Essays in Behavioral Public Economics

Abstract

This dissertation consists of three essays on economic design under empirically grounded models of how people form beliefs and make decisions. A running thread is that the design of institutions—markets, policies, and organizations—must account for the mental models decision-makers bring to them. Voters may underappreciate equilibrium effects, users may project human patterns onto artificial intelligence, and professionals may fail to correctly aggregate information. The first chapter studies the political feasibility of efficient reforms when voters underappreciate equilibrium effects, and shows how contingent compensation can make policies such as congestion pricing both welfare-improving and politically feasible. The second chapter studies how people learn about artificial intelligence (AI), showing that human-like interfaces can induce users to evaluate AI through human categories of difficulty, ability, and mistake reasonableness, distorting beliefs and adoption decisions. The third chapter studies information design for AI-assisted decisions, showing that calibrated coarsening of algorithmic signals can improve professional decision-making when users deviate from Bayesian updating. Methodologically, these essays combine formal models, experiments, and surveys to show how policies, interfaces, and information structures can be designed around the motives, beliefs, and decision processes of the people they affect.