Jörg Rieskamp talks with
ScienceWatch.com and answers a few questions about
this month's Fast Moving Front in the field of Social
Sciences, general.
Article: SSL: A theory of how people learn to
select strategies
Authors: Rieskamp, J;Otto, PE
Journal: J EXP PSYCHOL-GEN, 135 (2): 207-236 MAY 2006
Addresses: Max Planck Inst Human Dev, Lentzeallee 94,
D-14195 Berlin, Germany.
Max Planck Inst Human Dev, D-14195 Berlin, Germany.
Univ Warwick, Dept Psychol, Coventry CV4 7AL, W Midlands,
England.
Why do you think your paper is highly
cited?
The idea that people do not always solve a cognitive problem with the
identical cognitive tools has been assumed in various areas of psychology.
Even children already have access to different cognitive strategies that
could be applied to identical problems. This assumption explains why people
sometimes display different behavior in situations that are very similar.
However, when people have various strategies at their disposal, the
question arises about how people select strategies. Our article tackles
this strategy selection problem by suggesting that people learn to select
strategies on the basis of reinforcement learning. The theory is very
successful at predicting under which circumstances people select a specific
strategy.
Does it describe a new discovery, methodology, or
synthesis of knowledge?
Our investigation brings together two lines of research. The first line of
research argues that people make decisions by selecting different
strategies from a tool box. The second line of research argues that people
make decisions by learning from experience which decisions lead to good
outcomes. We brought both lines of research together, by suggesting a
learning theory that assumes that the objects of reinforcement are not the
single decisions, but rather people's cognitive strategies.
"My research follows a computational
approach that specifies the cognitive
mechanisms underlying people's
behavior."
On the basis of the strategies' success or failure, some strategies become
more or less likely to be selected. In a nutshell, we have developed a
cognitive learning theory that goes beyond pure behavioral reinforcement
learning.
Would you summarize the significance of your paper
in layman's terms?
It is often surprising to observe how differently people behave in the same
situation; even the same person often shows rather inconsistent behavior.
To explain this finding, it can be assumed that people have a repertoire of
strategies to solve judgment and decision problems. When they apply a
strategy that focuses on single pieces of information, their choices can be
different than if they had applied a strategy that integrates all available
information.
To explain which strategy a person is using, we assume that people monitor
the success of the strategies they are applying. After they have gained
substantial experience with a decision problem, they are most likely to
select a strategy that solves the problem well.
How did you become involved in this research and
were any particular problems encountered along the way?
The research approach that follows from the idea that people are equipped
with a repertoire of strategies to solve the problems they face has been
criticized repeatedly for not specifying a theory of how strategies are
selected. Furthermore, the approach has been criticized for being hard to
test against competing approaches, because it does not specify a general
theory of strategy selection. To address this criticism I became involved
in putting forward a general theory that could describe how people select
strategies.
The idea that learning processes could function as a selection device
appeared attractive from the beginning of the project. However, there are
various learning approaches and learning models that could have been
applied for this purpose. The difficulty was in finding a model that would
be able to describe the empirical findings accurately without creating a
too-complex theory that would face the problem of over-fitting. In the end,
we selected a fairly simple model that was able to describe the results
quite well.
Where do you see your research leading in the
future?
My research follows a computational approach that specifies the cognitive
mechanisms underlying people's behavior. I am convinced that cognitive
models of decision-making lead to better explanations of human behavior and
to more precise predictions. These models often make better predictions
when taking learning processes into account; that is, assuming that people
adapt to the environment they face. Cognitive models can explain why people
often make decisions inconsistent with traditional, normative models of
behavior.
Do you foresee any social or political implications
for your research?
Past research has shown that human behavior often does not obey the rules
of rationality as they have been stated in traditional economic theory. Our
research can explain this finding by arguing that people often select
simple strategies that are sufficiently able to solve decision problems.
However, the research also shows that when people are provided with
learning opportunities, they will select those strategies that perform well
for a particular problem. Our research implies that it can often be
important to structure a decision situation in a way that people will adapt
to it quickly.
Prof. Dr. Jörg Rieskamp
Department of Psychology
University of Basel
Basel, Switzerland