability of
preference mechanisms to implicitly perform multiplication. Moreover, rational
choice theories rely heavily on computations involving probabilities. Surely, by these
accounts, such probabilities are represented by magnitudes that functionally have two
decimals of precision, such as I use in the decimal exchange task to represent the
exchange rate. Thus, on a rational choice account, using a number with two decimal
places should be unproblematic.
Altogether, this implies that the task demands of the WTR task and the
exchange task are similar: both require computations that, at a minimum, involve the
multiplication of two numbers and then the comparison of two numbers. Indeed, there
are several ways in which the exchange task has fewer demands. First, there is no
need to transform objective money into subjective preferences; one need only
80
multiply one number by the exchange rate and then compare. Second, all information
required to solve the exchange task is presented simultaneously and there is no need
to access any stored preference information; for the WTR task not all information is
presented, requiring some to be drawn from memory. Therefore, the alternative
hypothesis that our results can be explained by general rationality predicts that the
exchange task will be just as easy, if not easier, to solve.
Comparisons of the WTR task with the two versions of the exchange rate task
can answer four questions: (1) Are there any differences on measures of consistency
between the WTR task and the other two tasks? (2) Are there differences in the time it
takes subjects to makes their decisions depending on the task? Greater time for the
exchange rate tasks would suggest those tasks are more difficult. (3) Are the
exchange rate tasks experienced as being more difficult than the WTR task? (4) Do
similar manipulations have similar effects on the WTR task and the exchange rate
tasks? If not, this would suggest that they are being processed by distinct algorithms
despite their mathematical similarity.
First, as expected, there were no differences on either measure of consistency.
To make the computation of consistency as similar as possible across the two task
types, instead of comparing subjects’ decisions to the given exchange, I calculate
consistency as for the WTR task and determine whether any switch point is consistent
with subjects’ decisions in the exchange task. (Note that the mean absolute difference
between the given and revealed exchange rate was .16 and that the median absolute
difference was .08, both less than the resolution of the instrument.) The mean
81
percentage of anchored sets that were perfectly consistent in the WTR task (M = 82%,
SD = 25%) was not different from the means percentage of sets that were perfectly
consistent in the decimal exchange rate task (M = 87%, SD = 22%), t(111) = -1.48, p
= .141 (two-tailed), r = -.14, or in the fractional exchange rate task (M = 82%, SD =
25%), t(111) = -.09, p = .932 (two-tailed), r = .01. Using the consistency
maximization method, the percentage of choices consistent with the assigned tradeoff
value in the WTR task (M = 98%, SD = 5%) was not different from the percentage of
consistent choices in the decimal exchange rate task (M = 98%, SD = 4%), t(111) = 1.35, p = .180 (two-tailed), r = -.13, or in the fractional exchange task (M = 98%, SD
= 4%), t(111) = -.25, p = .801 (two-tailed), r = -.02. These results show as expected
that subjects were equally able to give consistent responses to all choice tasks. Note,
moreover, that the switch points in the exchange rate tasks were given to the subjects
and were always displayed on the computer screen whereas subjects had to generate
the switch point in the WTR task.
Second, subjects took approximately 50%-80% longer to complete exchange
rate decisions relative to WTR decisions. The time per decision in the WTR task (M =
1823 msec, SD = 654 msec) was significantly shorter than the time per decision in the
decimal exchange rate task (M = 2939 msec, SD = 1629 msec), t(111) = 7.48, p = 1010
, r = .58, and in the fractional exchange task (M = 3309 msec, SD = 2345 msec),
t(111) = 6.45, p = 10-8, r = .52. We can exclude the first 10 decisions for each task
type to remove any effects that may be due to subjects becoming familiar with the
mechanics of the task. Here as well, the time per decision in the WTR task (M = 1376
82
msec, SD = 614 msec) was significantly shorter than the time per decision in the
decimal exchange rate task (M = 2378 msec, SD = 1374 msec), t(111) = 6.33, p = 108
, r = .52, and in the fractional exchange task (M = 2652 msec, SD = 1990 msec),
t(111) = 5.43, p =
10-6, r = .46. These results suggest that the exchange rate tasks were more difficult for
subjects.
Third, subjects experienced the WTR task as significantly easier than the other
two tasks. The rated difficulty of the WTR task (M = 2.42, SD = 1.49) was less than
the rated difficulty of the decimal exchange task (M = 3.56, SD = 1.62), t(100) = 5.45,
p = 10-6, r = .48, and the fractional exchange task (M = 4.05, SD = 1.84), t(100) =
6.84, p = 10-9, r = .56.
Finally, I can examine whether the same manipulation causes the same effects
on responses to both types of task. Specifically, I examine how increasing the size of
the anchored values in the anchored sets affects the switch point computed for each
set. Consider the anchored set depicted in Figure 1A. If the ratios of the amounts in
each decision are held constant but the anchored value (i.e. the value for the other) is
increased, the absolute amounts involved are increased. Although in principle one
might expect magnitude effects (as are observed for other types of discounting (Green
& Myerson, 2004)), I specifically chose anchored values to try to minimize them.
Nonetheless, across the first four samples, initial analyses suggested that there was a
small, but statistically reliable, effect of magnitude. The effect was such that the
greater the anchored value within a set, the smaller was the WTR assigned to that
83
set—when more money was at stake, subjects were less generous. On average, the
maximum difference between anchored sets was approximately 0.1, only half of the
distance between successive WTRs that could actually be assigned to an anchored set.
If decisions on the WTR task and the exchange rate tasks are being generated
by the same process, then one can expect that the revealed exchange rate in each
anchored set should decrease as the absolute value increases. (A subject’s revealed
exchange rate is potentially different from the exchange rate assigned to that subject
by the experimenter, see Methods.) This was not the case: As the absolute amount
involved increased, subjects’ WTRs decreased, but their revealed exchange rates
increased, see Figure 3. A focused contrast (Rosenthal, et al., 2000) revealed that the
slope of WTR task data was significantly different from the slope of the exchange
task data: The linear contrast for the WTR task was significantly different from the
average linear contrast of the exchange rate tasks, t(111) = 3.56, p = 10-3, r = .32. The
linear contrast for the WTR task shows a significantly negative slope (-0.48), t(111) =
-2.60, p = .011, r = .24; as the amount at stake increases, subjects become less
generous. The average linear contrast of the exchange rate tasks shows a significantly
positive slope (0.29), t(111) = 2.74, p = .007, r = .25; in these tasks, subjects appear to
assume that the greater the number in the anchored position, the greater its relative
value. It is not just that the slopes differ in magnitude, they differ in direction as well.
This supports the hypothesis that separate processes are involved in the WTR and
exchange rate tasks.
84
These data can also be used to further support Result 2: welfare tradeoff ratios
are computed by specialized machinery, not constructed deliberatively in response to
the task. As shown above, subjects’ WTRs as revealed in a given anchored set subtly
decreased as the absolute magnitude involved increased, even when the decisions
from all the anchored sets were randomly mixed together. Such subtle responses are
expected from a system designed to make adaptive, context-sensitive social
allocations. (See Appendix 6 for an additional experiment bearing on this point.) This
pattern would make sense if, for example, the mind were attempting to minimize risk
or opportunity cost—there is more to lose if one forgoes $75 relative to forgoing $19
(cf. Stewart-Williams, 2007). This pattern of responses is very unlikely if subjects are
constructing an arbitrary threshold online and attempting to make decisions consistent
with it. Why not simply use the same threshold at all times? Indeed, varying the
threshold as a function of the absolute amounts involved greatly increases how
challenging the task is. If subjects’ minds were not using a stored magnitude
embedded within a context-sensitive social allocation system—for instance, if
conscious deliberation were driving our results—it is difficult to see how or why such
subtle and patterned responses would emerge.
85
0.7
Decimal Exchange Task
Revealed Switch Points (+/- SEM)
0.65
0.6
0.55
Fraction Exchange Task
0.5
0.45
0.4
WTR Task
0.35
0.3
19
23
37
46
68
75
Anchor Size
Figure 3. Revealed exchange rates and WTRs as a function of the value of the anchor
in the anchored sets. As the anchored values increase, the amount at stake increases.
86
To make full use of the data, the above results were computed using a withinsubjects approach. However, one could wonder whether contrast effects, order
effects, or practice effects may have affected the results. To address this, I also
analyzed the data using a between-subjects approach, using only the first task each
subject completed. The pattern of the results was identical. For tests of differences in
consistency, there were again no differences, all ps > .45. All other tests revealed
significant effects, all ps < .006.
These results show that the underlying processes giving rise to WTR task
decisions and exchange rate task decisions are not identical. Although subjects are
ultimately just as consistent on all types of tasks, they nonetheless take at a minimum
50% longer to complete the exchange rate tasks, they experience the exchange rate
tasks as more difficult, and the same manipulation affects responses in the two tasks
in opposite ways. If subjects’ performance was due simply to general rationality or
mathematical abilities, all the tasks should have been just as difficult to complete.
Indeed, on this alternative account the WTR task might be even more difficult. For
the exchange rate tasks, subjects are given an exchange rate to use and are not
required to memorize it. For the WTR task, however, subjects must both generate and
remember their WTR if they are to perform consistently. Although some aspects of
rationality and mathematical abilities may be necessary components for making
welfare tradeoffs, they are not sufficient to explain them. Instead, the results are
consistent with the existence of a domain-specific social allocation decision system.
87
For further evidence regarding the specialized nature of welfare tradeoff computation,
see Appendix 7.
Correlations of the WTR Measure with Other Measures
If the WTR task is actually measuring the output of an internal variable
designed to regulate welfare tradeoffs—designed to index valuation of another
individual—then it should show positive correlations with other measures designed to
measure sharing, helping, relationship quality, and the phenomenology of
interpersonal closeness. As seen in Table 3, the four rating scale measures generally
had large and significant correlations with the switch point calculated from the WTR
task. This table also displays the psychometric reliability (Cronbach’s α) of the WTR
task (α was calculated for the six separate series of decisions subjects completed for a
single target); these values are uniformly very high.
88
Table 3
Reliability of the WTR Measure and Its Correlations with Other Scales
United States
Sample 1
WTR for
Friend
WTR for
Acquaintance
Argentina
Sample 2
WTR for
Friend
WTR for
Acquaintance
United States
Sample 3 (Pay/
Hypothetical)
WTR for
Friend
WTR for
Acquaintance
Reliability
(Cronbach’s
α)
Social Value
Orientation
Communal
Strength
Relationship
Quality
Inclusion of
Other in
Self
.92
.44***
.34***
.32***
.11
.95
.40***
.41***
.40***
.25***
.89
.31***
.20*
.27**
.10
.93
.35***
.21*
.14
.25**
.97
.45***
.30***
.17*
.05
.98
.51***
.45***
.44***
.27***
Note. See methods for a description of the measures. The reliabilities for Sample 4
(which did not include the other measures) ranged from .96 to .99. The lack of a
correlation between the Friend WTR and the Inclusion of Other in Self (IOS) is
possibly due to there being little variance in IOS measure; most people rated their
relationship with their friend very high on this measure. ***p ≤ .001, **p < .01, *p <
.05.
89
Summary of Chapter 2: The Existence of Welfare Tradeoff Ratios
The five studies of this chapter provide evidence consistent with the
hypothesis that the mind calculates an internal variable for regulating welfare
tradeoffs—a welfare tradeoff ratio. These studies showed seven primary results:
1. Subjects’ welfare tradeoff decisions were summarizable by a single number;
that is, their decisions were made consistently.
2. This consistency was not an artificial feature of the task. Subjects behaved just
as consistently regardless of whether decisions were made in an ordered or
random series.
3. Subjects’ WTRs were not the product of cheap talk: The magnitude of their
WTRs was essentially unchanged regardless of whether real money was at
stake.
4. Welfare tradeoff ratios were person-specific: Subjects had higher WTRs for
their friends, relative to acquaintances, and decisions for both friends and
acquaintances could not be summarized by a single number.
5. Subjects’ separate welfare tradeoff ratios toward a friend and an acquaintance
correlated with welfare tradeoffs between the friend and acquaintance.
6. Results from a mathematically identical task show that the WTR behaves as
other psychological magnitudes, allowing consistent decisions to be made.
7. Nonetheless, the pattern of responses to the WTR task—the ease and speed
with which subjects complete it—shows that welfare tradeoff computations
are handled by a specialized system.
90
Although the mind may calculate welfare tradeoff ratios toward others, does it
estimate WTRs stored in others’ minds? I address this question in Chapter 3.
91
Chapter 3: The Estimation of Welfare Tradeoff Ratios2
The difference between a full and an empty stomach can be a matter of life
and death. Not surprisingly, research in animal psychology and behavioral ecology
reveals exquisitely crafted psychological machinery for estimating variables
necessary to support efficient foraging (Gallistel, 1990; Stephens, Brown, &
Ydenberg, 2007). Humans are no exception: experimental and ethnographic evidence
reveals that the human mind also contains foraging specializations (Kameda &
Tamura, 2007; New, Krasnow, Truxaw, & Gaulin, 2007; Silverman & Eals, 1992;
Winterhalder & Smith, 2000). But are there other variables that need to be estimated,
aside from rates of return, especially for a social animal such as humans? In this
chapter, I investigate whether humans also estimate the welfare tradeoff ratios
contained in the minds of others.
To provide converging evidence with Chapter 2, the methodology of this
chapter involves substantial methodological changes. First, I examine whether WTR
machinery can be applied to any agent—not just individual humans. If so, then it
should be applicable to coalitions as well. Part of the evolution of multi-person
cooperation may have involved the ability to conceptualize cooperative groups as
agents (Tooby, Cosmides, & Price, 2006). Therefore, to establish the generality of
WTR psychology, subjects in the experiments below were (implicitly) asked to infer
the WTRs that individual humans have toward a coalition of humans.
2
The research presented in this chapter was conducted in collaboration with Theresa Robertson.
92
Second, instead of explicitly asking subjects what kinds of tradeoffs other
people would make, I used an implicit task to determine whether subjects viewed
individuals with different profiles of costs or benefits—and presumably different
WTRs—as belonging to separate categories. Although WTRs are conceptualized as
continuous magnitudes, a great deal of evidence suggests that humans can flexibly
turn dimensional distinctions into categorical distinctions (Hampton, 2007;
Jackendoff, 1983). Thus, although the measure is ultimately categorical (either a
person is or is not in a given category), it can still be used to probe an underlying
continuous dimension. Only after the categorization measure does the subject answer
explicit questions designed to test whether the categories are related to welfare
tradeoff content.
The Logic of Welfare Tradeoff Ratio Estimation
Does the mind estimate the welfare tradeoff ratios of others? If so, what inputs
would a system for WTR estimation use? Consider how standard WTR computation
proceeds: If I can potentially provide my group a benefit Bgroup at personal cost Cme, I
will only do so in actuality when Bgroup * WTRme,group > Cme, or, rearranging, when
WTRme,group > Cme / Bgroup. That is, the group will receive the benefit when my welfare
tradeoff ratio for the group—my valuation of the group—is greater than the costbenefit ratio. All else equal, as Cme increases, so should one’s (lower bound) estimate
of my WTR. Individuals willing to sacrifice more personal welfare will be inferred to
have a higher WTR toward the group. This is tested in Studies 6-8.
93
An examination of the equation relating WTRs to costs and benefits suggests
an additional hypothesis: As Bgroup increases, the entire fraction becomes smaller,
implying that smaller B’s should be linked with larger inferred (lower bound)
estimates of WTRs. Indeed, in other work the size of the benefits does matter in
inferences about WTRs (Sell, 2005). Instead of testing this aspect of welfare tradeoff
theory, however, I would like to examine variations in benefits for a slightly different
purpose, using a somewhat different manipulation than would be required to test the
more obvious prediction in this context. Specifically, I would like to further test
against rational choice and payoff-based approaches for understanding social
perception in social foraging. From these perspectives, information regarding benefits
provided should be used to make selective partner choices. However, the costs a
person pays to provide these benefits are immaterial. After all, if you are designed to
maximize your returns, why should you care how much my contributions cost me?
Even if costs did have some information value, variations in benefits should still have
a larger impact.
From welfare tradeoff theory, however, benefits and costs are only
inferentially powerful when they are revealing about underlying regulatory variables.
Thus, when variations in benefits delivered cannot reveal anything about a person’s
WTR, then they should not be encoded. This is tested in Studies 9 & 10. (Of course, it
is likely that the size of the benefits will have some cue value, even if it is not as
effective as variations in cost. Although I attempt to remove any explicit information
about efficiency or ability in this chapter’s experiments, the mind may nonetheless
94
assume that individuals who provide more benefits are somewhat more able and
therefore more valuable.)
To see whether the mind distinguishes between individuals who pay small or
large costs, or between individuals who provide large or small benefits, I use an
implicit categorization task (Taylor, Fiske, Etcoff, & Ruderman, 1978). This task
assumes that members of the same category are more likely to be confused with each
other than with members of different categories. For example, a person paying a
small cost should be confused with others who pay small costs, but not with others
who pay large costs. Although WTRs are conceptualized as continuous magnitudes, a
great deal of evidence suggests that humans can flexibly turn dimensional distinctions
into categorical distinctions (Hampton, 2007; Jackendoff, 1983). Thus, a categorical
measure can still be used to probe an underlying continuous dimension. Although the
categorization measure tells whether the mind makes certain distinctions, it does not
reveal the content of the categories. Therefore, after the categorization task, subjects
complete a number of rating scales to determine whether the categories have the
social consequences predicted by welfare tradeoff theory.
Study 6: Variations in Cost Incurred
Methods
Subjects. Fifty-eight students (33 female) at the University of California,
Santa Barbara (UCSB) participated in exchange for partial course credit.
Procedure and dependent variables. All materials and dependent variables
for this study and the remaining studies in Chapter 3 are presented in Appendix 8.
95
Subjects were asked to form impressions of eight fictitious individuals (each
represented by a facial photograph of a white man) whose plane had crashed on a
deserted island. The target individuals’ survival required that they all work together
and forage for food. For each of five “days” on the island, subjects read a sentence
describing each target’s foraging experience (40 total sentences). On three days, each
target found and shared a resource by paying a “default” cost (i.e., the time and
energy necessary to walk and look for food); because these sentences do no
discriminate between targets, I call them non-diagnostic sentences. For example,
“With some passion fruit in hand, he walked over a few fallen logs and went back to
camp.” On two additional days, half of the targets chose to pay an exceptionally large
cost to forage and the other half paid a small, accidental cost; because these do
discriminate, I call them diagnostic sentences. For example, “In order to catch some
crabs, he waded into the shark-infested shallows off the bay” versus “While picking
strawberries, he cracked and broke his watch against a rock.” The order of sentences
and the pairings of sentences to targets were randomly determined for each subject
with the constraints that all targets were paired with a non-diagnostic sentence on the
first day and no target was paired with a diagnostic sentence two days in a row.
After viewing the five foraging days, subjects completed a surprise memory
test requiring them to match each event to the target who originally performed it.
Categorization is assessed using the memory confusion paradigm of Taylor and
colleagues (1978). In this paradigm, the pattern of errors made by subjects in the
surprise memory test is used to infer social categorization. (Correct responses are not
96
analyzed because it is impossible to know if they are due to accurate memory, a
memory confusion, or random responding.) If subjects make more within- than
between-category confusions, this suggests that the hypothesized categories are
psychologically real to them. Within-category confusions occur when a subject
misattributes an action by (e.g.) an individual who paid a large cost to another
individual who paid a large cost. Between-category confusions occur when a subject
misattributes an action by (e.g.) an individual who paid a large cost to an individual
who paid a small cost. I compute a categorization score by subtracting betweencategory confusions from within-category confusions (after correcting for differing
base rates by multiplying between-category confusions by ¾; without such a
correction, random responding would appear as systematic misattribution to the
opposite category). If categorization is occurring, categorization scores should be
greater than zero. Based on previous work (Delton, Cosmides, Guemo, Tooby, &
Robertson, 2010), I expect categorization scores for the diagnostic sentences to show
the strongest, most consistent evidence of categorization, with comparatively weaker
effects for the non-diagnostic sentences. Thus, although I report results for both types
of sentences, my interpretations are based only on results for the diagnostic sentences.
After completing the categorization measures, subjects rated the targets on a
number of dimensions using 7-point scales with appropriate anchors at “1” and “7.”
Brief descriptions are displayed in Tables 1 & 2. At no point during the memory test
or the impression ratings did I provide subjects with information about targets’
behavior; only photographs were available.
97
Results
I examined every categorization and impression variable across all five studies
in this chapter for sex differences with a two-tailed α of .001 (slightly less
conservative than a Bonferroni correction for a total α of .05). No differences passed
this threshold. Sex differences do not appear to qualify my results.
All p-values are two-tailed.
Does categorization occur as a function of costs incurred? Yes, subjects
categorized targets as a function of costs incurred: Categorization scores for the
diagnostic sentences were reliably greater than zero, M = 2.73, SD = 3.50, t(57) =
5.94, p = 10-6, r = .62. There was also a tendency for categorization scores for the
non-diagnostic sentences to be greater than zero, M = 0.78, SD = 3.85, t(57) = 1.54, p
= .128, r = .20.
Do impressions differ as a function of costs incurred? Yes, subjects viewed
targets who incurred larger costs more positively on all items, all ps < .07 and all rs ≥
.24 (Tables 4 & 5).
98
Table 4
Effect sizes (r) of mean differences on the impression items
Study 6
Study 7
Study 8
Benefiting group at large cost
always viewed more positively
Study 9
Study 10
Providing large benefits
always viewed more
positively
Item
Feels altruistic
toward group
.36**
.40**
.42***
.22
.08
Cares about group
.58***
.14
.50***
.15
.18
Desirable as a
group member
.60***
.41**
.49***
.03
.25†
Desirable for 1:1
cooperation
–
.45***
.50***
.30†
.28*
Willing to
Contribute
–
.46***
.48***
.10
.28*
Trustworthy
.45***
.32*
.54***
.09
.26*
Deserves respect
.48***
.43***
.49***
.19
.28*
Likeable
.49***
.18
.45***
.32*
.14
Put in effort
.59***
.54***
.52***
.07
.24†
Competent
.61***
.46***
.49***
.21
.19
Is not selfish
.24†
–
–
–
–
Intended to pay
large cost
.48***
–
–
–
–
Worthy of being
leader
.57***
–
–
–
–
Does not deserve
punishment
.29*
–
–
–
–
99
Deserves reward
.50***
–
–
–
–
Note. Each effect size, r, represents the comparison of two means: One mean from
targets who pay a large cost and one from targets who pay a small cost (Studies 6-8)
or one mean from targets who provide a large benefit and one from targets who
provide a small benefit (Studies 9 & 10). Comparisons were made using repeated
measures t-tests. Interested readers can convert using r = Sqrt( t2 / ( t2 + df )); t = ( r2 *
df ) / ( 1 – r2 ); df = sample size minus 1. “–” indicates that data on this item were not
collected in a particular study. †p < .10, *p < .05, **p < .01, ***p < .001.
100
Table 5
Means (Standard Deviations) of Impression Items
Study 6
High
Low
WTR
WTR
4.78
4.48
(0.94)
(1.03)
Study 7
High
Low
WTR
WTR
4.92
4.60
(0.86)
(0.90)
Study 8
High
Low
WTR
WTR
4.70
4.13
(1.22)
(1.16)
Cares about
group
5.29
(0.84)
4.75
(1,00)
5.13
(0.96)
5.01
(0.93)
5.23
(0.92)
4.51
(1.09)
5.41
(0.92)
5.29
(0.99)
5.17
(1.17)
4.96
(1.15)
Desirable as a
group member
5.38
(0.93)
4.63
(1.20)
5.14
(0.94)
4.77
(0.94)
4.99
(1.05)
4.36
(0.94)
5.07
(1.12)
5.03
(0.80)
4.75
(1.17)
4.49
(1.04)
Desirable for 1:1
cooperation
–
–
4.94
(1.02)
4.49
(0.98)
4.60
(1.09)
3.89
(1.05)
4.98
(1.11)
4.66
(1.03)
4.32
(1.16)
4.05
(1.08)
Willing to
Contribute
–
–
5.57
(0.88)
5.14
(0.93)
5.24
(0.84)
4.63
(1.05)
5.51
(0.95)
5.41
(0.89)
5.07
(1.03)
4.81
(1.08)
Trustworthy
5.19
(0.86)
4.73
(1.08)
5.12
(0.94)
4.86
(0.98)
4.87
(0.91)
4.23
(0.95)
5.01
(0.96)
4.91
(0.80)
4.73
(1.13)
4.46
(1.07)
Deserves
respect
5.09
(0.91)
4.56
(1.05)
5.23
(0.94)
4.78
(0.95)
4.93
(0.78)
4.33
(0.84)
5.16
(0.94)
4.99
(1.07)
4.67
(0.92)
4.39
(0.95)
Likeable
5.04
(0.82)
4.56
(1.07)
4.92
(0.95)
4.76
(0.98)
4.81
(0.81)
4.24
(0.89)
5.03
(0.96)
4.70
(0.86)
4.43
(0.96)
4.31
(0.92)
Put in effort
5.45
(0.85)
4.73
(1.03)
5.54
(0.85)
4.99
(0.86)
5.24
(0.89)
4.54
(0.95)
5.40
(1.03)
5.33
(0.98)
5.04
(1.02)
4.80
(1.02)
Item
Feels altruistic
toward group
Study 9
Large
Small
Benefits
Benefits
5.02
4.84
(1.27)
(1.29)
Study 10
Large
Small
Benefits
Benefits
4.56
4.46
(1.17)
(0.98)
101
Competent
5.41
(0.82)
4.78
(1.06)
5.26
(0.97)
4.82
(0.88)
5.15
(0.99)
4.63
(1.02)
5.29
(0.97)
5.12
(0.90)
4.77
(1.07)
4.58
(1.06)
Selfish
2.72
(1.13)
2.95
(1.06)
–
–
–
–
–
–
–
–
Intended to pay
large cost
4.39
(1.18)
3.78
(1.20)
–
–
–
–
–
–
–
–
Worthy of being
leader
4.75
(1.07)
3.91
(1.35)
–
–
–
–
–
–
–
–
Deserves
punishment
2.16
(1.07)
2.43
(1.20)
–
–
–
–
–
–
–
–
Deserves reward
5.36
(0.99)
4.77
(1.27)
–
–
–
–
–
–
–
–
102
Note. “–” indicates that data on this item were not collected in a particular study. The median repeated measures correlation
was .57 (first quartile = .47; third quartile = .61).
Study 7: Varying Costs with Constant Intentions
Study 6 is consistent with the hypothesis that larger incurred costs lead to
greater estimated welfare tradeoff ratios. However, there is a potential confound in
the stimuli: All targets necessarily paid the default costs of foraging, but only those
who paid a large cost paid any additional cost intentionally; those who incurred a
small, incidental cost did so accidentally. Could the effects in Study 6 be driven by
whether an additional cost was paid intentionally, instead of the size of the cost? To
rule this alternative explanation out, Study 7 replicated the first study but removed
sentences involving incidental costs. Thus half the targets paid an exceptional cost on
some occasions on top of the default costs of foraging whereas the other half only
paid the default costs. In all cases, all costs were paid intentionally.
Method
Subjects. Fifty-nine students (31 female) at UCSB participated in exchange
for partial course credit.
Procedure and dependent variables. The procedure was identical to Study 6
except that all sentences with small incidental costs were removed. This required
having only four days of foraging such that there was no initial day during which only
non-diagnostic sentences were shown. Note that only targets who paid large costs
were paired with diagnostic sentences. I added two new questions to more directly
assess the value of targets as cooperation partners. I also reduced the total length of
the rating portion by removing a number of peripheral questions.
103
Results
Does categorization occur as a function of costs incurred? Yes, subjects
categorized targets as a function of costs incurred: Categorization scores for the
diagnostic sentences—which in this study only existed for people paying large
costs—were reliably greater than zero, M = 1.38, SD = 2.82, t(58) = 3.75, p < .0004, r
= .44. There was no tendency for categorization scores for the non-diagnostic
sentences to be greater than zero, M = -0.39, SD = 3.65, t(58) = -0.81, p = .210, r = .11.
Do impressions differ as a function of costs incurred? Yes, subjects viewed
targets who incurred larger costs more positively on most items, most ps < .05 and rs
≥ .30 (Tables 4 & 5). Although in the predicted direction, the effects for likeability
and caring about the group were somewhat small.
Study 8: Manipulating the Target of Costs Incurred
In Studies 6 & 7, subjects consistently perceived targets paying a larger cost
as more competent. Is a perceived difference in ability to contribute—not a perceived
difference in willingness—driving our results? To test against this possibility, in
Study 3 half the targets incurred an exceptional cost to benefit the group and the other
half benefited the group at the “default” cost and then incurred an exceptional cost to
benefit themselves (thus equating ability to incur costs). Indeed, the latter targets
actually found more food and so in an objective sense are more able.
The hypothesis that greater costs incurred lead to higher inferred WTRs
assumes that the mind is attempting to estimate group-specific welfare tradeoff ratios.
104
It thus predicts that the targets will be categorized as a function of who receives the
food (the group vs. the target himself) and that targets who provision the group will
be viewed more positively.
Method
Subjects. Seventy-six students (36 female) at the UCSB participated in
exchange for partial course credit.
Procedure and dependent variables. The procedure was identical to Study 6
except that half the targets were paired with diagnostic sentences depicting them as
paying a large cost to provision the group, whereas the other targets were paired with
diagnostic sentences depicting them as paying default costs to provision the group
and then paying a large cost to provision himself. For example, “He exposed himself
to the hazardous waves on the rocks of the bay so he could catch sea bass to bring to
the group” versus “After providing some food for the group, he waded through a river
full of deadly piranhas to hunt duck for himself.” The rating items were identical to
Study 7. Arranging the stimuli in this manner works against the hypothesis that
individuals incurring a large cost on behalf of the group will be viewed more
positively: These individuals actually produce less than individuals provisioning both
themselves and the group.
Results
Does categorization occur as a function of who received a benefit? Yes,
subjects categorized as a function of whether large costs were incurred on behalf of
the group or on behalf of the target himself: Categorization scores for the diagnostic
105
sentences were reliably greater than zero, M = 2.25, SD = 3.90, t(75) = 5.04, p = 10-5,
r = .50. Categorization scores for the non-diagnostic sentences were not appreciably
greater than zero, M = 0.40, SD = 3.65, t(75) = 0.96, p = .171, r = .11.
Do impressions differ as a function of who received a benefit? Yes,
relative to targets benefiting themselves at a large cost, subjects’ viewed targets who
provisioned the group at a large cost more positively on all items, all ps < .001 and all
rs ≥ .42. As in the earlier studies, targets who incurred large costs on behalf of the
group were viewed as more competent. This occurred despite the other targets
actually finding more food, suggesting that social factors can influence perceptions of
competence.
Discussion
Studies 6-8 reveal that targets are categorized by the size of the costs they
incur. Nonetheless, I note a potential caveat in interpreting these data. In these studies
targets elicited significantly different ratings of competence as a function of the costs
they incurred (a measure of efficiency or ability). Although it was not the goal of this
chapter to investigate the effects of ability on person perception, it is possible that the
stimuli inadvertently included subtle ability information. Although I tried to correct
for this experimentally in Study 8 by having the targets who placed low value on the
group be objectively more able, this study still showed differences in perceived
competence such that targets who placed greater value on the group were seen as
having greater competence. One interpretation is that this is due to a “halo” effect,
106
stemming from the more direct inference regarding care and altruism. Another
possibility, however, is that I was unable to remove inadvertent cues to ability.
Studies 9 & 10: Varying Benefits Provided
Studies 6-8 provided consistent evidence that targets willing to incur greater
costs on behalf of the group were categorized separately from those who weren’t;
moreover, they were viewed more positively, including on WTR-relevant dimensions
(e.g. on items of care and altruistic feelings toward the group). That costs are encoded
at all is not strongly consistent with rational choice or payoff-based approaches.
(Indeed, these approaches might predict that people who pay more to deliver the same
benefits—that is, have lower efficiency—might be viewed less positively.) What role
do variations in benefits have? According to welfare tradeoff theory, variations in
benefit delivery should be encoded when they cannot reveal underlying differences in
WTRs.
To test this, new diagnostic sentences were created: Half the targets in Studies
9 & 10 were twice depicted as paying default costs to provide a very small benefit
(e.g. a small handful of nuts); the remainder were twice depicted as paying default
costs to provide a very large benefit (e.g. a very large amount of fruit). These
differences are not revealing of underlying WTRs because it is random with respect
to effort whether a person will encounter a large or a small resource. All targets paid
the default costs of searching the island for food; only chance determined what
resources they would find.
107
Studies 9 & 10 used different sets of non-diagnostic sentences: In Study 9, on
three further occasions all targets provided a medium sized benefit at default costs. In
this study the environment is depicted as relatively resource rich—all targets always
find at least something. If positive sentiment is preferentially directed at targets
finding large benefits, it could be that subjects perceive targets finding small benefits
as making poor decisions—why settle for a handful of nuts when you can find a
basket of fruit? To test against this, in Study 10, on the three further occasions, all the
targets failed to find any food. Thus, resource acquisition was not certain. Note that a
manipulation check at the end of Study 9 showed that subjects perceived the large
benefits as much more beneficial than the small benefits (r = .83, p < .001). Although
subjects certainly noticed the size of the benefits, will this difference affect
perceptions of the targets?
Methods
Subjects. Forty-three students (24 female; Study 9) and 57 students (29
female; Study 10) at UCSB participated in exchange for partial course credit.
Procedure and dependent variables. The procedure was identical to Study 1
with the following exceptions. In Studies 9 & 10, half the diagnostic sentences
depicted the targets finding a small benefit at a default cost; the remainder depicted
the targets finding a large benefit at a default cost. For example, “He got food for the
group by climbing one the highest trees on the island to collect the few cashews he
saw at the top” versus “He noticed a large patch of peaches in the grove at the top of a
perilous cliff and climbed up to gather loads to bring back for the group” In Study 9,
108
the non-diagnostic sentences depicted targets as finding a medium benefit at a default
cost as in previous experiments. In Study 10, the non-diagnostic sentences depicted
targets as failing to find a benefit at a default cost. For example, “He searched all over
the island but found no food to bring back.” The rating items were identical to Study
2.
Results & Discussion
Does categorization occur as a function of the size of the benefit? No,
categorization scores for diagnostic sentences were not reliably different from zero
(and were in the wrong direction) in Study 9 and Study 10, respectively: M = -0.15,
SD = 2.44, t(42) = -0.41, p = .34, r = -.06; M = -0.29, SD = 2.25, t(56) = -0.96, p =
.17, r = -.13. This was also true for non-diagnostic sentences, Study 9 and Study 10,
respectively: M = 0.48, SD = 2.94, t(42) = 1.06, p = .15, r = .16; M = -0.07, SD =
2.77, t(56) = -0.18, p = .43, r = .02.
Do impressions differ as a function of the size of the benefit? There was
some tendency for targets who provided larger benefits to be viewed more positively,
although the size of the effects were relatively small and sometimes only marginally
significant (Tables 4 & 5): Study 9, rs ranged from .03 to .32 and ps from .03 to .82;
Study 10, rs ranged from .08 to .28 and ps from .03 to .53. Although all impression
measures favored individuals providing large benefits, the effect sizes in Studies 9 &
10 were smaller than in Studies 6-8. This is opposite the relative effect sizes that
would be predicted by rational choice or payoff-based accounts.
109
Summary of Chapter 3: The Estimation of Welfare Tradeoff Ratios
Consistent with predictions, subjects categorized targets by the size of costs
incurred and did not categorize targets based on unrevealing variations in benefits
delivered. This is unpredicted by rational choice and payoff-based approaches to
choosing cooperation partners. Under these perspectives, individuals are valuable as
cooperation partners when they provide large benefits; the costs others pay are much
less relevant. Altogether, the results of this chapter are consistent with the prediction
from welfare tradeoff theory that the mind can estimate the welfare tradeoff ratios of
others.
110
Chapter 4: Discussion and Conclusions
This research has explored the nature of a summary variable—the welfare
tradeoff ratio—and shown (a) that the mind computes this variable and (b) that the
mind can estimate this variable in the minds of others. The discovery of this
neurocomputational element helps to explain how a physical device such as the
human brain can engage in behavior as complex as cooperation, aggression, and
altruism. The characterization of human social decision-making in precise,
information processing terms allows researchers to build cumulative (and falsifiable)
theories of human social behavior. By adopting the common language of information
processing, researchers from the many disciplines interested in human behavior—
anthropology, economics, neuroscience, political science, and psychology, among
others—can begin to pool their knowledge and increase the rate at which human
nature is mapped.
Extensions and Future Directions
Ongoing Work on Welfare Tradeoff Theory
Chapter 3 provided preliminary evidence that the mind can estimate welfare
tradeoff ratios. In other work, Sznycer and colleagues (2010) have gone further. In
this experiment, pairs of friends were brought into the lab. Both members of the pair
filled out the WTR decisions used in Chapter 2 with respect to their friend who came
with them to the lab. Subjects also completed the scale as they believed their friend
would toward them (i.e., toward the subject). Results showed that the expected WTR
111
from the friend accurately tracked the friend’s actual WTR. Moreover, this accuracy
was person-specific: Each subject also completed the task with respect to a third-party
and completed the task as they believed the third-party would toward them.
Controlling for either of these variables—which serve as measures of generalized
generosity and generalized expectations of generosity—did not account for subjects’
accuracy regarding their friend.
Related to this, Lim and colleagues (2010) have shown that the mind can
distinguish between multiple levels of WTRs expressed toward the self. In this
experiment, subjects were paired with a (sham) partner. Through a computer, the
subject observed this partner make a number of welfare tradeoff decisions that
affected the subject and the partner, with the decisions all being consistent with a
particular WTR toward the subject. Between subjects, this WTR was varied from zero
to .9, with four intermediate values as well. Importantly, Lim et al. manipulated the
numbers at stake such that regardless of the partner’s welfare tradeoff ratio toward
the subject, the objective amount the subject received was in all cases identical.
Consistent with welfare tradeoff theory, subjects’ WTRs toward the partner increased
monotonically with the partner’s WTR toward the subject. Thus, subjects (a) extract
relatively precise WTRs from observing the patterns of decisions of others and (b)
sensitively regulate their behavior in response to differences in inferred WTR.
Although barely addressed in the present work, Lim et al. and Sznycer et al.
also explore the intimate links between welfare tradeoff ratios and emotional systems.
For instance, Lim et al. show that decisions by the partner that are especially
112
revealing of a high WTR toward the subject elicit gratitude; this is not true of
decisions that simply involve large benefits accruing to the subject. In a similar
fashion, decisions revealing an especially low WTR toward the subject elicit anger
and annoyance; this is not true of decisions that simply involve small benefits
accruing to the subject. Future work can benefit by further integrating welfare
tradeoff ratio methodology and emotions.
Studying Welfare Tradeoffs Without Numbers
One limitation of the studies in Chapter 2 is that they involved explicitly
presented numerals. Presumably, welfare tradeoff allocation systems evolved prior to
the cultural invention of numerals and should thus be able to operate without them.
(Although Chapter 3 did not use explicit numerals, it did not examine WTRs at a very
high level of resolution. These studies were only able to discriminate “large” from
“small” estimated WTRs.) It would provide strong converging evidence to show that
WTRs affect decision-making even without using explicit numbers.
One way to do this would take advantage of the matching law (Herrnstein,
Rachlin, & Laibson, 1997). When presented with two places to forage in, animals
usually divide their time in proportion to the relative rates of benefit delivery at the
two locations; this behavioral regularity is called the matching law. For example, if
equal sized pellets of food are available at both locations, but are delivered to patch 1
at twice the rate as patch 2, then the animal will spend twice as much time at patch 1.
(There are technical qualifications in exactly how the schedule of benefits must be
designed to ensure matching (Herrnstein, et al., 1997); I do not go into these here.)
113
This regularity can be used to investigate welfare tradeoff psychology in the
following way. First, consider that in normal matching experiments, the benefits of
choosing either foraging patch accrue to the chooser. Thus, if I had to choose between
patch 1 and 2, I would choose patch 1 at twice the rate. Starting from this baseline, an
experiment could manipulate whom the benefits of a foraging patch accrue to.
Foraging patch 1 still accrues to me, but when I select patch 2, any benefits I find
could accrue to my friend. Now, instead of dividing my time in a 2-to-1 ratio, I
should divide my time in a 2-to-(1*WTR) ratio. No explicit numerals are necessary.
Instead, “pellets” of identical value appear at two different patches and the
beneficiary of at least one patch is manipulated. Subjects’ behavior under these
conditions could reveal exactly how precise the internal magnitudes used in welfare
tradeoff calculations are.
Broadening the Breadth of Decisions
Chapter 2 mostly considered a relatively simple decision: Whether to allocate
one benefit to oneself or a different sized benefit to another. According to welfare
tradeoff theory, these decisions require computing whether cself < WTR * bother. Yet
many decisions do not fall into this pattern. As a more general description of welfare
allocation problems, consider the following framework. (This framework is presented
has an “in principle” mathematical formalism; in practice processing constraints or
other considerations may limit whether the mind could or would instantiate what is
described here.)
114
First, assume there is a set S of mutually exclusive states of the world that the
decision-maker could bring about, s1 through sn. (Although states are mutually
exclusive with another, they may contain shared sub-states.) Associated with each
state is a welfare allocation from the set A, a1 through an, such that the realization of
(e.g.) s1 causes a1 to obtain. Each state and welfare allocation impacts a set of agents
P, p1 through pn. When state si is realized, leading to welfare allocation ai, each agent
p1 through pn has their welfare affected by an amount vi,1 through vi,n, with the second
subscript indexing the agent that the welfare change applies to. (Each a is a set of
welfare changes, making A a set of sets. The welfare changes, v, can be any
cognizable magnitude able to enter into these calculations, including presumably
zero.) Finally, the decision-maker has a set of welfare tradeoff ratios, W, toward each
agent, w1 through wn. One of the ps, pself, represents the decision-maker; wself should
in theory equal 1. The total welfare gained by the decision-maker for a given si and ai
is the sum of all vi,k * wk, as k ranges from 1 to n. If the decision-maker is unilaterally
able to determine which state is realized, this becomes a maximization problem:
realize the state si that leads to the highest welfare gain for the decision-maker.
Each component of this mathematical formalism requires extensive
processing. For instance, determining the set of welfare changes v is a substantial
problem. Moreover, as the number of possible allocations and persons involved
increases, a computational explosion results, making computations involving large
numbers of allocations and person impossible. Nonetheless, as illustrated by Study 4
in Chapter 2, the mind can successfully perform welfare tradeoff tasks involving two
115
other people (aside from the self) and a quick reflection on everyday life suggests that
people can perform welfare allocations involving small groups of people. Although
an unbounded version of this extended formalism is surely not computationally
feasible, the breadth of the mind’s ability to perform welfare allocations suggests that
the machinery must allow more complex computations than those involving the self
and one other and only two possible allocations.
Comparisons to Other Research Programs
The previous chapters have developed reasons to believe that the mind estimates a
number of indices about others in its social world and integrates these estimates into a
summary representation, a person-specific welfare tradeoff ratio, to determine how to
allocate welfare between different parties. What are some existing theoretical
viewpoints that relate to, or perhaps contradict, this? I compare welfare tradeoff
theory to several below. In most cases, WTR theory is not at odds with existing
theory, but complements it by adding specificity to important aspects of welfare
allocation.
Affective Heuristic Approaches
A common view among many branches of cognitive and social psychology is
that judgment and decision making—when accomplished through means other than
conscious manipulation of magnitudes—is accomplished by inexact, often affective
means (Bechara, Damasio, & Damasio, 2000; Haidt, 2001). Consider how this view
would interpret choices people make between gambles (e.g., would you rather have
$50 for sure, a 50% chance of earning $120, or a 10% chance of earning $2000?).
116
The affective heuristic approach would predict that the mind generates a rough,
affective response to each option. Whichever option has the most positive affect (or
least negative affect) would be chosen. According to this view, there is no need for
precision computation; inexact, heuristic estimation satisfices. (Some proponents of
this view also argue that there is a “rational” system at work as well, but that this does
not often come into operation.)
The results of the present studies do not lend much support to this view.
Although subjects in my studies do not do conscious manipulation of magnitudes
when completing the welfare tradeoff task, their data nonetheless reveal extremely
precise patterns of behavior. Subjects’ informal descriptions of their decision process
often reduce to “I gave to me when it felt appropriate, and the other when that felt
appropriate.” Despite this vagueness in the phenomenology of their decision process,
their data can neatly be summarized by a single number—a welfare tradeoff ratio—
across a wide number of randomly mixed decisions. This is not due, moreover, to
subjects generally anchoring on switch points that would require little computation,
such as zero or one (leading one to always allocate to the self (except when the
amount is negative) or to always allocate the larger amount, respectively). An
examination of Figure 2 shows that the majority of subjects use switch points other
than these.
Of course, the present data do not directly bear on the type of data usually
accounted for by affective heuristic approaches: choices over gambles. However, data
directly investigating these domains often reveals surprisingly precise choice
117
behavior of subjects, in ways that seem to require precise complex calculations
(Birnbaum, 2008b; Rode, Cosmides, Hell, & Tooby, 1999). Preuschoff and
colleagues (2006) conducted a neuroimaging study of gambling decisions. Instead of
finding that the brain encodes rough, approximate magnitudes, they predicted and
found that the brain encoded in a precise fashion the expected value and the variance
of gambles. Moreover, these activations were found in brain areas typically thought to
be involved in affective heuristic processing.
The Elementary Forms of Human Interaction Approach
In a wide-ranging research program, Fiske (1992) has amassed a great deal of
evidence that, he argues, shows that human interaction is structured by four
elementary types of prosocial relationships. One of the major components of these
elementary relationships is that they determine how resources are to be shared. These
elementary types of interaction and their associated sharing rules are:
1. Communal Sharing: For a specified set of people, for a specified purpose and
resource, resources are pooled: Individuals contribute what they can and take
what they need—“from each according to his abilities, to each according to
his needs.”
2. Authority Ranking: Individuals are ranked, with higher ranked people being
entitled to more and better resources. (On this view, however, higher ranked
people are also expected to be beneficent toward their subordinates.)
3. Equality Matching: Everyone receives identical shares and contributes
equally.
118
4. Market Pricing: Exchanges and allocations of resources are accomplished by
implicit or explicit contracts. (See the section on social contract theory below
for relevant comments on this sharing rule.)
Given the wide range of cultures and situations in which Fiske finds these four
types of relationships, they appear to capture something real about human
psychology. What is the relationship between these modes of interaction and welfare
tradeoff theory? Although some relationships, such as nuclear families, may be
structured by communal sharing, there is no specification of when a person’s “needs”
become more than the rest of the set is willing to endure. Would members of a
communal sharing set be willing to let an individual whose self-reported needs are
essentially infinite take the vast majority of the resources created by the set? Welfare
tradeoff theory can help explain how limits to communal sharing are instantiated:
Although members of a communal sharing pool may all set high WTRs toward other
members, their WTRs are not infinitely high; eventually, taking according to one’s
need may come to be seen not as entitlement but as exploitation when such taking
crosses WTR boundaries.
A similar issue arises in authority ranking. Individuals gain authority because
of their ability to inflict costs and/or confer benefits. In virtue of this, according to
elementary relationship forms theory, lower ranked people allow them to take larger
or better shares. But what counts as too much? A variety of anthropological evidence
shows that lower ranked people will eventually rebel against a highly ranked person
who has overstepped what the lower-ranked feel is appropriate (Boehm, 1993). (A
119
similar phenomenon is even seen in capuchin monkeys (Gros-Louis, Perry, &
Manson, 2003).) The hypothesis that the mind computes relatively precise WTRs
with respect to higher and lower ranked people can explain why authority ranking
does not allow infinite benefit taking. At some point the benefits taken by a higher
ranked person, relative to the costs inflicted on lower ranked people, are too far from
the welfare tradeoff ratios that lower ranked individuals set toward the higher ranked
person.
Whereas systems for communal sharing and authority ranking may interact
with welfare tradeoff computations, equality matching may simply apply to situations
where nuanced WTR calculations are not possible. Such situations may include multiperson cooperative ventures where tracking the inputs of multiple contributors and
apportioning benefits proportionate to inputs may be difficult or impossible (Tooby,
et al., 2006). Here, there may not be a useful connection between the structures
instantiating the relationship type and WTR computation. Alternatively, people’s
decisions to join a collective endeavor may depend on their WTRs toward other
contributors. Consider that a person may perceive their personal benefits from a
collective endeavor to be subjectively less than the costs if they contributed. Thus, on
grounds of personal welfare alone, it would not be rational for them to contribute.
However, if the person’s WTRs toward one or more of the other parties is sufficiently
high, it nonetheless might be adaptive to contribute.
120
Altogether, welfare tradeoff theory and the elementary forms of interaction
approach do not conflict. Instead, welfare tradeoff theory can supply boundary
conditions to at least some of the forms.
Other-Regarding Preferences Approaches
Standard approaches in economic game theory, when applied to the behavior
of individuals, attempted to explain behavior in terms of rational computations
designed to maximize personal objective payoffs. (Although it is not, in principle,
necessary for agents to attempt to maximize personal objective payoffs, this is what
has been traditionally assumed.) Years of experimental evidence have falsified this
assumption: Subjects in behavioral economic games do not behave as predicted if
attempting to maximize personal payoff; instead, they often act in ways that benefit
others even at a cost to the self (Fehr & Henrich, 2003). Because of this, various
theories of other-regarding preferences have been developed to better explain
subjects’ economic decision-making (Camerer & Fehr, 2006). According to Fehr
(2009): “A ‘social preference’ is now considered to be a characteristic of an
individual’s behavior or motives, indicating that the individual cares positively or
negatively about others’ material payoff or well-being. Thus, a social preference
means that the individual’s motives are other-regarding—that is, the individual takes
the welfare of the other individuals into account” (p. 216).
Theories of other-regarding preferences potentially diverge from welfare
tradeoff theory on a number of points. First, except for other-regarding preference
theories explicitly drawing on theories of selection pressures (e.g., Henrich et al.,
121
2005), these theories do not usually have a principled reason to motivate their
functional form aside from intuition and the ability to account for data in the
circumscribed settings of experimental economics games (Fehr & Schmidt, 1999;
Rabin, 1993). Second, models of other-regarding preferences usually require
parameters to be estimated for each subject, but provide no principles to predict the
values these parameters might take. Welfare tradeoff theory, in contrast, provides a
number of principled predictions of individual differences in WTRs and related
internal variables (e.g., Lieberman, et al., 2007; Sell, Tooby, et al., 2009). Third, and
related to the first two points, because other-regarding preference theories do not
usually look to the theoretical constraints provided by theories of selection pressures,
there are an infinite number of possible other-regarding preference theories: Consider
that other-regarding preference theories retain the basic structure of behavioral game
theory— rational (but blinkered) computation operating over a preference structure
(Camerer, 2003). Although preference structures specifying short-term self-interest
(the usual assumption of past work) have been falsified, there are an infinite number
of other possible preference structures the mind could, in principle, adopt. Without
theoretical guidance and constraint, other-regarding preference theories are infinitely
flexible to account for almost any data set. Finally, although other-regarding
preference approaches allow for individual differences in preferences and strategy
(Camerer & Fehr, 2006), it is not clear that the theory can be extended to relationshipspecific valuation (at the very least, I have never encountered the approach applied in
this way).
122
This last issue, even if not a true theoretical difference, arises out of the
behavioral economics-centric approach of other-regarding preference theories. In a
recent review, Fehr (2009) makes strong statements about the appropriate
methodological tools for studying other-regarding preferences (emphasis his):
“Games in which an individual interacts repeatedly with the same partner—a repeated
PD [prisoners’ dilemma] for example—are clearly inappropriate tools…[Moreover,]
the game should NOT be a simultaneous move game (e.g., the simultaneous PD), but
should be played sequentially with the target subject being the second-mover who is
informed about the first mover’s choice” (p. 221). Because games cannot be repeated,
this means also that the games must be anonymous (if this were not the case, the
“game” could extend outside the laboratory). This rules out, by fiat, the possibility of
studying valuation toward family and friends, or toward any individual the subject
has knowledge of beyond the information directly given in a laboratory setting. Thus,
even if other-regarding preference theories are capable of predicting target-specificity
in valuation, the methodological approach of their practitioners prevents them from
examining it. In contrast, in testing predictions of welfare tradeoff theory, I developed
a novel procedure to allow real money to be at stake in a laboratory game where
subjects answered about a friend and an acquaintance who were not present in the lab
(see Study 4 in Chapter 2).
Interdependence Theory
Interdependence theory derives from decades of research in social psychology
(Kelley et al., 2003; Thibaut & Kelley, 1959). This theory weds game theoretic
123
notions, such as payoff matrices, with concepts from social psychology, such as
affect, attitudes, and expectancies. A major component of interdependence theory is
that it provides a theory of situations (Reis, 2008). It analyzes situations (using game
theory-like matrices) along a number of dimensions. The four most important
dimensions are the following (Rusbult & Van Lange, 2008): (1) The level of
dependence: the extent of influence of one person’s actions on another’s outcomes.
(2) The mutuality of dependence: the degree to which the level of dependence is
similar or different across interaction partners. (3) The basis of dependence: the
extent to which one person’s outcomes are influenced by personal choices or the
choices of others. (4) The covariation of interests: the extent to which outcomes
correspond or conflict.
Where interdependence theory diverges from typical game theory approaches
is that instead of considering only objective payoffs, outcomes include not only
payoffs but also emotional effects and other subjective payoffs that bear no direct
relation to the objective payoffs. (This contrasts with the other-regarding preferences
approach which, although it looks at preferences beyond self-interest, still ultimately
deals only with subjective valuations of distributions of objective payoffs.) Although
this broadening is in some ways a virtue of the approach, as it has the potential to
explain phenomena not explainable by standard game theory approaches alone, it is
also a weakness, as it drastically increases theoretical free parameters. Predicting
someone’s behavior becomes not just a matter of analyzing how actions influence
objective payoffs, but to what extent one actor cares about another, how an actor’s
124
emotional response to helping or exploiting another increases or decreases the
subjective value they attach to their outcome, etc. Filling out these promissory notes
requires more specific theories that make substantive predictions about cooperative
behavior. Ultimately, interdependence theory may be more of a meta-theory, a
language within which to talk about interdependent action, than a strongly predictive
theory of behavior. (Consider that to even apply notions of, e.g., the level of
dependence requires knowledge of how people value different actions and outcomes.
But, outside of intuition, these valuations cannot be known without a prior theory to
predict and interpret them.)
Welfare tradeoff theory can help fill in these gaps by articulating the various
determinants of interpersonal valuation (e.g., indices of relatedness) and by providing
a concrete description of how one person’s welfare is translated into units of personal
welfare. Combining this approach with the game theory-like, but psychologically
realistic, approach of interdependence theory may allowed for more refined
prediction.
Social Contract Theory
Multiple lines of evidence converge to show that the mind contains
mechanisms that enable social exchange, the mutual and contingent provisioning of
rationed benefit (reviewed in Cosmides & Tooby, 2005). Particular social exchanges
can take many forms, such as “If you give me meat now, then I will give you meat
later” or “If you give me stock options maturing within the next 10 years, then I will
design the next-generation alternative to the iPad for you.” How does the logic of
125
social exchange and the logic of welfare tradeoffs relate? I break this into two
questions: Are WTRs centrally involved in creating well-formed, sincere social
exchanges? Do WTRs play a more indirect role in bounding the scope of social
exchange obligations?
Suppose that a social exchange involves two people, named “you” and “me.”
A well-formed social exchange requires that I give you a benefit, byou, if you meet a
requirement, cyou. (Although the requirement may be costly to you, that is not
necessary.) Moreover, it requires that you give me a benefit, bme, if I meet a
requirement, cme. If I assume that you are sincere in offering your side of the social
contract, than I face a choice between being sincere also and following through on my
side of the exchange, or attempting to cheat you by accepting a benefit from you
without meeting the requirement you have set of me. Using welfare tradeoff theory,
let’s determine what my WTR toward you would need to be to cause me to be
sincere. The left side of the following equation represents my payoffs if I am sincere;
the right side represents my payoffs if I attempt to cheat you:
bme – cme + WTR * (byou – cyou) > bme – WTR* cyou
Algebraic manipulation reduces this to:
WTR * byou > cme
Or:
WTR > cme / byou
What does this imply? In some cases, my WTR toward you does not need to be very
high at all for me to be sincere in my offer. Consider the limiting case where your
126
requirement of me is not actually a cost (i.e., cme ≤ 0). If that if the case, then I should
be sincere in my offer to you whenever my WTR is even somewhat greater than zero.
But what happens if we consider a case where the requirement of me is quite high
relative to the benefit to you? This would often be the case when transacting with
large corporations: The subjective cost of purchasing a new Toyota is quite high to
me, but the subjective benefits of single additional sale are quite low to Toyota. In
this case, my WTR would have to be quite large for me to be sincere in my offer. Yet,
humans (modern ones, at least) often engage in exchanges that would seem to require
extremely high WTRs to be sincere about them. This suggests that welfare tradeoff
theory does not strictly apply to the domain of social exchanges. Indeed, the creation
of a psychology that enabled economic exchange independent of WTR considerations
may be a central element in allowing human-style trade.
Could welfare tradeoff considerations apply in a relatively indirect way to
social exchanges, however? One possible route would be through (usually implicit)
side-constraints on the entitlements and obligations created by a social contract
(Cosmides & Tooby, 2008). Assume that I have agreed to give you a ride to the
airport in exchange for you cleaning my apartment. What happens if at the last minute
my mother needs to be rushed to the hospital? Would you expect me to follow
through on my obligation? To the extent that most people’s intuition would excuse
me from my obligation (though also excusing you from yours), this suggests there are
side-constraints—conditions under which my costs from providing you with a
benefits are so great (that is, exceeding an implicit WTR), that you release me from
127
our contract. It remains for future work to study whether such side-constraints exist
and if they involve implicit WTRs.
Parting Thoughts: Existentialism Meets Evolutionary Psychology
Part of me hopes that this work frightens you. I know it frightens me. If the
theory presented here is even broadly correct, then our devotion to our friends, our
children, our spouses, and our families is ultimately the product of magnitudes
represented in our heads. Love and affection dissolve into syntactic operations,
necessarily conducted without thought or foresight or feeling. Such is the paradox of
the modern science of the mind: Sometimes looking up at humanity’s highest peaks
of warmth and compassion is no different than staring into the abyss.
128
Appendix 1: Example Materials for Study 1
Instructions
In the following questions, imagine you are given a decision between two options.
Your choice determines how much money you and another person receive. If you
choose the first option, a sum of money will be given to you and no money will be
given to the other person. If you choose the second option, a sum of money will be
given to the other person and no money will be given to you.
For example,
consider the
decisions on the
right:
You Other
Option 1 25
0
Option 2
0
20
You Other
15
0
0
20
You Other
-13
0
0
20
In the first decision, you must choose between $25 for you or $20 for the other
person. If you choose the first option, then you get $25 and the other person gets $0. If
you choose the second option, you get $0 and the other person gets $20.
In the third decision, you must choose between -$13 for you or $20 for the other
person. If you choose the first option, then you would have to pay $13 and the other
person gets zero. Choosing a negative value for yourself means you lose money. If you
choose the second option then you get $0 and the other person gets $20.
decision-maker. They are shown on the right:
As an example,
consider the
decisions made by a
hypothetical
Option 1
Option 2
You Other
25
0
0
20
129
You Other
15
0
0
20
You Other
-13
0
0
20
Based on the choices in the example above, for the first decision the decision-maker
would receive $25 and the other person $0. In the second decision the decision-maker
would receive $0 and the other $20. In the third decision the decision-maker would
have to pay $13 and the other would receive $0.
Please try to make each of your decisions independently of your other decisions. That
is, try not to let any decision you make influence any of the other decisions you make.
As you work through the decisions, assume that you cannot share any money you
receive with the other person and that they cannot share with you. Also assume that
neither the other person nor anyone else will know what choices you make. This
survey is anonymous so even the experimenter will not be able to connect your
decisions to your identity.
In the following pages, you will be faced with a series of such decisions. For each
decision, please circle your preferred option. There are no “right” or “wrong” answers
to the questions. Please respond based on what feels appropriate to you.
130
Please think of a same-sex acquaintance who is not a close friend, but is someone you
see on a regular basis.
Please write their initials in this space: _________________
Keep this person in mind as you answer the following questions.
131
Please answer the following questions with regard to the same-sex acquaintance who
is not a close friend but you see regularly.
Their initials: ______
As you work through the decisions, complete one column at a time.
Option 1
Option 2
You
54
0
Other
0
37
Option 1
Option 2
You
-8
0
Other
0
23
Option 1
Option 2
You
-26
0
Other
0
75
Option 1
Option 2
You
46
0
Other
0
37
Option 1
Option 2
You
-3
0
Other
0
23
Option 1
Option 2
You
-11
0
Other
0
75
Option 1
Option 2
You
39
0
Other
0
37
Option 1
Option 2
You
1
0
Other
0
23
Option 1
Option 2
You
4
0
Other
0
75
Option 1
Option 2
You
31
0
Other
0
37
Option 1
Option 2
You
6
0
Other
0
23
Option 1
Option 2
You
19
0
Other
0
75
Option 1
Option 2
You
24
0
Other
0
37
Option 1
Option 2
You
10
0
Other
0
23
Option 1
Option 2
You
34
0
Other
0
75
Option 1
Option 2
You
17
0
Other
0
37
Option 1
Option 2
You
15
0
Other
0
23
Option 1
Option 2
You
49
0
Other
0
75
Option 1
Option 2
You
9
0
Other
0
37
Option 1
Option 2
You
20
0
Other
0
23
Option 1
Option 2
You
64
0
Other
0
75
Option 1
Option 2
You
2
0
Other
0
37
Option 1
Option 2
You
24
0
Other
0
23
Option 1
Option 2
You
79
0
Other
0
75
Option 1
Option 2
You
-6
0
Other
0
37
Option 1
Option 2
You
29
0
Other
0
23
Option 1
Option 2
You
94
0
Other
0
75
Option 1
Option 2
You
-13
0
Other
0
37
Option 1
Option 2
You
33
0
Other
0
23
Option 1
Option 2
You
109
0
Other
0
75
132
Please answer the following questions with regard to the same-sex acquaintance who
is not a close friend but you see regularly.
Their initials: ______
As you work through the decisions, complete one column at a time.
Option 1
Option 2
You
54
0
Other
0
37
Option 1
Option 2
You
-8
0
Other
0
23
Option 1
Option 2
You
-26
0
Other
0
75
Option 1
Option 2
You
46
0
Other
0
37
Option 1
Option 2
You
-3
0
Other
0
23
Option 1
Option 2
You
-11
0
Other
0
75
Option 1
Option 2
You
39
0
Other
0
37
Option 1
Option 2
You
1
0
Other
0
23
Option 1
Option 2
You
4
0
Other
0
75
Option 1
Option 2
You
31
0
Other
0
37
Option 1
Option 2
You
6
0
Other
0
23
Option 1
Option 2
You
19
0
Other
0
75
Option 1
Option 2
You
24
0
Other
0
37
Option 1
Option 2
You
10
0
Other
0
23
Option 1
Option 2
You
34
0
Other
0
75
Option 1
Option 2
You
17
0
Other
0
37
Option 1
Option 2
You
15
0
Other
0
23
Option 1
Option 2
You
49
0
Other
0
75
Option 1
Option 2
You
9
0
Other
0
37
Option 1
Option 2
You
20
0
Other
0
23
Option 1
Option 2
You
64
0
Other
0
75
Option 1
Option 2
You
2
0
Other
0
37
Option 1
Option 2
You
24
0
Other
0
23
Option 1
Option 2
You
79
0
Other
0
75
Option 1
Option 2
You
-6
0
Other
0
37
Option 1
Option 2
You
29
0
Other
0
23
Option 1
Option 2
You
94
0
Other
0
75
Option 1
Option 2
You
-13
0
Other
0
37
Option 1
Option 2
You
33
0
Other
0
23
Option 1
Option 2
You
109
0
Other
0
75
133
Please think of your closest same-sex friend.
Please write their initials in this space: _________________
Keep this person in mind as you answer the following questions.
134
Please answer the following questions with regard to your closest same-sex friend.
Their initials: ______
As you work through the decisions, complete one column at a time.
Option 1
Option 2
You
54
0
Other
0
37
Option 1
Option 2
You
-8
0
Other
0
23
Option 1
Option 2
You
-26
0
Other
0
75
Option 1
Option 2
You
46
0
Other
0
37
Option 1
Option 2
You
-3
0
Other
0
23
Option 1
Option 2
You
-11
0
Other
0
75
Option 1
Option 2
You
39
0
Other
0
37
Option 1
Option 2
You
1
0
Other
0
23
Option 1
Option 2
You
4
0
Other
0
75
Option 1
Option 2
You
31
0
Other
0
37
Option 1
Option 2
You
6
0
Other
0
23
Option 1
Option 2
You
19
0
Other
0
75
Option 1
Option 2
You
24
0
Other
0
37
Option 1
Option 2
You
10
0
Other
0
23
Option 1
Option 2
You
34
0
Other
0
75
Option 1
Option 2
You
17
0
Other
0
37
Option 1
Option 2
You
15
0
Other
0
23
Option 1
Option 2
You
49
0
Other
0
75
Option 1
Option 2
You
9
0
Other
0
37
Option 1
Option 2
You
20
0
Other
0
23
Option 1
Option 2
You
64
0
Other
0
75
Option 1
Option 2
You
2
0
Other
0
37
Option 1
Option 2
You
24
0
Other
0
23
Option 1
Option 2
You
79
0
Other
0
75
Option 1
Option 2
You
-6
0
Other
0
37
Option 1
Option 2
You
29
0
Other
0
23
Option 1
Option 2
You
94
0
Other
0
75
Option 1
Option 2
You
-13
0
Other
0
37
Option 1
Option 2
You
33
0
Other
0
23
Option 1
Option 2
You
109
0
Other
0
75
135
Initials of closest same-sex friend: ______
Option 1
Option 2
You
28
0
Other
0
19
Option 1
Option 2
You
-16
0
Other
0
46
Option 1
Option 2
You
99
0
Other
0
68
Option 1
Option 2
You
24
0
Other
0
19
Option 1
Option 2
You
-7
0
Other
0
46
Option 1
Option 2
You
85
0
Other
0
68
Option 1
Option 2
You
20
0
Other
0
19
Option 1
Option 2
You
2
0
Other
0
46
Option 1
Option 2
You
71
0
Other
0
68
Option 1
Option 2
You
16
0
Other
0
19
Option 1
Option 2
You
12
0
Other
0
46
Option 1
Option 2
You
58
0
Other
0
68
Option 1
Option 2
You
12
0
Other
0
19
Option 1
Option 2
You
21
0
Other
0
46
Option 1
Option 2
You
44
0
Other
0
68
Option 1
Option 2
You
9
0
Other
0
19
Option 1
Option 2
You
30
0
Other
0
46
Option 1
Option 2
You
31
0
Other
0
68
Option 1
Option 2
You
5
0
Other
0
19
Option 1
Option 2
You
39
0
Other
0
46
Option 1
Option 2
You
17
0
Other
0
68
Option 1
Option 2
You
1
0
Other
0
19
Option 1
Option 2
You
48
0
Other
0
46
Option 1
Option 2
You
3
0
Other
0
68
Option 1
Option 2
You
-3
0
Other
0
19
Option 1
Option 2
You
58
0
Other
0
46
Option 1
Option 2
You
-10
0
Other
0
68
Option 1
Option 2
You
-7
0
Other
0
19
Option 1
Option 2
You
67
0
Other
0
46
Option 1
Option 2
You
-24
0
Other
0
68
136
Relationship Quality Index, Acquaintance Version
[Subjects did not see a title to this scale or any other scale]
Please answer the following questions with regard to the same-sex acquaintance you
listed before. Their initials: ______
How much do you like this person?
1
2
3
Not at All
4
5
6
Somewhat
7
Very Much
How close do you feel towards this person?
1
2
3
Not at All
4
5
6
7
Somewhat
Very
How much could you rely on this person if you needed help?
1
2
3
Not at All
4
5
6
Somewhat
7
Very Much
How much could this person rely on you if they needed help?
1
2
3
Not at All
4
5
6
Somewhat
7
Very Much
How willing would you be to do this person a favor?
1
2
3
Not at All
4
5
6
Somewhat
7
Very
How willing would this person be to do a favor for you?
1
Not at All
2
3
4
5
Somewhat
6
7
Very
137
Appendix 2: Screen Shots of Computerized Tasks
138
Appendix 3: Materials for Study 3
Script for Payment Condition
Hi, thank you for coming today. My name is ______ and I’m going to be
running today’s experiment. In today’s experiment, you’re going to be making a
series of monetary decisions involving yourself and several other people and filling
out some other survey questions.
Before we begin, please read and sign this consent form.
The first thing we need to do is determine who you will be making decisions
about. We need you to select your closest-same sex friend and a same-sex
acquaintance. However, they need to be people who are listed in the student directory.
Please follow the instructions on the screen to do this. As you look through the
directory, please mark the pages that list your friend and acquaintance with a stickynote. Raise your hand when you are done and I’ll come and introduce the next part of
the experiment. (Go away and wait for them to raise their hand.)
Okay, now let me show you how a decision works. (Hold up a large
example.) Each decision includes two options. In the first option, you will get a sum
of money and the other person will receive no money. In the second option, the other
person will get a sum of money and you will get no money. All you need to do is
click the option you’d prefer. There is no right or wrong answer, just pick whatever
feels appropriate to you.
In some of the cases, the sum of money for you will be negative. (Hold up
example.) This means that you would have to pay money if you choose option 1. So if
you choose option 1, the other person would get no money and you would lose
money. If you choose option 2, you get no money and the other person gets a sum.
You are going to be presented with many decisions like these. You, your
friend, or your acquaintance may receive actual money (on top of your show-up fee)
based on the decisions you make.
Let me explain how payment works. Please follow along on the sheet. No
matter what happens, you will receive $10 for coming today. After the experiment is
over, you will role these dice. (Show them the dice.) If you roll double 6’s, two things
happen. First, you will receive 30 additional dollars to your show up fee. Second, you
will randomly select one of the decisions you made and we will make that one
decision come true. Either you, your friend, or your acquaintance will receive money
based on the decision you randomly select. Any gains or losses that you personally
make will be added to or subtracted from the $30.
Because of this, please take all the decisions seriously. Make them as if they
were all real because you don’t know which one actually might be real.
Just to make sure we’re ready to begin, I need to ask you a few questions:
(Show the first example): What happens if you choose option 1? If you choose
option 2?
(Show the second example): What happens if you choose option 1? Option 2?
Do any of your decisions actually affect you or others monetarily?
139
What happens if you roll double 6s?
What happens if you roll anything but double 6s?
Your participation is entirely voluntary and you are free to withdraw from the
study at any time without losing your research credit or being penalized in any way.
Your responses will not be associated with your identity. That is, you will remain
anonymous.
The computer will have all the instructions you need from this point on. If you
have any questions as you go along, please let me know.
Script for No Payment Condition
Hi, thank you for coming today. My name is ______ and I’m going to be
running today’s experiment. In today’s experiment, you’re going to be making a
series of monetary decisions involving yourself and several other people and filling
out some other survey questions.
Before we begin, please read and sign this consent form.
The first thing we need to do is determine who you will be making decisions
about. We need you to select your closest-same sex friend and a same-sex
acquaintance. However, they need to be people who are listed in the student directory.
Please follow the instructions on the screen to do this. As you look through the
directory, please mark the pages that list your friend and acquaintance with a stickynote. Raise your hand when you are done and I’ll come an introduce the next part of
the experiment. (Go away and wait for them to raise their hand.)
Okay, now let me show you how a decision works. (Hold up a large
example.) Each decision includes two options. In the first option, you will get a sum
of money and the other person will receive no money. In the second option, the other
person will get a sum of money and you will get no money. All you need to do is
click the option you’d prefer. There is no right or wrong answer, just pick whatever
feels appropriate to you.
In some of the cases, the sum of money for you will be negative. (Hold up
example.) This means that you would have to pay money if you choose option 1. So if
you choose option 1, the other person would get no money and you would lose
money. If you choose option 2, you get no money and the other person gets a sum.
You are going to be presented with many decisions like these. The decisions
are hypothetical and you, your friend, and your acquaintance will not receive any
money based on them.
Let me explain how payment works. Please follow along on the sheet. No
matter what happens, you will receive $10 for coming today. After the experiment is
over, you will role these dice. (Show them the dice.) If you roll double 6’s, you will
receive 30 additional dollars to your show-up fee.
Although the decisions are hypothetical, please take them all seriously. Please
try to make them as if they were all real because you don’t know which one actually
might be real.
Just to make sure we’re ready to begin, I need to ask you a few questions:
140
(Show the first example): What happens if you choose option 1? If you choose
option 2?
(Show the second example): What happens if you choose option 1? Option 2?
Do any of your decisions actually affect you or others monetarily?
What happens if you roll double 6s?
What happens if you roll anything but double 6s?
Your participation is entirely voluntary and you are free to withdraw from the
study at any time without losing your research credit or being penalized in any way.
Your responses will not be associated with your identity. That is, you will remain
anonymous.
The computer will have all the instructions you need from this point on. If you have
any questions as you go along, please let me know.
Examples
You Other
Option 1
25
0
Option 2
0
20
You Other
Option 1
-13
0
Option 2
0
20
141
Appendix 4: Material for Study 4
Verbal Script
Hi, thanks for coming today my name is ________ and I’ll be running today’s study.
In today’s study we’re going to be asking you questions that involve yourself
and several other people. These others might be friends, acquaintances, family
members, or other people you know.
Most the questions you will be answering will involve hypothetical monetary
decisions. Let me give you a couple examples. In this first example, the decision
involves yourself and another person. You task is to choose between two options. In
this example, if you choose option 1, then you receive $25 and the other person
receives nothing. But if you choose option 2, then you receive nothing and the other
receives $20.
Let’s look at another example. Notice in this example, option 1 has a negative
amount for you. If you choose option 1 you would lose money, $13, and the other
person gets nothing. But if you choose option 2, you would get nothing and the other
person would get $20. Do you guys have any questions at this point?
Let’s look at a third example. In this example, the decision involves two other
people, but does not involve yourself. But still, your task is to choose the option that
YOU prefer. Your task task is NOT to guess what you think the other people would
choose. Your task is to choose the decision that you find most desirable. So you
might prefer option 1 which means that Other 1 gets $25 and Other 2 gets nothing. Or
you might prefer option 2 which means that Other 2 gets $20 and Other 1 gets
nothing. Does anyone have any questions about that? I just want to make sure it’s
clear: when a decision involves two other people and not you, your task is still to
choose the option that you find most desirable, not the option you think one of the
other people would choose.
In these questions as well, negative numbers will also appear. Here, too, they
mean that a person would lose money.
As you work through the decisions, please keep something very important in
mind. Do you best to make each decision independently of all the others. Do not let
previous decisions you’ve made influence any decisions you are thinking about. To
make this easier, imagine that at the end of the experiment we randomly select just
one of the decisions that you made and we make this one, single decision real. If only
one decision can become real, you want to make each decision independently of all
others because you just don’t know which decision might be real. In other words,
make each decision as if it were the only decision you were making and you had to
live with that decision alone. This is very important, so does anyone have any
questions at this point?
Just a few more quick things. There are no right or wrong answers: please
make whatever decisions feel appropriate to you. As you make your decisions, please
assume there is no sharing allowed: you cannot give any money you would earn to
the others and they cannot give any money they earn to you or to each other. The
142
decisions are hypothetical, so the people you make decisions about will never know
what those decisions were.
Computerized Instructions
For this experiment we'd like you to think of six different people that you know.
Please try to think of a range of people, from family members, to close friends, to notthat-close friends, to acquaintances, even to people you mildly dislike. Try to think of
a broad range. On the following six screens we'd like you to enter these people's
initials.
We'll be using the initials to refer to these people later, so please do NOT select
people with the same initials. We want to make sure that you know exactly who we're
asking about. Make sure to enter only ONE set of initials on each screen.
We recommend typing initials in all caps so it's easier to read later.
Please enter the initials of ONE person that you know. Try to select from a broad
range of people you know (family members, friends, acquaintances, even people you
don't like that much). Make that you do not use the same person twice or the same set
of initials twice. We'll be using the initials to refer to the person later. We recommend
typing the initials in all caps as it makes it easier to read later.
Remember, enter the initials of only ONE person on this screen.
You have already entered:
<person2> <person3> <person4> <person5> <person6>
In the following questions, you are given a decision between two options. The
decisions are all hypothetical, but please imagine that your choice determines how
much money you and several other people will receive.
Some of the decisions will involve yourself and one other person. If you choose one
of the options, a sum of money will be given to you and no money will be given to
the other person. If you choose the other option, a sum of money will be given to the
other person and no money will be given to you.
For example, consider the decisions below:
Option 1
Option 2
You Other
25
0
0
20
143
You Other
15
0
0
20
You Other
-13
0
0
20
In the first decision, you must choose between $25 for you or $20 for the other
person. If you choose the first option, then you get $25 and the other person gets $0.
If you choose the second option, you get $0 and the other person gets $20.
In the third decision, you must choose between -$13 for you or $20 for the other
person. If you choose the first option, then you would have to PAY $13 and the other
person gets zero. Choosing a negative value for yourself means you LOSE money. If
you choose the second option then you get $0 and the other person gets $20.
As an example, consider these decisions made by a hypothetical decision-maker:
You Other
Option 1 25
0
Option 2
0
20
You Other
15
0
0
20
You Other
-13
0
0
20
Based on the choices in the example above, for the first decision the decision-maker
would receive $25 and the other person $0. In the second decision the decision-maker
would receive $0 and the other $20. In the third decision the decision-maker would
have to PAY $13 and the other would receive $0.
Other decisions may involve two other people, but NOT you. In these decisions, we
would still like you to pick the option that YOU would prefer. Do not worry about
what the other people would prefer.
For example, consider the decisions below:
Other 1 Other 2
Other 1 Other 2
Other 1 Other 2
Option 1
25
0
15
0
-13
0
Option 2
0
20
0
20
0
20
In these decisions, your task is to choose the option that YOU prefer. So in the first
decision above, you may prefer to give Other 1 $25 and Other 2 $0. If so, you would
select Option 1. Or you may prefer to give Other 2 $25 and Other 1 $0. If so, you
would select Option 2. Remember, please do not answer based on what you think the
other people would choose.
Negative numbers work the same here as in decisions involving yourself. If you
choose an option with a negative number, one person would lose money and the other
person would receive nothing.
Please make sure to pay attention because which option is on top and which is on the
bottom will change throughout the experiment.
As you work through the decisions, keep the following in mind--it is very important:
Please make all of your decision independently of the others. DO NOT LET ANY
144
DECISIONS INFLUENCE ANY OTHER DECISIONS. To help you do that,
imagine that at the end of the experiment we randomly select JUST ONE of the
decisions you made and we make only that ONE decision come true. Because only
one of your decisions would become real in this scenario, please make each of your
decisions independently of your other decisions. After all, if only one can come true,
you should make each decision as if it were the only one you were making. Make
every decision as if you would have to live with that decision made in isolation.
As you work through the decisions, assume that you CANNOT share any money you
receive with other people and that they CANNOT share with you or with each other.
Also assume that neither the other people nor anyone else will know what choices
you make. This survey is anonymous so even the experimenter will not be able to
connect your decisions to your identity.
In the following screens, you will be faced with a series of such decisions. For each
decision, please choose your preferred option. There are no "right" or "wrong"
answers to the questions. Please respond based on what feels appropriate to you.
Please make sure to pay attention because which option is on top and which is on the
bottom will change throughout the experiment.
145
Appendix 5: Materials for Study 5
In the following questions, you are given a decision between two options. The
decisions are all hypothetical, but please imagine that your choice determines how
much money you and your [friendacq], [friacqInitials], will receive.
The decisions will involve yourself and this person. If you choose one of the options,
a sum of money will be given to you and no money will be given to the other person.
If you choose the other option, a sum of money will be given to the other person and
no money will be given to you.
Press any key to continue.
(end screen)
For example, consider the three decisions below:
25 vs 20
15 vs 20
-13 vs 20
In the first decision, you must choose between $25 for you or $20 for the other
person. If you choose the first option, then you get $25 and the other person gets $0.
If you choose the second option, you get $0 and the other person gets $20.
In the third decision, you must choose between -$13 for you or $20 for the other
person. If you choose the first option, then you would have to PAY $13 and the other
person gets zero. Choosing a negative value for yourself means you LOSE money. If
you choose the second option then you get $0 and the other person gets $20.
(end screen)
As an example, consider these decisions made by a hypothetical decision-maker.
The decision-maker's preferred choices have been emphasized.
25 vs 20
15 vs 20
-13 vs 20
Based on the choices in the example above, for the first decision the decision-maker
would receive $25 and the other person $0. In the second decision the decision-maker
would receive $0 and the other $20. In the third decision the decision-maker would
have to PAY $13 and the other would receive $0.
(end screen)
146
As you work through the decisions, keep the following in mind--it is very important:
Please make all of your decisions independently of the others. DO NOT LET ANY
DECISIONS INFLUENCE ANY OTHER DECISIONS.
To help you do that, imagine that at the end of the experiment we randomly select
JUST ONE of the decisions you made and we make only that ONE decision come
true. Because only one of your decisions would become real in this scenario, please
make each of your decisions independently of your other decisions. After all, if only
one can come true, you should make each decision as if it were the only one you were
making.
Make every decision as if you would have to live with that decision made in isolation.
(end screen)
As you work through the decisions, assume that you CANNOT share any money you
receive with the other person and that they CANNOT share with you. Also assume
that neither the other person nor anyone else will know what choices you make. This
survey is anonymous so even the experimenter will not be able to connect your
decisions to your identity.
(end screen)
In the following screens, you will be faced with a series of such decisions. For each
decision, please choose your preferred option. There are no "right" or "wrong"
answers to the questions. Please respond based on what feels appropriate to you.
Please make sure to pay attention because which option is on top and which is on the
bottom will change throughout the experiment.
As you work through these questions, you may NOT write anything down.
(end screen)
You will now make decisions regarding yourself and the [friendacq], [friacqInitials],
who came with you to the experiment.
(end screen)
In the following questions you will be asked to judge which of two sums of money is
larger. However, the sums are not expressed in the same currency. One sum will be
expressed in United States dollars, the other will be expressed in euros. (The euro is
the unit of currency in many European countries.)
147
(end screen)
To help you determine which sum is larger, we will give you an exchange rate for the
two currencies. An exchange rate allows you to convert one currency to another. We
will always be giving you an exchange rate that allows you to convert euros to
dollars.
For example, assume that the exchange rate that converts euros to dollars is [oneHalf]
(one half). (We are only using [oneHalf] as an example. When we ask you questions,
it is very likely the exchange rate will be something other than [oneHalf].)
If the two sums of money were 5 US dollars and 20 euros, you would determine
which sum was greater as follows.
First, multiply the number of euros by the exchange rate:
20 euros * [oneHalf] = 10 dollars
Then compare the number of US dollars (5) with the number of euros expressed as
dollars (10). As you can see, the sum originally expressed as euros is a larger sum of
money. So in this example you would select 20 euros as the larger.
(end screen)
Let’s do 3 more examples. In each case, your task is to select the option that has the
larger sum of money.
20 vs 30
5 vs 14
-15 vs 22
In the first example, you must decide whether 20 dollars is greater or smaller than 30
euros.
In the second example, you must decide if 5 dollars is greater or smaller than 14
euros.
In the third example, you must decide if -15 dollars is greater or smaller than 22
euros. Be sure to notice when one of the numbers is NEGATIVE. You can think of 15 dollars as a loss of 15 dollars.
Which options do you think have the largest sum of money.
(end screen)
We’ve highlighted the correct answers in red.
The first example asks which is bigger, 20 US dollars or 30 euros? Our exchange rate
is [oneHalf] (one half), so we multiply 30 euros by [oneHalf], which equals 15 US
dollars. This means that 30 euros are worth 15 US dollars. The sum of 20 US dollars
148
available in Option 1 is greater than 15, so in this example you would select Option 1
as having the greater amount of money.
In the second example, we multiply 14 euros by [oneHalf], which equals 7 US
dollars. 7 dollars is greater than 5 dollars, so in this example you would select Option
2 as having the greater amount of money.
In the third example, we multiply 22 euros by [oneHalf], which equals 11 US dollars.
11 dollars is greater than -15 dollars, so in this example you would select Option 2 as
having the greater amount of money. Be sure to notice whether a sum is negative.
Although 15 dollars is greater than 11 dollars, NEGATIVE 15 dollars is less than 11
dollars.
(end screen)
Although there is an actual exchange rate in the real world that allows you to convert
euros into dollars, we will not be using that exchange rate in this experiment.
Instead, we will be giving you a hypothetical exchange rate. For you, the exchange
rate will be:
[exchangeRate]
We will present the exchange rate along with every question, so you do not need to
memorize it.
(end screen)
As you work through these questions, please DO NOT write anything down.
(end screen)
If you are not sure of the answer, please take your best guess.
Please make sure to pay attention because which option is on top and which is on the
bottom will change throughout the experiment.
(end screen)
We would now like you to answer another set of questions about which of two sums
is larger. As before one sum is expressed as US dollars and the other as euros.
There are two changes that make this set of questions different from the first set:
First, the exchange rate will be presented as a [secondRate] instead of as a [firstRate].
149
Second, we will be giving you a new exchange rate to use in the questions.
Your new exchange rate is:
[exchangeRate]
Following this screen are all the original instructions, with the new [secondRate]
format of the exchange rate used in the examples. Please review the instructions
before you begin.
150
Appendix 6: Probing the Sensitivity of Welfare Tradeoff Computations
The experiments of Chapter 2 showed that welfare tradeoff decisions are
made in a consistent fashion and presented evidence to rule out the possibility that
WTR decisions were being implemented by conscious manipulation of numbers. For
instance, despite decisions being randomly mixed together, Study 5 showed that
subjects’ WTRs subtly decreased as the total amount at stake was increased. If WTR
decisions were being produced by the conscious manipulation of numbers in an
attempt to be consistent, why go through the trouble of systematically changing them
as a function of different absolute magnitudes—especially when decisions of different
magnitudes are all randomly mixed? The goal of the experiment presented in this
appendix is to further probe the sensitivity of welfare tradeoff computations to
apparently minor differences between different decisions, even when the different
decisions are all randomly mixed.
In the original “standard” version of the WTR task (“Do you want $5 to be
allocated to you or $10 to be allocated to your friend?), both possible allocations
involve one person staying at their baseline level of utility and the other person
gaining utility. Although this describes one type of social interaction, there may be
fundamentally different types of social interactions, each one defined by how
different actors are moved up or down from their baseline utility (see also the section
“The Elementary Forms of Social Interaction” in Chapter 4). For example, imposing a
cost brings a person’s utility below what it was before a decision, and therefore
affects them in a way that failing to change their utility does not. The consequences of
151
moving a person to a utility lower than in the absence of interacting with you may be
more serious than not affecting them at all. Such a stance is consistent with optimal
foraging theory (Rode, et al., 1999; Stephens, et al., 2007) and may explain why
psychologists routinely find that losses and gains are treated differently (e.g., Tversky
& Kahneman, 1992). Thus, imposing a cost on someone to benefit yourself might
well lead to different inferences on the part of the other person, provoking different
consequences (e.g., aggression), compared to failing to confer a benefit to gain a
benefit (which their utility unchanged), or even inflicting a cost to avoid a cost (since
in this case there is no way to avoid a situation where someone’s utility will go below
baseline).
Thus, as very simple test of these ideas, I modified the standard version of the
WTR task, preserving the sign and magnitude of the ratio between the allocations, but
changing whether the allocations represented positive or negative deviations from a
person’s baseline. An indifference to the alternatives in all four of these variations
would imply identical WTRs. But would the mind actually treat these tasks
identically? In the “loss” version of the WTR task, both deviations are negative
(allocate a loss of $5 to yourself or a loss of $10 to your friend?). In this version,
someone’s utility must be decreased, making it symmetric to the standard version,
which requires that someone’s utility increase. In the “entitlement” version, the
deviation for the self is positive and the deviation for the other is negative (allocate a
gain of $10 to yourself and simultaneously a loss of $10 to your friend, or leave both
at baseline?). Causing another person to incur a loss has potentially more severe
152
consequences (e.g. eliciting aggression) than forgoing an opportunity to allocate them
a gain. This may lead the mind to be more generous—to express a higher WTR—in
the entitlement version of the task. In the “help” version, the deviation for the self is
negative and the deviation for the other is positive (allocate a loss of $5 to yourself
and simultaneously a gain of $10 to your friend, or leave both of you at baseline?).
The mirror image of the entitlement task, the help task asks subjects to actively move
themselves below baseline to increase another’s utility. Because this is especially
costly for the self—and forgoing it would leave the other’s utility unchanged—this
may lead the mind to be more selfish—to express a lower WTR—in the help version
of the task. Because the differences between formats might depend on the nature of
the subjects’ relationships with the target, I also examined four different types of
relationships.
To test these ideas, a group of 52 subjects at the University of California,
Santa Barbara, completed the welfare tradeoff task used in Chapter 2, with several
important differences. First, each of the specific 60 choices (e.g. $54 for the self or
$37 for the other) was completed four times, once each in one of the above formats.
Importantly, the instructions for the task did not describe it as involving four types of
decisions. Instead, subjects learned that they would be making a number of decisions
and learned how the mechanics of gaining or losing money based on the sign of the
number worked. Second, for a specific target, all of the choices across all the formats,
were randomly mixed together, presented one-by-one by a computer. This means that
each subject made 240 decisions about, e.g., their friend. Third, all subjects answered
153
about an acquaintance, their best friend, and their mother (or “the female relative who
raised” them, if they were not raised by their biological mother). Fourth, a subset of
20 subjects also completed the task regarding an enemy (described as “someone you
do not like and feel is against you, a person you may even feel is an enemy”). The
order of the first three targets was randomized. If a subject was asked about an
enemy, this always came last. I included the enemy condition to further explore the
boundaries of WTR computation. Unfortunately, because of the large number of
decisions subjects were making, most did not have time within the allotted
experimental session to complete the enemy portion. (Regardless, subjects’ WTRs
toward an enemy were uniformly very low.)
Despite the four formats being randomly mixed together within a target, are
there systematic differences between how subjects complete the standard, loss, help,
and entitlement versions of the task? Yes, as can be seen in Figure 4, subjects clearly
revealed different welfare tradeoff ratios not only as a function of target, but,
importantly, as a function of the format.
Subjects’ sensitivity to the format and the target was tested statistically by
using repeated-measures ANOVAs with target and format as factors and WTR scores
as the dependent variable. Using data from all subjects (hence excluding data about
enemies), there was a main effect of target type (Wilks’ λ = .40, F(2,50) = 37.67, p =
10-9, partial η2 = .60), a main effect of format (Wilks’ λ = .23, F(3,49) = 54.46, p =
10-14, partial η2 = .77), and an interaction between the two (Wilks’ λ = .51, F(6,46) =
7.38, p = 10-4, partial η2 = .49). This was also true using only subjects who also had
154
data regarding enemies (hence including the enemy data): There was a main effect of
target type (Wilks’ λ = .13, F(3,17) = 39.39, p = 10-7, partial η2 = .87), a main effect
of format (Wilks’ λ = .23, F(3,17) = 18.72, p = 10-4, partial η2 = .77), and an
interaction between the two (Wilks’ λ = .18, F(6,11) = 5.57, p = .005, partial η2 =
.82).
155
1.4
1.2
1
WTR
0.8
Standard
0.6
Loss
Entitle
Help
0.4
0.2
0
-0.2
Enemy
Acq
Friend
Mom
-0.4
Figure 4. Revealed welfare tradeoff ratios as a function of target and format. Error
bars represent ±SEM.
Put in words, subjects were most generous toward their mothers, moderately
generous toward their friends, and the least generous toward their acquaintances.
Moreover, subjects had negative WTRs toward their enemies, meaning (e.g.) that
subjects would incur a cost to avoid allocating money to their enemy. The entitlement
version elicited the most generosity: Subjects were relatively unwilling to reduce
another person’s utility below baseline even if that would increase the subject’s
utility. The help version elicited the least generosity: Subjects were relatively
unwilling to reduce their own utility below baseline even if that would increase
another’s utility. The standard and loss version both revealed moderate and roughly
equivalent levels of generosity. The interaction between format and target is largely
driven by the help format: The relationship with the target (mom vs. friend vs.
156
acquaintance) makes relatively little difference in expressed WTRs on the help task.
For all other tasks, subjects’ expressed WTRs are strongly affected by the nature of
the target.
That there was a significant interaction, even when the data for enemies is
excluded, suggests that target and format are not simply additive effects. Instead, the
system for computing welfare tradeoffs simultaneously integrates both pieces of
information in its decision-making process. This speaks to an important question:
Does the mind really compute WTRs or does it only compute the upstream factors
(such as kinship indices) and simply add or average these factors when making
welfare tradeoffs? If the latter were true, then we would expect only additive effects
of manipulating the format of the decisions. This would occur because the internal
representations only add or subtract from one another to arrive at a final decision. By
contrast, the mechanisms for making welfare tradeoffs are hypothesized to integrate
multiple pieces of information, with some information potentially enhancing,
moderating, or attenuating the effects of other information on the final output.
Subjects clearly distinguish different formats and targets from each other.
Despite the randomization, if we examine their decisions within a given format and
target, do they make their decisions consistently? Yes: Analyzing the consistency
maximization scores shows that subjects are consistent in their responses (see Table
6). Relative to the random null of 70.84%, consistency maximization scores were
always at least 93% (all rs relative to the random null > .93, all ps < 10-10). Analyzing
the perfect consistency scores shows also that subjects are consistent in their
157
responses (see Table 6). Relative to the random null of 1.1%, perfect consistency
scores were always at least 57% (all rs relative to the random null > .84, all ps < 10-8).
Note that consistency scores, especially on the perfect consistency measure, were
lower in this study in than in the main studies. This may have occurred for at least
two reasons: First, subjects answered so many more questions in this study than in the
main studies that their attention may have wavered, leading to less consistency.
Second, the welfare tradeoff system may not be designed to switch so quickly
between decisions with different formats, leading to a decrement in performance.
Nonetheless, subject clearly distinguished between the different formats despite the
randomization within a target. The welfare tradeoff system appears sensitive to even
apparently minor variations on the task.
Table 6
Consistency Score Means (Standard Deviations) as a Function of Target and Format
Consistency Maximization
Mom
Friend
Acquaintance
Enemy
Perfect Consistency
Mom
Friend
Acquaintance
Enemy
Standard
94% (07%)
94% (07%)
95% (07%)
96% (07%)
Loss
96% (06%)
94% (07%)
96% (06%)
97% (05%)
Entitlement
95% (07%)
93% (09%)
93% (07%)
95% (08%)
Help
95% (07%)
95% (08%)
97% (06%)
98% (06%)
59% (32%)
61% (34%)
70% (31%)
78% (30%)
70% (30%)
60% (31%)
75% (30%)
78% (27%)
70% (36%)
58% (38%)
57% (31%)
70% (29%)
67% (31%)
70% (35%)
80% (26%)
88% (23%)
158
Appendix 7: Dissociating Welfare Tradeoff Computation from Other Intuitive
Economic Computation: A Pilot Functional Magnetic Resonance Imaging
Study3
The findings in Chapter 2 showed that the welfare tradeoff task dissociated
from a logically identical task involving explicit conscious manipulation of
magnitudes (the currency exchange task). But these data leave open an additional
alternative to the hypothesis that the mind contains specialized mechanisms for
computing welfare allocations and tradeoffs: Perhaps the mind has a general
economic preference calculator. Although no one has proposed a computational
model of a single mechanism that could adaptively compute all possible economic
valuations, if such a mechanism did exist, it would operate over welfare tradeoffs. But
it would also compute the valuation of hot dogs, of donating to charity, of investing in
a dot-com, or of going Dutch on a date. Showing that welfare tradeoff calculations are
handled, in part, by a separate system could not necessarily be shown by the type of
data (e.g., reaction time data) used in Chapter 2. Even if there are numerous separate
systems for computing different types of economic valuation, they may all operate
with considerable speed and make all such calculations feel phenomenologically
intuitive.
Previous work using neuroimaging and behavioral measures has dissociated
time preferences from risk preferences (Weber & Huettel, 2008) and work using only
behavioral measures has dissociated WTR-like preferences from time preferences
3
Scott Grafton collaborated in design of the study presented here. Any errors of analysis or
interpretation are completely mine, however.
159
(Rachlin & Jones, 2008b), suggesting that separate systems are involved. To directly
address this issue, I instead turned to a very different approach from my previous
studies: functional magnetic resonance imaging (fMRI). Briefly, fMRI measures
changes in regional BOLD (blood-oxygenation-level dependent) signal in the brain.
Greater BOLD signal implies that a population of neurons has been activated
(Hashemi, Bradley, & Lisanti, 2004). For my purposes, I imaged subjects while they
completed several experimental tasks. If a population of neurons shows significantly
greater BOLD signal for one task relative to another, that would imply that the first
task preferentially activates those neurons. This would demonstrate a neural
dissociation between the two tasks.
In this study, I compared the welfare tradeoff task to two versions of a
temporal discounting task (see Chapter 2 on temporal discounting; details of the
method are given at the end of this appendix). The basics of the welfare tradeoff task
were retained from previous studies. Temporal discounting is an appropriate
comparison because it is also an economic preference and tradeoffs involving
temporal discounting are made intuitively; i.e., subjects do not consult a conscious
magnitude; they instead choose what “feels” right. The first temporal discounting task
asked subjects to choose between an earlier smaller benefit and a larger later benefit
as they the subject would want (self temporal discounting). For instance, in the self
temporal discounting task, subjects would need to decide whether they want $54
tomorrow or $98 in six months. The other temporal discounting task asked subjects to
make the same choices except to complete them as they believe their closest friend
160
would want (friend temporal discounting). For example, in the friend temporal
discounting task, subjects would need decide if their friend would want $54 tomorrow
or $98 in six months. This latter task not only controls for the economic nature of the
task, but also for the presence of another social agent. If a brain region is
differentially active in the WTR task relative to both temporal discounting controls,
then this would strongly suggest that the different task are handled by different
neurocognitive systems.
Before proceeding to the data, I note a major caveat: This study is only a pilot
study consisting of just eight subjects and its results must be treated with extreme
caution and only interpreted as preliminary.
Despite this, do the data support a neurocognitive dissociation between
welfare tradeoffs and temporal tradeoffs? First, note that reaction time data in this
experiment did not reveal that the WTR task was either substantially easier or more
difficult than the temporal discounting tasks. Subjects completed WTR decisions on
average in 1.61sec (95% Confidence Interval: 1.22 to 2.00). Moreover, they
completed the self temporal discounting task on average in 1.59sec (95% CI: 1.15 to
2.03) and the friend temporal discounting task on average in 1.71sec (95% CI: 1.22 to
2.19). (As a point of comparison, in Study 5 of Chapter 2, subjects completed the
WTR in 1.8sec on average, but took about 3.1sec on average to complete the currency
exchange tasks.) Interestingly, a paired-samples t-test revealed the friend temporal
discounting task to be significantly slower than the self version, t(7) = 2.94, p = .041.
The differences between the WTR and self temporal discounting task and the WTR
161
and other temporal discounting task were not significant, ps > .38. Thus, as expected,
the three tasks employed in this pilot imaging study are roughly equivalent in ease for
the subjects. Moreover, the WTR task was neither completed the fastest nor the
slowest.
Consistent with the hypothesis that there is a specialized system for
computing welfare tradeoffs, a group-level analysis showed that the welfare tradeoff
task was associated with significant activation in right dorsomedial prefrontal cortex
(rDMPFC), Brodmann’s area (BA) 8, relative to both the self and friend temporal
discounting tasks (see Figure 5 and the first rows of the top and bottom halves of
Table 7). The peak activation for the cluster revealed in the self temporal discounting
comparison was located at (23, 37, 44) and for the cluster revealed in the friend
comparison at (21, 31, 49).
The WTR/self discounting and WTR/other discounting activations are roughly
8mm apart. Do they really represent activation in the “same” place? Although fMRI
involves measuring activity at many thousands of voxels (cube-shaped “volume
elements” representing tiny portions of the brain), voxels are not statistically
independent. Instead, there are only a much smaller number of statistically
independent regions within the brain, called resels (for “resolving elements”). (Resels
are not actual physical regions of the brain; instead, they are a theoretical estimate of
the number of independent comparisons.) If two points fall within the same resel,
then they represent activation of the “same” region of the brain. For the self temporal
discounting comparison, the size of a resel is 239 voxels, with each voxel being 2mm
162
on a side. If a resel was perfectly spherical, then two points could be at most 16mm
apart. (A spherical shape works against concluding that two points belong to the same
resel.) For the friend comparison, the size of a resel is 204 voxels, implying that two
points could be at most 14mm apart. In either case, the two peak activations in the
rDMPFC in the self and friend comparisons are closer than this, suggesting that they
do lie within the “same” region of the brain and may represent identical activations.
(The calculations about maximum distance used the native coordinates calculated by
SPM—not the Talairach coordinates presented in the tables—because the SPM
coordinates are the coordinates that the resels are defined within. In SPM’s native
coordinates, the two activations are roughly 9mm apart.)
Consistent with the hypothesis that the self and friend temporal discounting
task activate the same region, and this region differs from that activated by the WTR
task, I found that the activation in the WTR–self temporal discounting comparison at
(23, 37, 44) was in the “same” place as another activation in the WTR–friend
temporal discounting comparison. This latter activation was located in right
dorsomedial prefrontal cortex, BA 6, at (21, 31, 35). Both BA 6 and 8 have been
implicated in various types of social reasoning tasks (Barbey, Krueger, & Grafman,
2009; Ermer, Guerin, Cosmides, Tooby, & Miller, 2006; Fiddick, Spampinato, &
Grafman, 2005), including moral reasoning tasks that involve trading off welfare
(Greene, Sommerville, Nystrom, Darley, & Cohen, 2001).
163
Finally, there was an additional shared activation in left caudate, located at (1,
1, 12) for the WTR–self temporal discounting comparison and (-3, -1, 17) for the
WTR–friend comparison.
For transparency, I also present significant activations comparing the two
temporal discounting tasks to the WTR task (Table 8) and the two temporal
discounting tasks to each other (Table 9). The most direct comparison to previous
neuroscientific literature on intertemporal choice would be the self temporal
discounting–WTR comparison at the top of Table 8. Several of these activations, such
as in the caudate and cingulate cortex, are consistent with past work (Kable &
Glimcher, 2007; McClure, Laibson, Loewenstein, & Cohen, 2004; Weber & Huettel,
2008).
In sum, welfare tradeoff decisions dissociate from another intuitive economic
preference, temporal discounting. This holds true relative to two controls tasks,
discounting as the self prefers and discounting as a friend prefers. This result is not
predicted by an alternative account holding that welfare tradeoff computations are
made by a general economic preference calculation system. It is, however, consistent
with the hypothesis that the mind contains specialized neurocognitive mechanisms for
computing welfare allocations and tradeoffs.
164
WTR > Self Temporal Discounting
10
5
0
WTR > Friend Temporal Discounting
10
5
0
Figure 5. Neural activations comparing the welfare tradeoff task to the two temporal
discounting tasks. The figure for the self temporal discounting contrast is positioned
at (17, 34, 46) in Talairach coordinates; the activation in rDMPFC at Brodmann’s
area (BA) 8 can be seen. The figure for the friend temporal discounting contrast is
positioned at (21, 35, 34) in Talairach coordinates; the activations in rDMPFC at BAs
6 & 8 can be seen. Color map represents t-values.
165
Table 7
Significant Activations of the Welfare Tradeoff Task Relative to the Temporal
Discounting Tasks
Brain Area
WTR > Self Temporal
Discounting
Superior Frontal
Gyrus
Medial Frontal
Gyrus
Middle Frontal
Gyrus
Precentral Gyrus
Postcentral Gyrus
Caudate
WTR > Friend
Temporal Discounting
Superior Frontal
Gyrus
Medial Frontal
Gyrus
Hemi
BA
Voxels
x
y
z
t-value
R
L
L
8
8
10
106 23
22 -18
16 -10
37
37
53
44
47
-2
10.4
7.79
5.1
L
10
11 -16
47
10
5.68
R
R
L
R
L
L
6
6
9
5
-
17 49
6
11 45
2
16 -40 17
16
2 -40
25
1
1
12 -8 13
42
25
39
71
12
7
3.97
4.71
4.33
4.8
5.35
4.05
R
L
8
6
29
39
21
-2
31
27
49
59
5.07
5.38
R
L
L
6
9
10
198
15
10
21
-7
-5
31
53
61
35
38
21
7.51
4.24
4.03
Middle Temporal
Gyrus
R
22
110 53 -35
0
7.96
R
39
44 45 -60
21
5.79
L
21
11 -54 -4 -13
4.98
Cingulate Gyrus
R
24
18 13
3
30
11.06
L
32
111 -20 12
39
7.72
Caudate
R
19 23 -22
22
4.99
L
38 -3 -1
17
4.98
Claustrum
L
37 -27 11
12
6.72
Note. Hemi = Hemisphere. BA = Brodmann area based on stereotaxic coordinates.
Voxels = number of voxels in a cluster. (x,y,z) are in Talairach coordinates.
166
Table 8
Significant Activations of the Temporal Discounting Tasks Relative to the Welfare
Tradeoff Task
Brain Area
Self Temporal
Discounting > WTR
Caudate
Caudate Tail
Cerebellar Tonsil
Cingulate Gyrus
Culmen
Cuneus
Inferior Frontal Gyrus
Lentiform Nucleus,
Putamen
Lingual Gyrus
Lingual Gyrus
Middle Occipital
Gyrus
Parahippocampal
Gyrus
Parahippocampal
Gyrus, Hippocampus
Posterior Cingulate
Sub-Gyral,
Hippocampus
Superior Temporal
Gyrus
Friend Temporal
Discounting > WTR
Anterior Cingulate
Culmen
Cuneus
Declive of Vermis
Inferior Parietal
Hemi
BA Voxels
x
y
z
t-value
R
R
L
L
R
L
R
R
31
24
13
130 38 -29
83 23 -36
10 -25 -52
13 -16 -27
15
8
1
48 -1 -68
229
2 -92
14 34
9
-5
12
-41
47
37
-6
3
-14
7.56
6.46
5.15
6.71
4.74
4.25
6.61
4.84
R
R
R
L
19
-
18 25
4
77 29 -10
62 25 -62
52 -16 -76
18
-9
1
-1
5.41
4.81
5.98
5.29
L
19
10 -27 -89
8
4.54
L
R
30
30
203 -20 -47
10 16 -39
8
-3
5.09
4.23
R
R
29
13
21
34 -13
4 -45
-18
7
4.41
4.36
R
-
23
28 -41
3
9
R
R
R
42
38
38
19
13
10
62 -27
40
0
46 17
14
-17
-11
5.38
5.29
5.18
L
L
L
R
R
L
L
32
30
18
19
40
56 -1
11 27
83 -3
61
4
13
9
28 -3
13 -55
-6
-22
8
7
36
-10
24
7.69
4.35
5.94
4.78
4.51
7.13
4.17
167
46
-31
-74
-90
-86
-72
-43
Lobule
Inferior Temporal
Gyrus
R
20
14 51 -49 -10
4.13
Lentiform Nucleus,
Putamen
R
19 31 -10
-7
6.22
Middle Occipital
Gyrus
R
18
106 32 -82
1
7.26
Paracentral Lobule
R
5
21
8 -38
49
5.76
Parahippocampal
Gyrus
R
19
20 21 -54
0
6.11
Parahippocampal
Gyrus, Amygdala
R
20 21 -11 -13
4.93
Precentral Gyrus
L
4
26 -28 -29
53
4.92
Precuneus
R
7
13 15 -69
52
5.13
Superior Temporal
Gyrus
R
42
12 60 -29
12
5.15
Note. Hemi = Hemisphere. BA = Brodmann area based on stereotaxic coordinates.
Voxels = number of voxels in a cluster. (x,y,z) are in Talairach coordinates.
168
Table 9
Significant Activations of the Temporal Discounting Tasks Relative to Each Other
Brain Area
Friend > Self Temporal
Discounting
Middle Frontal Gyrus
Cingulate Gyrus
Insula
Self > Friend Temporal
Discounting
Caudate Tail
Caudate Body
Cingulate Gyrus
Culmen
Declive
Inferior Parietal
Lobule
Insula
Lentiform Nucleus,
Putamen
Lingual Gyrus
Middle Frontal Gyrus
Middle Temporal
Gyrus
Parahippocampal
Gyrus, Hippocampus
Postcentral Gyrus
Posterior Cingulate
Precentral Gyrus
Superior Temporal
Gyrus
Thalamus
Thalamus, VLN
Transverse Temporal
Gyrus
Hemi
BA Voxels x
y
R
R
L
6
31
13
12 43 -1
10 17 -22
13 -34 20
R
R
L
L
L
L
L
24
24
-
53
23
R
R
R
40
13
13
13
L
L
R
R
L
z
t-value
54
33
16
7.17
3.86
5.96
-37
-41
-3
-1
-6
-58
-79
16
5
20
38
29
2
-12
10.4
5.66
4
9.01
4.37
6.4
6.25
54 -42
34 -27
41 -14
44
10
11
4.8
5.37
5.11
18
19
19
9
28 -23 -7
-20 -76
15 25 -73
25 -63
17 -48 28
18
-7
-4
-1
30
5.12
4.86
4.96
3.66
5.11
L
37
18 -53 -66
10
8.68
R
L
R
R
R
L
3
30
30
30
4
10
33
-61
2
23
4
13
42 -59
-11
-18
-53
-67
-65
-10
-14
34
7
12
11
31
4.93
4.08
4.41
5.92
5.18
6.78
R
R
L
38
-
12
31
42
6
19 -28
-18 -12
-23
0
12
4.92
5.57
5.1
R
41
66
43 -22
12
5.72
21
36
-12
291 -11
-3
35 -5
89 -16
169
Note. Hemi = Hemisphere. BA = Brodmann area based on stereotaxic coordinates.
Voxels = number of voxels in a cluster. VLN = Ventral lateral nucleus. (x,y,z) are in
Talairach coordinates.
170
Method
Subjects
Eight graduate and undergraduate students associated with the UCSB
Psychology Department were recruited through email advertisements. Two subjects
were left-handed; three were female. Subjects were compensated $30 for
approximately 1.5 hours of their time. Their decisions were hypothetical; thus they
received no additional compensation for their choices.
Tasks
Subjects completed three types of tasks: a welfare tradeoff task, a temporal
discounting task according to their own preferences, and a temporal discounting task
according to what they believed their closest friend’s preferences would be. Subjects
completed three versions of the welfare tradeoff task: one with respect to subjects’
closest same-sex friend, one with respect to a same-sex acquaintance, and one with
respect to a same-sex “stranger,” defined as someone the subject saw in their day-today life (as opposed to through television or other media), but with whom the subject
had never interacted (e.g., a person in the subject’s class the subject had never talked
to). Subjects also completed three versions each of the temporal discounting tasks:
one with the larger benefit available at 6 months, one at 8 months, and one at 12
months; in all versions the smaller benefit was available tomorrow. Prior to entering
the scanner, subjects determined the targets of their welfare tradeoff decisions and
completed practice trials of all three task types. (The practice trials used a different
target than the main experimental trials and involved a delay of 10 months.)
171
Subjects were scanned as they completed the task in nine blocks. Each block
consisted of 42 trials of the same type. Thus, for example, all and only trials of
welfare tradeoffs involving an acquaintance were completed within a single block.
Similarly, all and only trials of temporal discounting according to personal
preferences with a delay of 6 months were completed within a single block. Between
blocks, subjects were given as much rest as they desired. The order of the blocks was
randomized across subjects
Individual trials proceeded as follows. For the welfare tradeoff task, subjects
first saw a fixation/instruction screen. This screen had (a) the word “Choose”
centered near the top, (b) near the middle at the left the word “You,” and (c) near the
middle at the right the initials of the target (e.g., the initials of the subject’s friend).
This screen appeared for 250, 500, 750, 1000, 1250, or 1500msec, with the exact
length randomly determined. Next, specific values appeared below “You” and the
initials; these represented the amounts that would be allocated to the subject or the
target of their choices based on subjects’ decisions. Subjects responded with a fingerpress response box. Subjects had as long as they desired to make their decisions. As
soon they made their decision, the fixation/instruction screen reappeared. (This means
that timing of subjects’ decisions was in no way linked to time points within the
scanner’s repetition time.) The temporal discounting tasks were similar. Instead of
“You” and the initials, however, “Tomorrow” and the longer delay (e.g. “8 months)
appeared. Moreover, instead of “Choose,” the word “You” appeared if the task was to
172
be completed according to the subject’s preferences and the friend’s initials appeared
if it was to completed according the friend’s preferences.
The same 42 pairs of numbers (e.g., $54 tomorrow or $98 in 6 months) were
used in all blocks. The specific pairs used were chosen to allow WTRs to range
between 0 and 1. The lengths of delay (6, 8, and 10 months) were chosen so that the
specific numbers used would also be roughly centered on the distribution of
discounting parameters normally found in the literature (Kirby & Marakovic, 1996).
The numbers are shown in Table 10. Note that subjects did not get paid based on their
decisions; decisions were hypothetical and subjects were aware of this.
173
Table 10
Numbers Used in the WTR and Temporal Discounting Tasks
You/Tomorrow
92
87
82
82
73
79
77
66
60
54
53
45
38
39
28
24
19
14
9
5
1
Other/Later
93
92
91
96
91
105
110
101
100
98
105
101
94
110
94
97
97
93
94
96
100
You/Tomorrow
41
47
46
41
34
45
29
27
30
26
24
19
23
19
18
12
10
8
5
3
1
Other/Later
41
49
51
48
43
60
42
41
50
48
47
43
57
54
59
48
50
50
52
51
59
Imaging Methods and Analysis
Imaging Details
Images were acquired with a 3T Tim Siemens Magnetom Trio scanner with a
12 channel phased array head coil. Functional runs were acquired with an echo planar
gradient-echo imaging sequence sensitive to BOLD contrast, with a 2000msec
repetition time (TR), 37 interleaved slices per TR (3mm thickness, .5mm gap), a
30msec echo time (TE), a flip angle of 90°, a field of view (FOV) of 192mm, and a
64 x 64 matrix. Prior to functional runs, a high-resolution T1-weighted mprage image
174
of the whole brain was acquired (TR = 2300msec, TE = 2.98msec, flip angle = 9°, 3D acquisition, FOV = 256mm, slice thickness = 1.1mm, 256 x 256 matrix).
Image Preprocessing
Images were preprocessed using SPM8 (www.fil.ion.ucl.ac.uk/spm). All
subjects were checked for motion artifacts; none were found. Functional images were
realigned and then co-registered with each subjects’ anatomical image. (No slice
timing correction was applied.) Subjects’ anatomical and functional images were
normalized to the smoothed Montreal Averaged 152 atlas (template “T1” in SPM8),
with anatomical images being resliced to 1 x 1 x 1mm voxels and functional images
to 2 x 2 x 2mm voxels. Functional images were then smoothed with an 8 x 8 x 8mm
FWHM Gaussian kernel filter.
Statistical Analysis
Images were analyzed using SPM8. I used a simple block design defined by
three conditions: (1) the three blocks of the WTR task, (2) the three blocks of the
personal temporal discounting task, and (3) the three blocks of the friend temporal
discounting task. SPM’s canonical hemodynamic response function was used as a
basis function (no derivatives were used). Classical estimation was used at both the
individual and group level. Given that this was a pilot study with a limited number of
subjects, I used a relatively liberal statistical threshold to define significant clusters at
the group-level: p < .005 (uncorrected) with a cluster extent of 10. (No activations
would be significant using either family-wise error correction or false discovery rate
techniques.) The stereotaxic locations of significant voxels were converted from
175
SPM’s native MNI space to Talairach space using the Lancaster transform (icbm2tal),
implemented in GingerALE (Laird et al., 2005). Anatomical regions and Brodmann’s
areas were determined based on nearest gray matter using the Talairach Client
program (Lancaster et al., 1997; Lancaster et al., 2000).
176
Appendix 8: Materials for Chapter 3
Cover Stories
Cover Story (Studies 6, 7, and 9)
In today's study you'll be learning about a group of people who were traveling
together on a small chartered plane. While crossing the Pacific Ocean, the plane hit a
violent storm and was forced very far off course. A bolt of lightning ripped through
the plane and damaged the electrical system as well as one of the engines. With no
radio and failing power, the pilot managed to crash land on a small island.
The pilot died during the impact and many of the passengers were seriously injured.
For two days the passengers waited by the plane for help. They eventually ate all the
food they had.
Realizing that they might be on the island for a while and that they needed supplies,
those who were not injured decided to go out and collect food. They knew that they
needed to work together and cooperate so that everyone, including those who were
seriously injured, would survive.
Everyone who was going out searching agreed that any and all food they found they
would bring back and share with the entire group. Each person carried a bag to collect
food as well as a spear made from wood found near the crash site. In order to cover
more ground, each person went out on their own.
Cover Story (Study 8)
In today's study you'll be learning about a group of people who were traveling
together on a small chartered plane. While crossing the Pacific Ocean, the plane hit a
violent storm and was forced very far off course. A bolt of lightning ripped through
the plane and damaged the electrical system as well as one of the engines. With no
radio and failing power, the pilot managed to crash land on a small island.
The pilot died during the impact and many of the passengers were seriously injured.
For two days the passengers waited by the plane for help. They eventually ate all the
food they had.
Realizing that they might be on the island for a while and that they needed supplies,
those who were not injured decided to go out and collect food. They knew that they
needed to work together and cooperate so that everyone, including those who were
seriously injured, would survive.
Everyone who was going out searching agreed that most of the food they found they
would bring back and share with the entire group, although they could also keep some
for themselves. Each person carried a bag to collect food as well as a spear made
from wood found near the crash site. In order to cover more ground, each person went
out on their own.
177
Cover Story (Study 10)
In today's study you'll be learning about a group of people who were traveling
together on a small chartered plane. While crossing the Pacific Ocean, the plane hit a
violent storm and was forced very far off course. A bolt of lightning ripped through
the plane and damaged the electrical system as well as one of the engines. With no
radio and failing power, the pilot managed to crash land on a small island.
The pilot died during the impact and many of the passengers were seriously injured.
For two days the passengers waited by the plane for help. They eventually ate all the
food they had.
Realizing that they might be on the island for a while and that they needed supplies,
those who were not injured decided to go out and collect food. Food on the island was
scarce and hard to come by. They knew that they needed to work together and
cooperate so that everyone, including those who were seriously injured, would
survive.
Everyone who was going out searching agreed that most of the food they found they
would bring back and share with the entire group, although they could also keep some
for themselves. Each person carried a bag to collect food as well as a spear made
from wood found near the crash site. In order to cover more ground, each person went
out on their own.
Sentences
Standard Background Sentences
Fruit
He watched a flock of birds fly overhead and then took the coconuts he’d found back
to camp.
With some passion fruit in hand, he walked over a few fallen logs and went back to
camp.
He whistled to himself while walking back to camp with some papaya fruit.
Monkeys chattered off in the distance as he walked back to camp carrying plums.
He slung his bag, now full of apricots, over his shoulder and made the trip back to
camp.
While bringing some guava fruit back to camp, he felt the warm sun on the back of
his neck.
Carrying cantaloupes to camp, he thought about warming his hands in the fire that
night.
Before heading to camp with some honeydew melons, he drank from a spring.
Meat
He stopped to tie his shoe and then continued back to the camp with the oysters he’d
collected.
The sun was setting as he made his way back to camp with the wild pig he’d caught.
178
While returning to camp with a deer he had caught, he looked at some interesting
rocks.
He watched the white clouds overhead while taking the goose he had caught back to
camp.
He wiped the sweat from his brow and walked back to camp with the rabbit he had
trapped.
Brushing sand from his legs, he started the journey back to camp with the mussels he
had found.
He listened to his own footsteps while taking the pheasant he’d caught to camp.
He could smell the salt in the air while taking the quail he’d caught back to camp.
Vegetable
He brushed some briars off of his clothes and took the lettuce he found to camp.
He rinsed his hands and set off back towards camp with the yams he had dug up.
He watched the ocean waves while walking back to camp with the zucchini he had
found.
He heard the buzzing of insects while taking the carrots he had gotten back to camp.
Taking a path back to camp with the avocados he had picked, he heard a bird call
high above.
Going back to camp with the squash he collected, he noticed the island’s humidity.
He saw some flowers starting to bloom as he took the spinach he’d collected back to
camp.
While walking back to camp with tomatoes he’d found, he felt the wind blowing
through the island.
High Cost (Studies 6 & 7)
Fruit
After the group ate dinner, he worked all night to gather peaches for everyone.
He risked going to a dangerous part of the island to collect the kiwi fruit that grew
there.
After someone else dropped oranges into a cave, he ventured into the dark alone to
get them back.
After cougars had been seen by the pineapple tree, he was the only one willing to get
fruit from them today.
Meat
He caught a sea bass using pieces of his watch as lures knowing that it would destroy
the watch.
He ruined his shirt by using it as a harness to climb a tree and collect the cashews.
In order to catch some crabs, he waded into the shark-infested shallows off the bay.
He risked drowning by holding his breath long enough to swim to the depth where
shrimp lived.
179
Low Cost (Study 6)
Fruit
While he was taking bananas back to camp, the wind blew his hat away and he could
not find it.
A few dollar bills fell out of a hole in his pocket while he carried pears he’d found
back to camp.
While picking strawberries, he cracked and broke his watch against a rock.
After setting down what he was carrying in order to drink, he took the mangos but
forgot his comb and couldn’t find it later.
Meat
While bringing a lobster back to camp, he dropped his Walkman and it broke into
pieces.
While taking a duck back, a bat landed on his head knocking his sunglasses over the
cliff.
Carrying some snapper back, he tripped on a buried root and crushed his camera.
While he was taking honey back to camp, his keys fell from his coat into a patch of
ivy.
Big Cost, Benefit to Group (Study 8)
Fruit
He risked going to a dangerous part of the island to collect the kiwi fruit that grew
there and then took them to the group.
After someone else dropped oranges into a cave, he ventured into the dark alone to
get them and bring them to camp.
Even after cougars had been seen by the pineapple trees, he went and got fruit from
them to take back to the group.
He climbed to the top of a perilous cliff in order to gather peaches from the grove
there and then brought them to camp.
Meat
He exposed himself to the hazardous waves on the rocks of the bay so he could catch
sea bass to bring to the group.
In order to catch some crabs to bring back to the group, he waded into the sharkinfested shallows off the bay.
To get food for the group, he risked drowning by holding his breath long enough to
swim to the depth where shrimp lived.
During a windstorm, he got food for the group by climbing the highest tree on the
island to collect the cashews at the top.
180
Big Cost, Benefit to the Self (Study 8)
Fruit
When the group finished eating food he provided, he collected strawberries for
himself in an area full of dangerous bears.
After he brought food to camp, he went through a rapidly flooding valley during a
rainstorm to get himself some mangos.
He brought food to the camp and then collected some bananas for himself by
swinging from tree-top to tree-top in the forest.
When he was finished collecting food for the group, he scaled a tall, vertical
mountain to get pears for himself.
Meat
After bringing food to the group, he collected snapper for himself to eat at the very
edge of a quickly moving waterfall.
After providing some food for the group, he waded through a river full of deadly
piranhas to hunt duck for himself.
After getting food to bring to camp, he swam far, far out into treacherous open waters
to collect lobster for himself.
Later, after bringing food to camp, he got honey for himself by carefully moving
through a dense cloud of angry bees.
Big Benefit (Studies 9 & 10)
Fruit
He risked going to a dangerous part of the island to collect the ample, ripe kiwi fruit
he knew grew there and then took them to the group.
After someone else dropped a great many juicy oranges into a cave, he ventured into
the dark alone to get them and bring them to camp.
Even after cougars had been seen by the pineapple trees, he went and got for the
group the copious armfuls of fruit he had seen there earlier.
He noticed a large patch of peaches in the grove at the top of a perilous cliff and
climbed up to gather loads to bring back for the group.
Late in the day he brought back to camp the bags and bags worth of strawberries he
collected in an area full of dangerous bears.
He went through a rapidly flooding valley during a rainstorm and gathered for the
group the huge tasty mangos he had seen there.
Moving from precarious tree-top to precarious tree-top, he collected for the group
many bunches of yellow bananas he had seen from the ground.
He scaled the sheer face of a tall vertical mountain and gathered for the group the tons
of pears he saw growing there.
181
Meat
Seeing many huge sea bass, he exposed himself to the hazardous waves on the rocks
of the bay so he could catch them to bring to the group.
To bring back to the group the large number of crabs he could see in the bay, he
waded into the bay's shark-infested shallows.
To get food for the group, he risked drowning by holding his breath long enough to
swim to the depth where he saw that an enormous group of shrimp lived.
He got food for the group by climbing one the highest trees on the island to collect
the pounds and pounds of cashews he saw at the top.
At the very edge of a quickly moving waterfall, he speared the twenty or thirty
snapper he saw there and took them to camp.
Moving carefully as he waded through a river full of deadly piranhas, he hunted a
flock of ducks and took most to the group.
Seeing numerous large lobsters there, he swam far out into treacherous open waters
and collected them to bring to camp.
He managed to get many cups worth of honey for the group by carefully moving
through a dense cloud of angry bees.
Small Benefit (Studies 9 & 10)
Fruit
He risked going to a dangerous part of the island to collect the few kiwi fruit he knew
grew there and then took them to the group.
After someone else dropped a pair of oranges into a cave, he ventured into the dark
alone to get them and bring them to camp.
Even after cougars had been seen by the pineapple trees, he went and got for the
group the handful of fruit he had seen there earlier.
He noticed a few peaches in the grove at the top of a perilous cliff and climbed up to
gather them to bring back for the group.
Late in the day he brought back to camp the handful of strawberries he collected in an
area full of dangerous bears.
He went through a rapidly flooding valley during a rainstorm and gathered for the
group the little mangos he had seen there.
Moving from precarious tree-top to precarious tree-top, he collected for the group a
pair of finger-sized bananas he had seen from the ground.
He scaled the sheer face of a tall vertical mountain and gathered for the group the two
or three pears he saw growing there.
Meat
Seeing a small sea bass, he exposed himself to the hazardous waves on the rocks of
the bay so he could catch it to bring to the group.
To bring back to the group the few crabs he could see in the bay, he waded into the
bay's shark-infested shallows.
182
To get food for the group, he risked drowning by holding his breath long enough to
swim to the depth where he saw that a small group of shrimp lived.
He got food for the group by climbing one the highest trees on the island to collect
the few cashews he saw at the top.
At the very edge of a quickly moving waterfall, he speared the single snapper he saw
there and took it to camp.
Moving carefully as he waded through a river full of deadly piranhas, he hunted a
lone duck and took it to the group.
Seeing a solitary lobster there, he swam far out into treacherous open waters and
collected it to bring to camp.
He managed to get a few thimbles worth of honey for the group by carefully moving
through a dense cloud of angry bees.
[Unlike other studies, subjects did not see all sentences in Studies 9 & 10. Whether a
given resource was depicted as being in abundance or scarce was randomly
determined for each subject. Thus, there are twice as many diagnostic sentences
created for these studies.]
Could not find food (Study 10)
He searched all over the island but found no food to bring back.
Everyone was hungry so he tried very hard to catch some food but did not manage to.
After he heard an animal in the forest, he tracked it for several hours but did not find
it again.
He spent all day trying to get to fruit trees he saw off in the distance, but they turned
out to just have flowers.
He worked hard to get a beehive out of a tree only to find it abandoned and without
honey.
He quietly tracked an antelope all day, but someone walking loudly through the forest
scared it off.
He spent a long time making a fishing net, but an unusually strong current tore it
before he caught anything.
While climbing a tall palm to get coconuts, he slipped and twisted his ankle and
couldn't search for food the rest of the day.
After searching the entire day all through the deep ravine, he still came back empty
handed.
He scoured the underbrush of a large glen, hoping to find food, but found nothing
instead.
A flock of birds made noise in a pond ahead, but flew away before he could catch one
to bring back.
After a long walk, he came to a fruit grove, but the trees were long dead and had no
fruit.
Although he thought that taking a less traveled route might lead to some food, he
found nothing.
183
He combed through the tall grass for anything edible but found nothing to take back
to camp.
He fished in the river all day, walking up and down it, but the fish were just not
biting.
After he heard an animal in the forest, he tracked it for several hours but did not find
it again.
He spent the day hiking to a place where someone else had seen fruit earlier, but it
was gone.
Out of all the shells on the beach, he could find none that still had a crab in them.
Despite swimming all over the lagoon, he could not find any fish for the group.
No matter where he went, no matter where he looked, he could not find any food for
the group.
He followed a winding stream, hoping to catch animals drinking from it, but none
were there today.
He checked the trunks of trees looking for edible mushrooms; unfortunately, animals
had already eaten them.
Someone thought edible tubers might be in the ground nearby, but no matter where he
dug, he could not find any.
He eventually went back to camp empty-handed after many hours climbing trees
looking for bananas.
184
Questions
Study 1
Does this person deserve to be punished for his actions on the island?
1
2
3
4
5
6
7
Not At All
Very Much
Does this person deserve to be rewarded for his actions on the island?
1
2
3
4
5
6
7
Not At All
Very Much
Would you be willing to have this person on your "team"?
1
2
3
4
5
6
7
Not At All
Very Much
To what extent is this person trustworthy?
1
2
3
4
5
6
7
Not At All Trustworthy
Very Trustworthy
To what extent is this person competent?
1
2
3
4
5
6
7
Not At All Competent
To what extent is this person likeable?
1
2
3
4
5
6
7
Not At All Likeable
Very Competent
Very Likeable
To what extent is this person altruistic?
1
2
3
4
5
6
7
Not At All Altruistic
Very Altruistic
To what extent is this person selfish?
1
2
3
4
5
6
7
Not At All Selfish
Very Selfish
How much do the other members of the group respect him?
1
2
3
4
5
6
7
Not At All
Very Much
How much do the other members of the group want him to be a leader?
1
2
3
4
5
6
7
Not At All
Very Much
185
How much does he care about the group?
1
2
3
4
5
6
7
Not At All
Very Much
How much effort did this person put into getting food for the group?
1
2
3
4
5
6
7
Very Little
Very Much
Sometimes when people went out searching for food, they ended up paying a cost
such as putting themselves in danger or losing personal items. When this happened to
the person above, to what extent did he end up paying the cost on purpose?
1
2
3
4
5
6
7
Not At All
Very Much
Studies 2-5
How much does the person on the left care about the group as a whole? [pictures were
shown]
1
2
3
4
5
6
7
Not at All
Very Much
How altruistic does the person on the left feel towards the group as a whole? [pictures
were shown]
1
2
3
4
5
6
7
Not At All Altruistic
Very Altruistic
Would you be willing to work with this person in a group?
1
2
3
4
5
6
7
Not At All
Very Much
Would you be willing to work with this person one on one?
1
2
3
4
5
6
7
Not At All
Very Much
To what extent is this person trustworthy?
1
2
3
4
5
6
7
Not At All Trustworthy
To what extent is this person competent?
1
2
3
4
5
6
7
Not At All Competent
Very Trustworthy
Very Competent
186
To what extent is this person likeable?
1
2
3
4
5
6
7
Not At All Likeable
Very Likeable
To what extent is this person willing to provide for the group?
1
2
3
4
5
6
7
Not At All Willing
Very Willing
How much effort did this person put into getting food for the group?
1
2
3
4
5
6
7
Very Little
Very Much
How much do the other members of the group respect him?
1
2
3
4
5
6
7
Not At All
Very Much
187
References
Aron, A., Aron, E. N., & Smollan, D. (1992). Inclusion of Other in the Self Scale and
the Structure of Interpersonal Closeness. Journal of Personality and Social
Psychology, 63(4), 596-612.
Axelrod, R., & Hamilton, W. D. (1981). The Evolution of Cooperation. [Article].
Science, 211(4489), 1390-1396.
Barbey, A. K., Krueger, F., & Grafman, J. (2009). An evolutionarily adaptive neural
architecture for social reasoning. Trends in Neurosciences, 32(12), 603-610.
Bechara, A., Damasio, H., & Damasio, A. R. (2000). Emotion, decision making and
the orbitofrontal cortex. Cereb Cortex, 10(3), 295-307.
Birnbaum, M. H. (2008a). Evaluation of the priority heuristic as a descriptive model
of risky decision making: Comment on Brandstatter, Gigerenzer, and Hertwig
(2006). Psychological Review, 115(1), 253-260.
Birnbaum, M. H. (2008b). New paradoxes of risky decision making. Psychological
Review, 115(2), 463-501.
Boehm, C. (1993). Egalitarian Behavior and Reverse Dominance Hierarchy. Current
Anthropology, 34(3), 227-254.
Boyd, R., & Richerson, P. J. (1989). The Evolution of Indirect Reciprocity. Social
Networks, 11(3), 213-236.
Brandstatter, E., Gigerenzer, G., & Hertwig, R. (2006). The priority heuristic: Making
choices without trade-offs. Psychological Review, 113(2), 409-432.
188
Budesheim, T., & DePoala, S. (1994). Beauty or the beast? The effects of appearance,
personality, and issue information on evaluations of political candidates.
Personality and Social Psychology Bulletin, 20, 339-348.
Buss, D. M., Shackelford, T. K., Kirkpatrick, L. A., & Larsen, R. J. (2001). A half
century of mate preferences: The cultural evolution of values. Journal of
Marriage and the Family, 63(2), 491-503.
Camerer, C. F. (2003). Behavioral game theory: Experiments in strategic interaction:
Princeton University Press.
Camerer, C. F., & Fehr, E. (2006). When does "economic man" dominate social
behavior? Science, 311(5757), 47-52.
Clutton-Brock, T. (2002). Behavioral ecology - Breeding together: Kin selection and
mutualism in cooperative vertebrates. Science, 296(5565), 69-72.
Clutton-Brock, T. (2009). Cooperation between non-kin in animal societies. Nature,
462(7269), 51-57.
Connor, R. C. (1986). Pseudo-reciprocity: Investing in mutualism. Animal Behaviour,
34(5), 1562-1566.
Cosmides, L. (1989). The Logic of Social-Exchange - Has Natural-Selection Shaped
How Humans Reason - Studies with the Wason Selection Task. Cognition,
31(3), 187-276.
Cosmides, L., & Tooby, J. (2005). Neurocognitive adaptations designed for social
exchange. In D. M. Buss (Ed.), The Handbook of Evolutionary Psychology
(pp. 584-627). Hoboken, NJ: Wiley.
189
Cosmides, L., & Tooby, J. (2008). Can evolutionary psychology assist logicians? A
reply to Mallon. In W. Sinnott-Armstrong (Ed.), Moral psychology (pp. 131136). Cambridge, MA: MIT Press.
Dehaene, S., Izard, V., Spelke, E., & Pica, P. (2008). Log or linear? Distinct intuitions
of the number scale in western and amazonian indigene cultures. Science,
320(5880), 1217-1220.
Delton, A. W., Cosmides, L., Guemo, M., Tooby, J., & Robertson, T. E. (2010). The
psychosemantics of free riding: Dissecting the architecture of a moral
concept. Unpublished manuscript.
Downs, C., & Lyons, P. (1991). Natural observations of the links between
attractiveness and initial legal judgments. Personality and Social Psychology
Bulletin, 17, 541-547.
Dunbar, R. I. M. (2004). Gossip in evolutionary perspective. Review of General
Psychology, 8(2), 100-110.
Ellsberg, D. (1961). Risk, ambiguity, and the Savage axioms. Quarterly Journal of
Economics, 75(4), 643-669.
Ermer, E., Guerin, S., Cosmides, L., Tooby, J., & Miller, M. (2006). Theory of mind
broad and narrow: Reasoning about social exchange engages ToM areas,
precautionary reasoning does not. Social Neuroscience, 1(3-4), 196-219.
Fehr, E. (2009). Social preferences and the brain. In P. W. Glimcher, C. F. Camerer,
E. Fehr & R. A. Poldrack (Eds.), Neuroeconomics: Decision making and the
brain (pp. 213-232). Amsterdam: Academic Press.
190
Fehr, E., & Henrich, J. (2003). Is strong reciprocity a maladaptation? On the
evolutionary foundations of human altruism. In P. Hammerstein (Ed.),
Genetic and cultural evolution of cooperation (pp. 55-82). Cambridge, MA,
US: MIT Press.
Fehr, E., & Schmidt, K. M. (1999). A theory of fairness, competition, and
cooperation. Quarterly Journal of Economics, 114, 817-868.
Fessler, D. M. T., & Navarrete, C. D. (2004). Third-party attitudes toward sibling
incest: Evidence for Westermarck's hypotheses. Evolution and Human
Behavior, 25, 277-294.
Fiddick, L., Spampinato, M. V., & Grafman, J. (2005). Social contracts and
precautions activate different neurological systems: An fMRI investigation of
deontic reasoning. Neuroimage, 28(4), 778-786.
Fiske, A. P. (1992). The four elementary forms of sociality: framework for a unified
theory of social relations. Psychol Rev, 99(4), 689-723.
Gallistel, C. R. (1990). The organization of learning. Cambridge, MA: MIT Press.
Gardner, A., West, S. A., & Barton, N. H. (2007). The relation between multilocus
population genetics and social evolution theory. American Naturalist, 169(2),
207-226.
Gigerenzer, G., Hell, W., & Blank, H. (1988). Presentation and Content - the Use of
Base Rates as a Continuous Variable. Journal of Experimental PsychologyHuman Perception and Performance, 14(3), 513-525.
191
Grafen, A. (1990). Biological Signals as Handicaps. Journal of Theoretical Biology,
144(4), 517-546.
Green, L., & Myerson, J. (2004). A discounting framework for choice with delayed
and probabilistic rewards. Psychol Bull, 130(5), 769-792.
Greene, J. D., Sommerville, R. B., Nystrom, L. E., Darley, J. M., & Cohen, J. D.
(2001). An fMRI investigation of emotional engagement in moral judgment.
Science, 293(5537), 2105-2108.
Gros-Louis, J., Perry, S., & Manson, J. H. (2003). Violent coalitionary attacks and
intraspecific killing in wild white-faced capuchin monkeys ( Cebus
capucinus). Primates, 44(4), 341-346.
Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist approach
to moral judgment. Psychological Review, 108(4), 814-834.
Haig, D. (1998). Genomic imprinting. American Journal of Human Biology, 10(5),
679-680.
Hamilton, W. D. (1964). The genetical evolution of social behaviour. Journal of
Theoretical Biology, 7(1), 1-52.
Hammerstein, P., & Parker, G. A. (1982). The asymmetric war of attrition. Journal of
Theoretical Biology, 96(4), 647-682.
Hampton, J. A. (2007). Typicality, graded membership, and vagueness. Cognitive
Science, 31(3), 355-384.
Hashemi, R. H., Bradley, W. G., Jr., & Lisanti, C. J. (2004). MRI: The basics (2nd
ed.). Philadelphia: Lippincott, Williams, & Wilkins.
192
Henrich, J. (2004). Cultural group selection, coevolutionary processes and large-scale
cooperation. [Review]. Journal of Economic Behavior & Organization, 53(1),
3-35.
Henrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., Gintis, H., et al. (2005).
"Economic man" in cross-cultural perspective: Behavioral experiments in 15
small-scale societies. [Review]. Behavioral and Brain Sciences, 28(6), 795-+.
Herrnstein, R. J., Rachlin, H., & Laibson, D. I. (Eds.). (1997). The matching law:
Papers in psychology and economics. New York: Russell Sage Foundation.
Hoffman, E., McCabe, K., & Smith, V. L. (1996). Social distance and other-regarding
behavior in dictator games. American Economic Review, 86(3), 653-660.
Jackendoff, R. (1983). Semanctics and cognition. Cambridge, MA: MIT Press.
Jackendoff, R. (2002). Foundations of language: Brain, meaning, grammar,
evolution. Oxford: Oxford University Press.
Jackendoff, R. (2006). Language, consciousness, culture: Essays on mental structure.
Cambridge, MA: MIT Press.
Johnson, E. J., Schulte-Mecklenbeck, M., & Willemsen, M. C. (2008). Process
models deserve process data: Comment on Brandstatter, Gigerenzer, and
Hertwig (2006). Psychological Review, 115(1), 263-272.
Jones, B. A., & Rachlin, H. (2006). Social discounting. Psychological Science, 17(4),
283-286.
193
Jones, B. A., & Rachlin, H. (2009). Delay, Probability, and Social Discounting in a
Public Goods Game. Journal of the Experimental Analysis of Behavior, 91(1),
61-73.
Kable, J. W., & Glimcher, P. W. (2007). The neural correlates of subjective value
during intertemporal choice. Nature Neuroscience, 10(12), 1625-1633.
Kameda, T., & Tamura, R. (2007). "To eat or not to be eaten?" Collective riskmonitoring in groups. Journal of Experimental Social Psychology, 43(2), 168179.
Kaplan, H., & Hill, K. (1985). Food sharing among Ache foragers: Tests of
explanatory hypotheses. Current Anthropology, 26(2), 223-246.
Kaplan, H., Hill, K., Lancaster, J., & Hurtado, A. M. (2000). A theory of human life
history evolution: Diet, intelligence, and longevity. Evolutionary
Anthropology, 9(4), 156-185.
Kelley, H. H., Holmes, J. G., Kerr, N. L., Reis, H. T., Rusbult, C. E., & Van Lange, P.
A. M. (2003). An atlas of interpersonal situations. Cambridge University
Press, New York.
Killeen, P. R. (2009). An additive-utility model of delay discounting. Psychol Rev,
116(3), 602-619.
Kirby, K. N., & Marakovic, N. N. (1996). Delay-discounting probabilistic rewards:
Rates decrease as amounts increase. Psychonomic Bulletin & Review, 3(1),
100-104.
194
Kurland, J. A., & Gaulin, S. J. C. (2005). Cooperation and conflict among kin. In D.
M. Buss (Ed.), Handbook of evolutionary psychology (pp. 447-482).
Hoboken, NJ: Wiley & Sons.
Laird, A. R., Fox, P. M., Price, C. J., Glahn, D. C., Uecker, A. M., Lancaster, J. L., et
al. (2005). ALE meta-analysis: controlling the false discovery rate and
performing statistical contrasts. Hum Brain Mapp, 25(1), 155-164.
Lancaster, J. L., Rainey, L. H., Summerlin, J. L., Freitas, C. S., Fox, P. T., Evans, A.
C., et al. (1997). Automated labeling of the human brain: a preliminary report
on the development and evaluation of a forward-transform method. Hum
Brain Mapp, 5(4), 238-242.
Lancaster, J. L., Woldorff, M. G., Parsons, L. M., Liotti, M., Freitas, C. S., Rainey,
L., et al. (2000). Automated Talairach atlas labels for functional brain
mapping. Hum Brain Mapp, 10(3), 120-131.
Leimar, O., & Hammerstein, P. (2001). Evolution of cooperation through indirect
reciprocity. Proceedings of the Royal Society B-Biological Sciences,
268(1468), 745-753.
Leslie, A. M. (1994). ToMM, ToBy, and agency: Core architecture and domain
specificity. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind:
Domain specificity in cognition and culture (pp. 119-148). Cambridge, U.K.:
Cambridge University Press.
195
Leslie, A. M., Knobe, J., & Cohen, A. (2006). Acting intentionally and the side-effect
effect - Theory of mind and moral judgment. Psychological Science, 17(5),
421-427.
Lieberman, D., Tooby, J., & Cosmides, L. (2003). Does morality have a biological
basis? An empirical test of the factors governing moral sentiments relating to
incest. Proceedings of the Royal Society of London Series B-Biological
Sciences, 270(1517), 819-826.
Lieberman, D., Tooby, J., & Cosmides, L. (2007). The architecture of human kin
detection. Nature, 44, 727-731.
Lim, J., Sznycer, D., Delton, A. W., Robertson, T. E., Tooby, J., & Cosmides, L.
(2010). Interactive welfare tradeoff games. Unpublished manuscript.
Marr. (1982). Vision. San Francisco: W. H. Freeman & Company.
Maynard Smith, J. (1982). Evolution and the theory of games. Cambridge, U.K.:
Cambridge University Press.
Maynard Smith, J., & Price, G. R. (1973). The logic of animal conflict. Nature,
246(5427), 15-18.
McClure, S. M., Laibson, D. I., Loewenstein, G., & Cohen, J. D. (2004). Separate
neural systems value immediate and delayed monetary rewards. Science,
306(5695), 503-507.
McElreath, R., & Boyd, R. (2007). Mathematical models of social evolution: A guide
for the perplexed. Chicago: University of Chicago Press.
196
Milinski, M., Semmann, D., Bakker, T. C. M., & Krambeck, H.-J. (2001).
Cooperation through indirect reciprocity: Image scoring or standing strategy.
Proceedings of the Royal Society B-Biological Sciences, 268(1484), 24952501.
Mills, J., Clark, M. S., Ford, T. E., & Johnson, M. (2004). Measurement of communal
strength. Personal Relationships, 11(2), 213-230.
New, J., Krasnow, M. M., Truxaw, D., & Gaulin, S. J. C. (2007). Spatial adaptations
for plant foraging: women excel and calories count. Proceedings of the Royal
Society B: Biological Sciences, 274(1626), 2679-2684.
Noe, R., & Hammerstein, P. (1994). Biological Markets - Supply-and-Demand
Determine the Effect of Partner Choice in Cooperation, Mutualism and
Mating. Behavioral Ecology and Sociobiology, 35(1), 1-11.
Noe, R., & Hammerstein, P. (1995). Biological Markets. Trends in Ecology &
Evolution, 10(8), 336-339.
Nowak, M. A. (2006a). Evolutionary dynamics: Exploring the equations of life.
Boston: Belknap Press.
Nowak, M. A. (2006b). Five rules for the evolution of cooperation. [Review].
Science, 314(5805), 1560-1563.
Nowak, M. A., & Sigmund, K. (2005). Evolution of indirect reciprocity. [Review].
Nature, 437(7063), 1291-1298.
197
Ohtsubo, Y., & Watanabe, Y. (2003). Contrast effects and approval voting: An
illustration of a systematic violation of the independence of irrelevant
alternatives condition. Political Psychology, 24(3), 549-559.
Ohtsuki, H., & Iwasa, Y. (2006). The leading eight: Social norms that can maintain
cooperation by indirect reciprocity. Journal of Theoretical Biology, 239(4),
435-444.
Oliver, A. (2003). A quantitative and qualitative test of the Allais paradox using
health outcomes. Journal of Economic Psychology, 24, 35-48.
Pacheco, J. M., Santos, F. C., Souza, M. O., & Skyrms, B. (2009). Evolutionary
dynamics of collective action in N-person stag hunt dilemmas. Proceedings of
the Royal Society B-Biological Sciences, 276, 315-321.
Panchanathan, K., & Boyd, R. (2003). A tale of two defectors: the importance of
standing for evolution of indirect reciprocity. J Theor Biol, 224(1), 115-126.
Pinker, S. (2007). The stuff of thought: Language as a window into human nature.
New York: Viking.
Preuschoff, K., Bossaerts, P., & Quartz, S. R. (2006). Neural differentiation of
expected reward and risk in human subcortical structures. Neuron, 51(3), 381390.
Price, G. R. (1970). Selection and covariance. Nature, 227(5257), 520-521.
Price, G. R. (1972). Extension of covariance selection mathematics. Annals of human
genetics, 35(4), 485-490.
198
Rabin, M. (1993). Incorporating fairness into game theory and economics. American
Economic Review, 83, 1281-1302.
Rachlin, H., & Jones, B. A. (2008a). Altruism among relatives and non-relatives.
Behavioural Processes, 79(2), 120-123.
Rachlin, H., & Jones, B. A. (2008b). Social discounting and delay discounting.
Journal of Behavioral Decision Making, 21(1), 29-43.
Reis, H. T. (2008). Reinvigorating the concept of situation in social psychology.
Personality and Social Psychology Review, 12(4), 311-329.
Rieger, M. O., & Wang, M. (2008). What is behind the priority heuristic? A
mathematical analysis and comment on Brandstdtter, Gigerenzer, and Hertwig
(2006). Psychological Review, 115(1), 274-280.
Rode, C., Cosmides, L., Hell, W., & Tooby, J. (1999). When and why do people
avoid unknown probabilities in decisions under uncertainty? Testing some
predictions from optimal foraging theory. Cognition, 72(3), 269-304.
Rosenthal, R., Rosnow, R. L., & Rubin, D. B. (2000). Contrasts and effect sizes in
behavioral research: A correlational approach. Cambridge, U.K.: Cambridge
University Press.
Rusbult, C. E., & Van Lange, P. A. M. (2008). Why we need interdependence theory.
Social and Personality Psychology Compass, 2/5, 2049-2070.
Sell, A. (2005). Regulating welfare-tradeoff ratios: Three tests of an evolutionarycomputational model of human anger. Unpublished doctoral dissertation.
199
Sell, A., Cosmides, L., Tooby, J., Sznycer, D., von Rueden, C., & Gurven, M. (2009).
Human adaptations for the visual assessment of strength and fighting ability
from the body and face. Proceedings of the Royal Society B-Biological
Sciences, 276(1656), 575-584.
Sell, A., Tooby, J., & Cosmides, L. (2009). Formidability and the logic of human
anger. Proceedings of the National Academy of Sciences of the United States
of America, 106(35), 15073-15078.
Shepard, R. N. (1984). Ecological Constraints on Internal Representation - Resonant
Kinematics of Perceiving, Imagining, Thinking, and Dreaming. Psychological
Review, 91(4), 417-447.
Silverman, I., & Eals, M. (1992). Sex differences in spatial abilities: evolutionary
theory and data. In J. Barkow, L. Cosmides & J. Tooby (Eds.), The Adapted
Mind: evolutionary psychology and the generation of culture (pp. 533-549).
New York, NY: Oxford University Press.
Spelke, E. S. (2000). Core knowledge. American Psychologist, 55(11), 1233-1243.
Stephens, D. W., Brown, J. S., & Ydenberg, R. C. (Eds.). (2007). Foraging. Chicago:
University of Chicago Press.
Stewart-Williams, S. (2007). Altruism among kin vs. nonkin: effects of cost of help
and reciprocal exchange. Evolution and Human Behavior, 28(3), 193-198.
Stewart, J. (1985). Appearance and punishment: The attraction-leniency effect in the
courtroom. Journal of Social Psychology, 125(373-378).
200
Sugiyama, L. S. (2005). Physical attractiveness in adaptationist perspective. In D. M.
Buss (Ed.), Handbook of evolutionary psychology (pp. 292-343). Hoboken,
NJ: Wiley & Sons.
Sznycer, D., Lim, J., Delton, A. W., Robertson, T. E., Tooby, J., & Cosmides, L.
(2010). The mind accurately estimates others' welfare tradeoff ratios.
Unpublished manuscript.
Taylor, S. E., Fiske, S. T., Etcoff, N. L., & Ruderman, A. J. (1978). Categorical and
Contextual Bases of Person Memory and Stereotyping. Journal of Personality
and Social Psychology, 36(7), 778-793.
Thibaut, J. W., & Kelley, H. H. (1959). The social psychology of groups. New York:
John Wiley.
Tooby, J., & Cosmides, L. (1989). Evolutionary Psychologists Need to Distinguish
between the Evolutionary Process, Ancestral Selection Pressures, and
Psychological Mechanisms. Behavioral and Brain Sciences, 12(4), 724-724.
Tooby, J., & Cosmides, L. (1990). The Past Explains the Present - Emotional
Adaptations and the Structure of Ancestral Environments. Ethology and
Sociobiology, 11(4-5), 375-424.
Tooby, J., & Cosmides, L. (1996). Friendship and the banker’s paradox: Other
pathways to the evolution of adaptations for altruism. In W. G. Runciman, J.
Maynard Smith & R. I. M. Dunbar (Eds.), Evolution of Social Behaviour
Patterns in Primates and Man (Vol. 88, pp. 119-143).
201
Tooby, J., Cosmides, L., & Barrett, H. C. (2005). Resolving the debate on innate
ideas: learnability constraints and the evolved interpenetration of motivational
and conceptual functions. In P. Carruthers, S. Laurence & S. Stitch (Eds.), The
innate mind: structure and content (pp. 305-337). New York, NY: Oxford
University Press.
Tooby, J., Cosmides, L., & Price, M. E. (2006). Cognitive adaptations for n-person
exchange: the evolutionary roots of organizational behavior. Managerial and
Decision Economics, 27(2-3), 103-129.
Tooby, J., Cosmides, L., Sell, A., Lieberman, D., & Sznycer, D. (2008). Internal
regulatory variables and the design of human motivation: A computational
and evolutionary approach. In A. J. Elliot (Ed.), Handbook of approach and
avoidance motivation (pp. 251-271). Mahwah, NJ: Lawrence Erlbaum
Associates.
Trivers, R. L. (1971). Evolution of Reciprocal Altruism. [Article]. Quarterly Review
of Biology, 46(1), 35-&.
Tversky, A. (1969). Intransitivity of preferences. Psychological Review, 76(1), 31-48.
Tversky, A., & Kahneman, D. (1992). Advances in Prospect-Theory - Cumulative
Representation of Uncertainty. Journal of Risk and Uncertainty, 5(4), 297323.
Van Lange, P. A. M. (1999). The pursuit of joint outcomes and equality in outcomes:
An integrative model of social value orientation. Journal of Personality and
Social Psychology, 77(2), 337-349.
202
Van Lange, P. A. M., Otten, W., DeBruin, E. M. N., & Joireman, J. A. (1997).
Development of prosocial, individualistic, and competitive orientations:
Theory and preliminary evidence. Journal of Personality and Social
Psychology, 73(4), 733-746.
Weber, B. J., & Huettel, S. A. (2008). The neural substrates of probabilistic and
intertemporal decision making. Brain Res, 1234, 104-115.
Westermarck, E. (1891). The history of human marriage. London: Macmillan & Co.
Williams, G. C. (1957). Pleiotropy, Natural-Selection, and the Evolution of
Senescence. Evolution, 11(4), 398-411.
Williams, G. C., & Williams, D. C. (1957). Natural-Selection of Individually Harmful
Social Adaptations among Sibs with Special Reference to Social Insects.
Evolution, 11(1), 32-39.
Wilson, D. S., & Sober, E. (1994). Reintroducing Group Selection to the Human
Behavioral-Sciences. [Review]. Behavioral and Brain Sciences, 17(4), 585608.
Wilson, M., & Daly, M. (2004). Do pretty women inspire men to discount the future?
Biology Letters, 271, S177-S179.
Winterhalder, B., & Smith, E. A. (2000). Analyzing adaptive strategies: Human
behavioral ecology at twenty-five. Evolutionary Anthropology: Issues, News,
and Reviews, 9(2), 51-72.
203