London Judgement & Decision Making Group
Autumn term 2014 – 2015
Organizers
Neil Bramley
University College London
Contact details:
Department of Cognitive, Perceptual & Brain Sciences
Room 201, 26 Bedford Way, London, WC1H 0AP
UK
Telephone: +44 (0) 79 1441 9386
E-mail: neil.bramley.10@ucl.ac.uk
Leonardo Weiss-Cohen
University College London
Contact details:
Department of Cognitive, Perceptual & Brain Sciences
Room 200, 26 Bedford Way, London, WC1H 0AP
UK
Telephone: +44 (0) 20 7679 7570
E-mail: l.cohen.12@ucl.ac.uk
LJDM website
http://www.ljdm.info
LJDM members’ (Risk & Decision) list
Contact: Dr Marianne Promberger (marianne.promberger@kcl.ac.uk)
Seminar Schedule
October – December 2014
Wednesdays, 5:00 pm in Room 313, 26 Bedford Way, UCL Psychology
1st October
Why are we all here? The origins and history of LJDM
Peter Ayton1, Nigel Harvey2 and Larry Phillips3
1
City University, 2University College London and 3London School of
Economics
8th October
The role of expectancy in human Pavlovian Conditioning
Gabrielle Weidemann
University of Western Sydney
15th October
Subjective and objective decision making in scientific peer review
Mike Peacey
University of Bristol and University of Bath
22nd October
Giving and probability
David Reinstein
University of Essex
29th October
How do risk attitudes affect measured confidence?
Chris Starmer
University of Nottingham
5th November
READING WEEK (NO SEMINAR)
12th November
The dynamics of evidence accumulation and choice
Takao Noguchi
University College London
19th November
Cultural norms and identity in coordination games
Joanne Laban
University of Exeter
26th November
Heuristics as special cases of Bayesian Inference
Paula Parpart
University College London
3rd December
Separable processes in the generation of random sequences: Data and a model
Rick Cooper
Birkbeck University
10th December
Developmental differences in causal learning
Chris Lucas
University of Edinburgh
Abstracts
01.10.2014
Why are we all here? The origins and history of LJDM
Peter Ayton1, Nigel Harvey2 & Larry Phillips3
1
City University, 2University College London & 3London School of Economics
To start the year’s LJDM seminars, we thought that it would be interesting to those attending to
learn a little about how the London Judgment and Decision Making Group came into being. We
will talk about the origins of the seminar series, the development of research into Judgment and
Decision Making in London and the rest of the UK, and the influences on that research of other
work within cognition and related areas (e.g., thinking and reasoning, choice models of reaction
time, psychophysical judgment).
08.10.2014
The role of expectancy in human Pavlovian Conditioning
Gabrielle Weidemann
University of Western Sydney
Pavlovian conditioning, in which a neutral stimulus (usually referred to as a conditioned stimulus,
CS; e.g., a picture on a computer screen) is paired together with a biologically significant stimulus
(usually referred to as an unconditioned stimulus, US; e.g., a puff of air to the eye), results in the
acquisition of a reflexive response (usually referred to as a conditioned response, CR; e.g., a blink
to the picture). The extent to which this learning in humans is the result of a conscious expectancy
for the US is the source of much debate. In this talk I will present the results of some experiments
which show how expectancy precedes and predicts conditioned eyeblink responses, as well as the
results of some studies which show that it is possible to dissociate expectancy from responding. I
will explain how it is that these two sets of findings might be reconciled and how this might
illuminate the role of expectancy in human Pavlovian Conditioning.
15.10.2014
Modelling the effects of subjective and objective decision making in scientific peer review
Mike Peacey
University of Bath and University of Bristol
The objective of science is to advance knowledge, primarily in two interlinked ways: circulating
ideas, and defending or criticizing the ideas of others. Peer review acts as the gatekeeper to these
mechanisms. Given the increasing concern surrounding the reproducibility of much published
research, it is critical to understand whether peer review is intrinsically susceptible to failure, or
whether other extrinsic factors are responsible that distort scientists' decisions. Here we show that
even when scientists are motivated to promote the truth, their behaviour may be influenced, and
even dominated, by information gleaned from their peers' behaviour, rather than by their personal
dispositions. This phenomenon, known as herding, subjects the scientific community to an
inherent risk of converging on an incorrect answer and raises the possibility that, under certain
conditions, science may not be self-correcting. We further demonstrate that exercising some
subjectivity in reviewer decisions, which serves to curb the herding process, can be beneficial for
the scientific community in processing available information to estimate truth more accurately. By
examining the impact of different models of reviewer decisions on the dynamic process of
publication, and thereby on eventual aggregation of knowledge, we provide a new perspective on
the ongoing discussion of how the peer-review process may be improved.
22.10.2014
Giving and Probability
David Reinstein
University of Essex
When and how should a fundraiser ask for a donation from someone facing an uncertain “bonus”
income?” A standard model of expected utility over outcomes predicts that her Before choice—
her ex-ante commitment conditional on her income—will be the same as her choice After the
income has been revealed. Deciding “if you win, how much will you donate?” involves a
commitment (1) over a donation for a state of the world that may not be realized and (2) over
uncertain income. Models involving reference-dependent utility, tangibility, and self-signaling
predict more giving Before, while theories of affect predict more giving After. In our online field
experiments at UK universities, where commitments to donate were binding giving was larger and
more frequent in the Before treatment for male subjects, with a significant gender differential. Our
laboratory evidence also found significant effects for men and significant gender differences. Lab
treatments isolated distinct mechanisms: for men, donations were higher in all treatments where
the donation’s collection was uncertain whether or not the income was known. This supports a
self-signaling explanation: commitments realized with a lower probability must involve larger
amounts to have the same signalling power. Our results are directly relevant to fundraising and
volunteer-recruitment strategies, and offer further evidence that we need to exercise caution in
applying expected utility theory in the presence of social preferences.
29.10.2014
How do risk attitudes affect measured confidence?
Chris Starmer
University of Nottingham
We examine confidence in own absolute performance using two elicitation procedures: selfreported (non-incentivised) confidence and an incentivised procedure that elicits the certainty
equivalent of a bet based on performance. The former procedure reproduces the “hard-easy effect”
(overconfidence in easy tasks and underconfidence in hard tasks) found in a large number of
studies
using
non-incentivised
self-reports.
The
latter
procedure
produces
general
underconfidence, which is somewhat reduced when we filter out the effects of risk attitudes.
However, even after controlling for risk attitudes our incentivised procedure leads to significant
underconfidence, and does not lead to better calibration between confidence and performance than
non-incentivised self-reports. Finally, we find that self-reported confidence correlates significantly
with features of individual risk attitudes including parameters of individual probability weighting.
12.11.2014
The dynamics of evidence accumulation and choice
Takao Noguchi
University College London
When judging whether a bug is poisonous or whether a person is trustworthy, we often consider
multiple pieces of information. Some of the information may help us make better judgement,
while others may mislead our judgement. In this talk, I will address how order of information
considered impacts on speed and accuracy of judgement, and discuss implications for sequential
sampling models.
In sequential sampling models, an individual is assumed to accumulate evidence supporting a
judgement, and when the accumulated evidence reaches a response criterion, the individual makes
a judgement. In these models, accumulation rate is a key to understanding speed and accuracy of
judgement, and accumulation rate is typically assumed to be fixed until an accumulation reaches a
response criterion. Previous studies, however, report that especially when under time pressure, an
individual is likely to evaluate an alternative as attending attribute dimensions (e.g., Lamberts,
1995). These findings indicate that accumulation rate may be based on a subset of dimensions at
first, and gradually, more dimensions become incorporated --- accumulation rate is dynamic.
Dynamic nature of accumulation rate predicts that when an individual first attends to a dimension
that provides misleading information, this individual accumulates evidence for an incorrect
judgement at first. As a result, 1) the individual should be slower to make a correct judgement, as
the individual has to override initially accumulated evidence, and also 2) the individual should be
less likely to make a correct judgement. These predictions were empirically tested and generally
supported in various domains.
19.11.2014
Cultural Norms and Identity in Coordination Games
Joanne Laban
University of Exeter
We run experiments with a stag hunt and bargaining coordination game. Using a between-subjects
design, we vary the identity of the opponent between someone of the same culture or a different
culture. The idea is to see whether cultural norms or identity play a part in coordination decisions.
We compare the responses of British and Asian students at the University of Exeter and show the
cultural identity of the opponent by physical appearance. The players appear to use cultural
stereotypes to predict behaviour, especially in the bargaining game which may require more
strategic thought than the stag hunt game. In particular, the British act in way that indicates they
believe the Asians will behave more cautiously than other British. According to our results, the
stereotype of Asians being cautious is misleading.
26.11.2014
Heuristics as special cases of Bayesian Inference
Paula Parpart
University College London
Bayesian inference models have usually been cast as being rational and at odds with simple
heuristic approaches. We show that prominent decision heuristics, take-the-best and tallying, are
special cases of Bayesian inference. Why is it possible that simple heuristic are sometimes able to
outperform more complex strategies like linear regression? In my talk I will argue that covariance
between cues plays an important role and the fact that simple heuristics completely disregard
covariance in their cue validity estimates. We developed a Bayesian learning model by extending
a popular regularized regression approach (ridge regression) from machine learning. This
regularized regression approach contains a penalty term corresponding to a Bayesian prior that
reflects the amount of covariance in the environment. By relying on a latent state variable model
similar to multivariate regression, our Bayesian model enables us to smoothly move between a
model with lots of covariance (linear regression) to one with no covariance (heuristics).
Importantly, in the limit, i.e., for a large penalty parameter, the Bayesian model becomes
equivalent to the heuristics (cue validities). Thus, our model subsumes both regression and frugal
heuristics as extreme cases on a continuum of decision strategies.
Interestingly, the optimal performance can be found in the middle most of the time, i.e. between a
fully integrative approach and a frugal heuristic approach. This implies that true environmental
structure and potentially psychological processing might often lie somewhere between the
assumptions of heuristic and standard regression approaches.
03.12.2014
Separable processes in the generation of random sequences: Data and a model
Rick Cooper
Birkbeck University
When participants are required to generate “random” sequences of responses their behaviour
typically betrays a number of well-known biases, such as the avoidance of immediate repeats and
a tendency to produce stereotyped sub-sequences. These have been accounted for within verbal
models in which executive processes modulate schema-based behaviour. In this talk I will first
present data from a dual-task study which fractionates the involvement of at least two executive
processes in random generation. Following this, I will demonstrate how participant behaviour
within the dual-task study can be captured by a general computational model of the interaction of
executive sub-processes.
10.12.2014
Developmental differences in causal learning
Chris Lucas
University of Edinburgh
Children learn causal relationships quickly and make far-reaching generalizations from what they
observe. I will discuss results showing that in some cases they do so in a manner that is more
flexible -- and arguably better -- than adult learners. These results are compatible with at least two
kinds of explanations: (1) children have weaker a priori assumptions or expectations than adults,
which makes them more sensitive to atypical evidence; and (2) that children and adults respond to
evidence and update their beliefs in fundamentally different ways, reflecting differences in their
goals and circumstances. I will detail some variations on this latter explanation, drawing on
analogies to inference procedures in Bayesian models and optimization, as well as the exploreexploit trade off in active learning. Finally, as time permits I will mention some work in progress
that is designed to distinguish between these kinds of explanations and shed light on the processes
underlying causal inference.