Variability in learning and decision-making:

failure or feature?

Workshop description

Humans do not react to their volatile environments in a reflexive manner, but rather choose which action to perform in response to a given, often multidimensional context. Despite the importance of understanding how we make judgments and decisions, it remains unclear what are the fundamental principles that explain a large variety of apparently irrational phenomena such as systematic biases – that are often surprisingly preserved across species. This observation stands in addition to the puzzle behind what actually causes behavioral variability over short and long time scales, which is usually ascribed to distinct sources of noise corrupting otherwise ideal behavior. A recent line of research proposes to address these two puzzles using a single 'resource-efficient coding' framework. Should the neural sources of noise that guide seemingly 'suboptimal' behavior be seen as unavoidable failures of limited biological systems? Or can biological agents take advantage of these noise sources as useful features to 'optimize' their behavior?

One of the reasons for these puzzles might be the lack of consensus across disciplines (neuroscience, psychology and economics) on how decisions emerge and what actually constitutes an 'optimal' or 'rational' decision in light of variability observed both within and across individuals. This ambiguity is caused, perhaps, by the fact that each field of research has attempted to solve distinct behavioral phenomena in isolation. Fortunately, this has been changing in the last years, as neuroscientists, psychologists and economists are starting to work together, based on the realization that several behavioral phenomena may have a common foundation spanning across disciplines. The objective of this workshop is to bring people from these different fields of research together, and create a constructive debate and eventual consensus on how current findings and future research from different disciplines might be brought together. The long-term goal of such multidisciplinary effort is to move toward a general computational framework for understanding the sources of decision variability and biases in humans and other animals.

Date, time & location

The workshop will take place at the 4th Multidisciplinary Conference on Reinforcement Learning and Decision-Making (RLDM 2019) at McGill University in Montréal, QC, Canada, on July 10, 2019, from 1.00pm till 5.00pm.

The workshop will be chaired by Rafael Polania (ETH Zürich, Switzerland) and Valentin Wyart (Ecole Normale Supérieure, Paris, France). For more information, please contact us by email ( or or Twitter (@RafaPolania or @valentinwyart).