Working papers
Categorization in Games: A Bias-Variance Perspective (with Jehiel, P.) R&R AEJ micro
Idea: We present a framework for endogenous categorisation in games. On-path situations (nodes, information sets, or types) are distinguished perfectly, due to abundance of data, whereas off-path situations have to be categorised coarsely due to scarcity of data, bundling similar situations together. Application of the framework to classic examples yield novel predictions.
This work incorporates our previous working paper “Cycling and Categorical Learning in Decentralized Adverse Selection Economies”
"Preferences as Heuristics" (with A. Rigos) submitted
Idea: We propose a theory of how subjective preferences for actions develop in response to objective payoffs. People regularly make mistakes when figuring out what action to take. Adjusting subjective preferences can change the likelihood of different objective mistakes. Strengthening a preference for an action decreases the chance of wrongly avoiding it but raises the risk of wrongly choosing it. Preferences adapt to balance these mistakes. Cultural differences in preferences, and the resulting behaviours, are thereby explained in terms of differences in incentive structures across societies. In very general terms, our model implies that people will develop preferences for actions that tend to be beneficial given the incentives that prevail in their social environment.
Work in progress
“An Experiment on Observations on Cooperation” (with Heller, Y., and Embrey, M.)
"Comparing Models of Learning in Games" (with Afsar, A., Fudenberg, D., and Karreskog Rehbinder, G.)
"Normatively Relevant Narratives" (with Rigos, A., and Andersson, L.)
Retired working paper
“Asymptotically Optimal Regression Trees”
Idea: Optimal locally adaptive bin-size is derived for a version of regression trees. Consistency and asymptotic normality of the estimator is proved.