9515421 Kalai This project studies issues in rational interactive social learning. This includes learning in long term strategic interaction, e.g., firms engaged in market competition, and in recurring short interactions, e.g., real estate transactions involving different buyer-seller pairs. It develops further theoretical results and applies the theoretical findings to specific problems in economics. The theoretical developments include the following subjects: (1) asymptotic accuracy of forecasts as captured by notions of merging; (2) testing forecast performance using past realization as done by notions of calibration; (3) relationships between merging and calibration; and (4) economic and game-theoretic calibrated forecasts. The applications include the following issues: (1) how long-term learning affects classical models of industrial organization, e.g., pricing, bargaining and auctions; (2) how social learning can be guided to bring about social efficiency; (3) developing long-term implementation schemes; and (4) developing tests that enable experimentalist to determine, based only on observed past realizations, whether a long-run or recurring equilibrium is being played.
