Paradox of Animal Sociality,
Lecture #2002-10
TELLING "NEARLY WELL-FORMED" DARWINIAN
STORIES ABOUT BEHAVIOR
A. Sociobiological Explanation is a Degenerate form of
Darwinian Explanation. An explanation is degenerate when it has lost crucial features of
the original from which it is derived. Explanations offered as "Darwinian" differ in their degree of
degeneracy.
1. A Full Darwinian explanation of behavior would be a theory
about the cause of behavioral evolution, i.e., the cause of
adapted phyletic descent of behavior. It would assert that evolution is caused by
natural selection, i.e., by the differential replication of alternative types of organism.
2. Ethology is of necessity a degenerate theory, a theory that
provides natural selection explanations for behavioral adaptation but
is relatively silent about the phenomenon of evolution, itself. Since
direct evidence of phyletic descent of behavior is very rare, ethologists must usually
content themselves that behavior always evolves in concert with structure or with using
the comparative and methods to develop evidence of behavior adaptations and then using
field experiments to demonstrate that the adaptations were providing reproductive
benefits.
3. Sociobiology is a still more degenerate theory, one that explains
behavior without attributing any particular descriptive property to it.
Systematic study of evolution or adaptation of behavior is relatively unusual.
Sociobiology is mostl reticent concerning what sorts of organisms are likely to be
selected. Its project is to provide explanations for social behavior on a case-by-case
basis, with a focus on behaviors that seem to be difficult to explain by reference to
natural selection. Since it has no way of identifying up front which organisms are likely
to be selected, it is not only degenerate, it is circular The same is true, in the main, of
behavioral ecology and much of what passes for evolutionary psychology.
. For their explanations, sociobiologists rely heavily on "Nearly well-formed
Darwinian stories." I call these Darwinian stories Nearly Well formed, because the
ultimately well formed Darwinian story, explain established design features. Nearly-well
formed Darwinian stories explain, as I have argued, traits on a trait by trait basis. Still, so
as not to drive us all crazy in the next few weeks, I will refer to these Nearly Well formed
Darwinian stories as simply, “Darwinian Stories”.
B. Darwinian stories answer questions of the form, "Why do
contemporary members of species, S, display behavioral
trait,T?"
1. Answers are of the form, "Because over the history of Species, S,
those members that did T tended to have more offspring than those
that did not-T AND T-doers tended to have offspring that did T." Such
stories are Darwinian stories because they invoke crucial elements of the Natural Selection Model: all
species-members are taken to be members of a population (~ herd) are taken to vary in the possession of
some trait (T and not-T), this variation is taken to be heritable I call the stories told by sociobiologists,
"nearly well-formed", because they contain all the elements of the model EXCEPT the eugenic concept.
They offer no general characterization of the sort of organisms that nature is selecting.
2. Darwinian stories are often simplified and rendered in the form of
mathematical models that are called Evolutionary "Games Theory"
Models.
a. Games Theory Models originated in political science where they were used to
explore negotiating situations in which the best outcome for one participant depends
on what the other participant does. In a game theory model, two different behavioral
strategies are pitted against one another as follows. Each player is presumed to be able
to choose either of the two strategies and the game "designer" specifies the "payoff" that
each player will receive for each combination of strategies. These features of a games
theory model are often represented in tables such as the following:
Figure 2002-10-1
Payoff received by
Strategy A
Strategy B
Strategy A
Payoff {A against A}
Payoff of {A against B}
Strategy B
Payoff {B against A}
Payoff of {B against B}
player 1, when. …
Player 1 Plays…
Against Player 2, playing
The table can be read as a narrative with blanks in it that can be filled by reading out the
central cells of the table, thus "Payoff received by Player 1, when Player 1 plays
Strategy A (or b) against a second player who plays either Strategy A or B.
'
b. The most well known of these games models is the Prisoners' Dilemma Model.
The Prisoners' Dilemma game is a model of a conflict between individual and collective
interest. Anybody who has watched the crime show, Lore and Ordure, is familiar with
the model situation. Two robbers are caught, but the police lack sufficient evidence to
convict both of the crime. So, they take the robbers into separate rooms and offer each
the following deal. " If you confess to the crime and testify against your partner we will
give you a probation. If you don’t confess to the crime, and your partner testifies against
you, we will give you a heavy sentence. If you both confess, both will receive a moderate
sentence." What the police do not say, but which is implied is that if neither confesses,
both robbers will get off with a light sentence.
Figure 2002-10-2
Payoff received by
Prisoner 1, when ….
Against Prisoner 2, playing
Tough it
Rat
Prisoner 1 Plays
Tough it
Light Sentence
(-2 years)
Severe Sentence
(-5 years)
Rat
Probation
(0 years)
Moderate Sentence
(-4 years)
I have to admit that the prisoner's dilemma has never made sense to me, as a dilemma
that might occur on Lore and Ordure. . Why exactly would two confessors get a lighter
sentence? Moreover, even more implausible, why do two prisoners who Tough It end up
with more of a sentence than the one who Rats? Isn't the whole point that the police
cannot make a conviction unless one of the prisoners rats?1
1
This a wonderful example of untended surplus meaning of a model that is seriously misleading.
Detectives Cypovitz and Sorensen have the two lunkheads in custody, one in the cage and one in the coffee
room. They go back and forth between the two trying by any wile to undermine their faith in one another.
Now I don’t know how realistic these dramas are, but in every episode I have ever watched (and I confess
to having watched to many of them), each member of a pair of lunkheads in which both played "hang
tough" would do better than an individual who played "rat" to his partner's "hang tough". If Andy and
Danny didn't need to make their case, they wouldn't bother to interrogate, so if both lunkheads hang tough
they will both probably "walk". Furthermore, when negotiating with a potential "rat", the most the two
detectives can ever offer is to "put in a good word for you with the D.A." So unlike the classic "Prisoners
Dilemma", where both players are always paid to rat, the NYPD Blue dilemma actually involves an
quandary in which each lunkhead should hang tough if he thinks his partner is going to hang tough and rat
if he thinks his partner is going to rat.
c. The prisoners dilemma model should be known as the cooperation dilemma
model. Lets imagine that your room mate is a biology major and you are a psychology
major. Each of you is taking a course in the other’s major Department to fulfill a
perspective and neither of you is doing very well in that course. In fact, if you don’t take
some drastic measures, you both will get C’s . Your roommate suggests the following
deal: “Let’s help each other out. You can help me out preparing for my psychology
exam and then I will help you out preparing for your biology exam and we’ll both get
B’s. Sounds good, right? But wait a minute! If you fulfilled your part of the bargain and
your roommatedidn’t, you would get a D in the course because you would have been
distracted from your own studying and got nothing in return. So, if your room mate
doesn’t fulfill his side of the bargain, you certainly shouldn’t fulfill your side. Now
imagine that your roommate fulfilled his part of the deal. If you reneged on your part,
Figure 2002-10-3
And Student 2
Grade points received by
student 1 and by the
average of both players
when ….
Student 1
Follows Through
Reneges
Follows
Through
Reneges
3.0
Av. 3.0
1.0
Av.2.5
4.0
Av. 2.5
2.0
Av. 2.0
you would have his help and wouldn’t have wasted your time helping him. You might
get an A! Therefore, WHICHEVER your roommate does, follow through or renege, your
best course of action is to renege.
However, this isn't the worst of a social cooperation situation. Remember that
your roommate is capable of making the same calculations as you have and deciding to
renege. Once he has made them, he will see, as you just have seen, that he is always paid
to renege. So you know up front that if your roommate is a rational actor, he is going to
renege. And knowing that, why should you fulfill you part of the bargain.
Notice that this logic is true, even though the average grade that each of you would
receive would go up a half grade point if you both were to follow through on the deal.
e. Another example of a cooperation dilemma model is the Tragedy of the
Commons, so named by Garrett Hardin in a famous book by that name. The word
"Commons" in the title is a reference to the practice familiar in colonial America of
putting aside a space of pasture at the center of a village on which villagers could graze
their cattle. To make the example simple, lets imagine the simplest possible village, one
with just two farmers living around its common. The grass on the common is sufficient
to feed six cows that give 20 lbs of milk each. Therefore, if each farmer puts out three
cattle, both get 60 lbs of milk. If more than six cattle are put out to pasture, the pasture is
damaged and provides less milk, lets say, 10 pounds less for each additional cow.
Nevertheless, one of the farmers does the numbers and realizes that if he puts out an
additional cow he will make more milk. For seven cattle, the pasture produces only 110
pounds of milk, or just under 16 lbs per cow. However, the farmer with 4 cows takes
home 62.8 lbs, which is 2.8 lbs more than he took home with only three cows. The
second farmer will get just over 47 lbs, a reduction of nearly 15 lbs.
So now, the second farmer also does the numbers. If he puts a fourth cow out on
the pasture, the pasture will have eight cows on it and provide 100 lbs of milk. So now,
each former gets 12.5 pounds per cow. With his 4 cows out on the pasture the second
farmer now gets 50 lbs of milk, more than he got since the first farmer increased HIS
herd but less than he would have gotten if neither had increased his herd. But the first
farmer now gets only 50 lbs of milk, which is much less than the nearly 63 lbs he was
getting before farmer 2 put our his fourth cow, and even less than farmer 1 was getting
before he upset that balance by putting out his forth cow. You see why this scenario is
called a tragedy -- at each stage, each farmer is doing something that betters himself but
the overall effect is to ruin both farmers.
The Tragedy of the Commons is represented as a cooperation dilemma in the
diagram just above. .
Cooperation dilemma models -- whether they be a Prisoners' Dilemma models, a
Figure 2002-10-4
And Farmer 2 Puts out
Milk received by farmer
1 and by the average of
both farmer when. …
Farmer 1 Puts out
3 cows
4 cows
3 Cows
60lbs
Av. 60lbs
63lbs
Av. 55lbs
4 Cows
47lbs
50lbs
Av.55lbs
Av. 50lbs
c-student models, or a tragedy of the commons models -- all have certain common
features. What a non-cooperator gets from a cooperator is greater than what a cooperator
gets from a cooperator, which in turn is greater than what two non-cooperators get
working independently get which is greater than what a cooperator gets in an interaction
with a non-cooperator. In addition, the total received by two cooperators working
together must be greater than the sum of a cooperator and a non-cooperator "working
together".
Cooperation dilemma models occur so often in the sociobiological literature that they
appear in textbooks and articles in the following highly standard format.
Figure 2002-10-5
Against Player 2, playing
Payoff received by
Cooperate
player 1, when ….
Player 1 Plays
Cooperate
Defect (Cheat)
Defect (Cheat)
R
S
Reward for Cooperation
Sucker's Payoff
T
P
Temptation to Defect
Punishment for failure to
Cooperate
A table is a cooperation dilemma table if (and only
if) T > R > P > S and 2 R> (T + S)
The terms in this standard cooperation table may require some clarification . For
the purposes of such tables, to "cooperate" is to honor the contract with one's associate,
to "defect" is to fail to honor it. Defecting is sometimes called "Cheating". Those of
you who watch a lot of crime shows may notice that the word "cooperate" has the
opposite meaning in this table than it does in Lore and Ordure. In Lore and Ordure , to
"Cooperate" is to work with the police to maximize your own individual outcome -- in
other words, "cooperate" means "rat". However, in the standard sociobiological
representation of cooperation dilemma tables, "cooperate" means "work with your partner
to maximize your joint outcome.” Conventionally, the cells an a cooperation dilemma
table are identified with the letters R (= Reward for Cooperation), T (=Temptation to
Defect), S (= Sucker's Payoff), and P (=Punishment for Failure to Cooperate). For a table
to represent a social dilemma model, it must be the case that payoffs represented by these
letters have the indicated relative magnitudes.
d. Evolutionary Games Tables. When the Strategies of a games table are
alternative heritable types of organisms within a species and the payoffs in the table are
increases to their fitness, then the table becomes and evolutionary games table. Here is
a standard Evolutionary Games Table.
Figure 2002-10-6
Payoff received
by individuals
that…..
….play…
Strategy
A
…..against individuals playing….
Strategy A (of
Strategy B (of which
which there are p in there are (1 - p) in the
the population)
population
Total
Payoff
( p)(P{ A:A})
+
(1-p)(P{A:B})
p(Payoff {A against A})
(1-p)(Payoff of {A against B})
p(Payoff {B against A})
(1-p)( Payoff of {B against B}) ( p)(P{ B:A})
+
(1-p)(P{B:B})
Strategy
B
Strategy A will come to characterize the population when the total payoff to Strategy A
is greater than the total payoff to Strategy B: i.e., when ( p)(P{ A:A}) + (1-p)(P{A:B})
is greater than ( p)(P{ B:A}) + (1-p)(P{B:B})
Evolutionary games tables are grossly similar to ordinary games tables. The
similarities are found in the payoffs at the center of both kinds of tables, payoffs that
occur to the individuals playing one strategy against one another and against individuals
playing an alternative strategy. In an evolutionary games table, the payoffs are fitness
values. They represent increases (or decreases) to the reproductive rate of individuals
playing one strategy as opposed to the other. To pack the identifications of payoffs into
the cells of the table , we use a symbol system, such that, for example, the “Payoff of A
playing against B” is shortened to “(Payoff{A against B} or even more compactly,
“P{A:B}”.
The two types of tables are also different in some very important ways. Instead of
individuals playing different strategies, the strategies are played by different types of
individuals within the population. Each individual plays one and only one strategy, A or
B.. An ordinary games table answers the question, what will be the payoff for each
player for each combination of plays the players might make in a single round game in
which there are two players and two strategies. An evolutionary games table answers the
question, "Assuming that there are two types of players within the species, what will be
the effect on these two types if they come in contact with one another at random within
the population?" In an ordinary games table only one of the four outcomes can happen in
a round of game. In an evolutionary games table, each possible outcome occurs with a
relative frequency determined by the frequency of the two types within the population,
and we calculate the total outcome by multiplying the frequency payoff by its value.
Confusingly for our purposes, the proportion of the A-type individuals in the population
is usually represented by a small p and the proportion of the alternative, B-type, is
represented by 1-p. So there are two kinds of p’s in the table, Big P’s that stand for
“Payoff” and little p’s that stand for “proportion..
The model underlying an evolutionary game table is thus substantially different.
Imagine that a population has two mutually exclusive, alternative genetic types, A and B,
say, blue eyes and brown eyes. . The two types associate at random within the
population, A's meeting A's, A's meeting B's, B's meeting A's and B's meeting B's with a
frequency that is proportional to their representation in the population. So, for instance,
if A's represent a third of the population, then A's will encounter other A's a third of the
time and B's two thirds of the time.
One of the values of evolutionary games tables is that it makes it possible to talk
about how the fortunes of two different types within a population will be affected by their
relative numbers. You might get an intuitive grasp of how initial frequencies can effect
outcomes in an evolutionary game, if you think about an old childhood game, rock,
scissors, paper. The game is ordinarily played as a two-player, three strategy game, in
which each of two children can deploy one of the three strategies and in a given round,
the outcome is determined by what the other player deploys. In the language of the
game, Rock crushes Scissors, Scissors cuts Paper, and Paper covers Rock. The two
players hold their hands behind their back and at the count of three; make their choices
simultaneously by putting out a fist (Rock), a v-shape with their middle and index finger
(Scissors) a flat hand with palm down (Paper). . Rock wins over Scissors, Scissors over
Paper, Paper over Rock.
To turn this children's game into something like an evolutionary game, we need
only imagine a tournament in which 100 people each playing one of the three strategies,
rock, scissors or paper, as they mill about encountering one another at random. The
outcome for any strategy obviously depends on the proportions of players playing the
remaining two strategies. The best outcome for Rock-players occurs in a tournament
with many Scissors-players and few Paper-players. The best outcome for Scissorsplayers occurs in a tournament with lots of Paper-players and no Rock-players. Finally,
the best outcome for Paper-Players occurs in a tournament with many Rock-players and
few Scissors players. Thus, which strategy is the winning strategy depends on which
OTHER strategies are present in the game. Think about what would happen for different
proportions of the three strategies. We will encounter this situation frequently as we
discuss the evolution of cooperation, but before we get to this complicated material we
have to get a firmer grip on how an evolutionary games table relates to Darwinian
explanation.
e. Any Nearly -Well-formed Darwinian Story can be represented as an
evolutionary games table. Imagine that there are two sorts of creatures in the
population. In this scenario, all the creatures have the same genome except for one
difference. For our purposes, he term "genome" just means "the collection of inherited
Figure 2002-10-7
Payoff received by
Individuals that. …
…. Against individuals playing…
Strategy
Strategy (G) with
(G+T) with
probability (1 - p)
probability p
…play …
Strategy (G+T) (p)(b)
(1-p)(b)
p(0)
(1-p)( 0)
Strategy (G)
Total
Payoff
(p)(b)+(1-p)b=
pb+b-pb=
b
( p)(0)+(1-p)( 0)=
0
Strategy G+T will come to characterize the population when the total payoff to
Strategy G+T is greater than the total payoff to Strategy G: i.e., when b > 0, which
will be true whenever trait T provides a positive benefit to its bearers.
traits that an individual has". We'll call these two types the G-type and the (G+T) type.
To see how this works, let's imagine that the Trait conveys some benefit to trait bearers,
call it 'b'. A benefit is any event that increases the reproductive rate of an individual to
which it occurs. For the moment, we'll imagine that the trait has nothing to do with social
interaction, so it conveys the same benefit whether the trait bearer is in the company of
other trait bearers, or whether it is in the company of non-trait-bearers. Now we can fill
in an evolutionary games table for this trait as in figure 10-7. This table represents a
Darwinian story which can be used to explain why in contemporary times, genomes in
the species tend to have the particular trait. They do, the Darwinian Story says, because
in the history of the species Genomes with the trait T (G+T) had a higher payoff across
its two kinds of interactions than genomes without that trait. Notice by looking at the
totals column that G+T has a higher number just so long as b is greater than zero. Thus,
as long as the benefit provided by b really IS a benefit (a positive number) then traitbearers will increase in the population by comparison with non-trait-bearers.
Figure 2002-10-8
Payoff received by
Individuals that ….
…against individuals playing….
Strategy
Strategy (G) with
(G+T) with
probability
probability p (1 - p)
… play ….
Strategy (G+T) (p)(-c)
Strategy (G)
p(0)
(1-p)(-c)
(1-p)( 0)
Total
Payoff
(p)(-c)+(1-p)(-c)=
- pc-c+pc=
-c
( p)(0)+(1-p)( 0)=
0
Strategy G+T will not come to characterize the population unless the total payoff to
Strategy G+T is greater than the total payoff to Strategy G: i.e., -c > 0, which will
be true only when trait T provides a negative cost, a "cost" that is really a benefit.
Notice that the opposite occurs if Trait T levies a cost on individuals to which it is
added and if a cost is understood as any event that decreases the reproductive rate of any
individual to which it occurs.
Of course, many traits convey both costs and benefits. To model such traits, we need to
enter both cost and benefit terms into the evolutionary games table, as we do in Figure
2002-10-09 on the next page. Again, we look at the Totals column, where we see that
G+T will be greater than G just so long as b-c is greater than 0, or b>c. Thus, so long as
the fitness benefits of a trait outweigh its costs, the trait will come to characterize the
species.
Figure 2002-10-09
Payoff received by
Individuals that ….
...play…
Strategy
(G+T)
Strategy (G)
…against individuals playing…
Strategy
Strategy (G)
(G+T) with
with probability
probability p (1 - p)
(p)(b-c)
(1-p)(b-c)
p(0)
(1-p)( 0)
Total
Payoff
(p)(b-c)+(1-p)(b-c)=
1 (b-c)=
b-c
( p)(0)+(1-p)( 0)=
0
Strategy G+T will come to characterize the population if the total payoff to
Strategy G+T is greater than the total payoff to Strategy G: i.e., b-c > 0: i.e.,
whenever trait T provides costs that are greater than the benefits it provides. .
To clarify these concepts, consider the following example. What sort of
Darwinian story would we tell to explain the fact that many animals -- cats, for instance - spend considerable time grooming themselves. Given fleas, mites and other skin
parasites, keeping the fur tidy and clean presumably conveys benefits to cats that do it.
But the time spent grooming is time taken away from other activities -- looking out for
predators, seeking food, or just catching up one one's rest. The table above shows that
just so long as
 at least a few members of the species are capable of performing the behavior, and
 the behavior is an inherited one, and
 the benefits of the behavior exceed its costs,
the behavior will come to characterize the species. Alternatively, to put the explanation
as a nearly well formed Darwinian story, cats groom themselves because, in the history of
the species, some cats were capable of grooming, grooming cats had more offspring than
non-grooming cats, and grooming cats had offspring that tended to groom.
Benefits are necessary to natural selection but not all benefits lead to natural
selection. Returning to our grooming cats, for instance, we can say that being groomed
by Cat B would benefit Cat A. But this fact will have no obvious implications for
natural selection unless "being groomed by Cat B" is somehow a consequence of some
heritable trait of Cat A. Natural selection is blind to any benefit that does not provide a
reproductive advantage to the heritable trait that makes the benefit possible. However,
this blindness of evolutionary theory poses an important problem for sociobiologists. If
natural selects for traits that benefit only the bearers of the traits, then how do
contemporary biologists explain the many behaviors that animals perform that seem to
provide benefits to others. Why would animals ever perform costly behaviors that benefit
others? Such behaviors would seem to be a challenge to Darwinism because any costly
trait that caused individuals to pass out benefits to their competitors would seem always
to be contrary to Darwinian interest. Why, then, are creatures ever nice to one another?