Speaker(s)Anujit Chakraborty (University of California, Davis)
LocationUniversity of Amsterdam, Roeterseilandcampus, Room E0.09
Date and time
March 23, 2023
16:00 - 17:15
In dynamic settings, rational agents must reason contingently about future states and actions, but failures of contingent reasoning are ubiquitous. How should we model boundedly rational agents who fail to correctly forecast future actions? We conduct an experiment in which subjects must forecast their own actions, finding that they often fail to do so. We use the rich dataset to compare five potential models of behavior which differ in how boundedly rational subjects perceive future payoffs. A model in which subjects expect to make more mistakes when the utility consequences of their future actions are more similar best explains behavior.
(This is joint work with Chad Kendall)