London Judgment & Decision Making Group
Spring term 2012 – 2013
Organizers
Emmanouil Konstantinidis
University College London
Contact details:
Department of Cognitive, Perceptual & Brain Sciences
Room 204c, 26 Bedford Way, London, WC1H 0AP
UK
Telephone: (+44) 020 7679 5364
E-mail: emmanouil.konstantinidis.09@ucl.ac.uk
Neil Bramley
University College London
Contact details:
Department of Cognitive, Perceptual & Brain Sciences
Room 201, 26 Bedford Way, London, WC1H 0AP
UK
E-mail: neil.bramley.10@ucl.ac.uk
LJDM website
http://www.ljdm.info
Web administrator:
Dr Stian Reimers (Stian.Reimers.1@city.ac.uk)
LJDM announcement emails
Contact: Dr Marianne Promberger (marianne.promberger@kcl.ac.uk)
Seminar Schedule
January – March 2013
5:00 pm in Room 313, 26 Bedford Way, UCL Psychology
9th January
The irrationality of categorical perception
Steve Fleming
New York University and University of Oxford
16th January
Multiple developments in children's counterfactual thinking
Sarah Beck
University of Birmingham
23rd January
Limits in decision making arise from limits in memory retrieval
Bradley Love
UCL, University of London
30th January
Children’s causal structure learning
Teresa McCormack
Queens University, Belfast
6th February
An integrative view on algebraic models and heuristics y of risky choice
Thorsten Pachur
Max Planck Institute for Human Development, Berlin
13th February
NO SEMINAR – UCL READING WEEK
20th February
Inductive Logic for Automated Decision Making
Jon Williamson
Kent
27th February
Decisions, Variability and Risk
Ulrike Hahn
Birkbeck, University of London
6th March
Criminal Sentencing as a Quasi-rational Cognitive Activity
Mandeep Dhami
University of Surrey
13th March
Does the “Why” Tell Us the “When”?
Christos Bechlivanidis
University College London
20th March
Agency Under Associative Control
Robin Murphy
University of Oxford
Abstracts
09.01.2013
Steve Fleming
New York University and University of Oxford
The irrationality of categorical perception
Categorical perception is ubiquitous in psychology. The perceptual system often settles on one or
other interpretation of an ambiguous stimulus, such as a Necker cube, even when a behavioural
response is not required. Such categorization is in direct tension with normative decision theory,
which mandates that in the face of uncertainty, the utility of various courses of action should be
weighted by the agent’s belief in alternate states of the world. If belief is collapsed to a single
state, then choices may be suboptimal due to neglecting their costs and benefits under other
possible states. We tested for such irrationality in a task that required observers to combine
sensory evidence with action-outcome uncertainty. Observers made rapid pointing movements to
targets on a touch screen, with rewards determined jointly by uncertainty in stimulus identities and
movement endpoints. Across both visual and auditory decision tasks, observers consistently
placed more weight on sensory evidence than action consequences. This asymmetry was
accounted for by a model in which an internal evidence threshold led to categorical perception on
a subset of trials, thus precluding sensitivity to utilities associated with the alternate perceptual
state. Our findings indicate that normative decision-making may be fundamentally constrained by
the architecture of the perceptual system.
16.01.2013
Sarah Beck
University of Birmingham
Multiple developments in children's counterfactual thinking
The first studies on the development of counterfactual thinking focussed on one question: whether
there was a shift in children's speculation about what might have been at 3-4 years of age. Since
then findings from a diversity of tasks have suggested that children's abilities develop somewhat
earlier (German & Nichols, 2003; Harris, 1997), later (Beck et al., 2006; Rafetseder, CristiVargas, & Perner, 2010), or that the emergence of adult-like counterfactual thinking (e.g. shown
by regret) might be separate from the basic reasoning abilities (e.g. Guttentag & Ferrell, 2004;
Weisberg & Beck, 2010; in press). I will explore which of the developmental data offer good
evidence for counterfactual thinking and identify questions that remain.
23.01.2013
Bradley Love
University College London
Limits in decision making arise from limits in memory retrieval
Do humans and machine systems make difficult decisions in a similar fashion? Some decisions,
such as predicting the winner of a baseball game, are challenging in part because outcomes are
probabilistic. One view is that humans stochastically and selectively retrieve a small set of
relevant memories when making such decisions. We show that optimal performance at test is
impossible when retrieving information in this fashion, no matter how extensive training is. One
implication of this view of human memory retrieval is that people, unlike machine systems, will
be more accurate in predicting future events when trained on idealized than on the actual
distributions of items. In others words, we predict the best way to convey information to people is
to present it in a distorted, idealized form. Idealization of training distributions is predicted to
reduce the harmful noise induced by immutable bottlenecks in people's memory retrieval
processes. These conjectures are strongly supported by several studies and supporting analyses.
People's test performance on a target distribution is higher when trained on an idealized version of
the distribution than on the actual target distribution. Optimal machine classifiers modified to
selectively and stochastically sample from memory match the pattern of human performance.
These results have broad implications for how to train humans tasked with important classification
decisions, such as radiologists, baggage screeners, intelligence analysts, and gamblers.
30.01.2013
Teresa McCormack
Queen’s University, Belfast
Children’s causal structure learning
In a series of studies, we have examined children’s ability to learn the structure of simple threevariable mechanical causal systems. We have found that children seem to have difficulty using
statistical information to learn causal structure, whether this is provided through observation of
the operation of a probabilistic system or through demonstrating the effects of interventions on the
system. Children, and in some cases adults too, are likely to rely on simple temporal cues to make
such judgments, even when this conflicts with statistical information. Children also have difficulty
predicting the effects of interventions on a causal system even if they are explicitly told its
structure. These findings are not straightforwardly predicted by the Causal Bayes Net account of
children’s causal structure learning.
6.02.2013
Thorsten Pachur
Max Planck Institute for Human Development, Berlin
An integrative view on algebraic models and heuristics of risky choice
Two prominent approaches to describe how people make decisions between risky options are
algebraic models and heuristics. The two approaches are based on fundamentally different
algorithms and thus are usually treated as antithetic, suggesting that they might be
incommensurable. Using cumulative prospect theory (CPT; Tversky & Kahneman, 1992) as an
illustrative case for an algebraic model, we demonstrate how algebraic models and heuristics can
mutually inform each other. Specifically, we highlight that CPT characterizes decisions in terms
of psychophysical characteristics (e.g., diminishing sensitivity to probabilities and outcomes) and
descriptive constructs such as risk aversion and loss aversion, and we argue that this holds even
when the underlying process is heuristic in nature. Fitting CPT to choices generated by various
heuristics, we find (a) that CPT is able to represent choices generated by heuristics with a good
model fit; and (b) that the heuristics generate characteristic parameter profiles in CT that reflect
the process architectures of the different heuristics in a psychologically meaningful way. Using
this approach we illustrate how CPT can be used to track how a heuristic’s degree of risk aversion
changes across different environments. Despite CPT’s ability to accommodate heuristic choices
rather well, model recovery analyses showed that the heuristics and CPT can still be well
distinguished under reasonable amount of noise in the data. Our results demonstrate that algebraic
models and heuristics offer complementary rather than rival modeling frameworks and highlight
the potential role of heuristic principles in information processing for prominent descriptive
constructs in risky choice.
20.02.2013
Jon Williamson
University of Kent
Inductive Logic for Automated Decision Making
According to Bayesian decision theory, one's acts should maximise expected utility. To calculate
expected utility one needs not only the utility of each act in each possible scenario but also the
probabilities of the various scenarios. It is the job of an inductive logic to determine these
probabilities, given the evidence to hand. The most natural inductive logic, classical inductive
logic, attributable to Wittgenstein, was dismissed by Carnap due to its apparent inability to
capture the phenomenon of learning from experience. I argue that Carnap was too hasty to dismiss
this logic: classical inductive logic can be rehabilitated, and the problem of learning from
experience overcome, by appealing to the principles of objective Bayesianism. I then discuss the
practical question of how to calculate the required probabilities and show that the machinery of
probabilistic networks can be fruitfully applied here. This culminates in an objective Bayesian
decision theory that has a realistic prospect of automation.
27.02.2013
Ulrike Hahn
Birkbeck, University of London
Decisions, Variability and Risk
The talk examines critically whether Expected Utility theory, the dominant normative framework
for the evaluation of decisions within philosophy, psychology, and economics deals appropriately
with variability or risk. It may be argued that the axiomatic foundations of utility theory provide
insufficient grounds for acceptance as a normative framework, in particular in contexts of highly
skewed distributions and one-off gambles. Simulation results aimed at the question of whether
alternative strategies may fare equally well or better under such circumstances will be discussed
and related to psychological data.
06.03.2013
Mandeep Dhami
University of Surrey
Criminal Sentencing as a Quasi-rational Cognitive Activity
Criminal sentencing is a complex cognitive activity that is often performed under suboptimal
conditions. It requires judges to apply intuitive and analytic judgment. Evidence suggests that
sentencers may not behave according to legal policy and training, and so several jurisdictions have
introduced guidelines to aid sentencing practice. I present recent research demonstrating that
sentences meted out in real cases are not being tailored to fit the characteristics of the individual
offence and offender, and that both the existing and new guidelines may be ineffective in helping
sentencers achieve this goal. Sentencing is not an intractable problem, and I propose the design of
more precise and comprehensive flowchart-type guidelines. However, this requires making our
notions of justice and fairness explicit and defining the concept of quasi-rationality.
13.03.2013
Christos Bechlivanidis
University College London
Does the “Why” Tell Us the “When”?
Traditional approaches to human causal judgment assume that the perception of temporal order
informs judgments of causal structure. We present two experiments where people follow the
opposite inferential route, where perceptual judgments of temporal order are instead influenced by
causal beliefs. By letting participants freely interact with a software-based “physics world”, we
induced stable causal beliefs that subsequently determined participants’ reported temporal order of
events, even when this led to a reversal of the objective temporal order. We argue that for short
timescales, even when our temporal resolution capabilities suffice, our perception of temporal
order is distorted to fit existing causal beliefs.
20.03.2013
Robin Murphy
University of Oxford
Agency Under Associative Control
Current research exploring the causes of volition argue for a ‘neurodualism’ in which the classic
distinction between instrumental action and Pavlovian reflex is found in separate neural pathways.
This endeavour pushes the causal mystery of voluntary behaviours around the brain without
necessarily developing our understanding of how the decisions related to voluntary actions
emerge. Associative learning mechanisms suggest that, at least part of, our sense of agency is the
product of a competitive learning network. Several experiments in which participants are required
to learn to control the occurrence of a novel outcome are described exploring 1) how perceptions
of control seem to emerge from learning processes and 2) that even with the same learning
experience individuals vary in how agentic they feel. This variability might be due to cognitive
processes correlated with mood state. Two interference experiments one using a neuroinhibitor
(Chase et al., 2011) and another a neuroenhancer (Msetfi et al., in prep) illustrate the role of the
serotonergic pathways for learning and our perception of agency.