Investigating the functional match between punishment and learning.

Brown University, Providence RI

Investigators

Abstract

This research proposal seeks to explain the way we punish by investigating the way we learn. It is motivated by a conclusion of research in psychology, biology and economics: Punishment functions to change the behavior of social partners, converting them from harm-doers to do-gooders. Consequently, human intuitions about punishment - and legal institutions that in part reflect those intuitions - may exhibit a match to the capacities and constraints of human learning. In other words, people will punish in ways and contexts that facilitate learning, and will avoid punishing in ways and contexts that fail to do so. The proposed research targets two case studies. The first is the punishment of accidental outcomes. For instance, consider two drunk drivers who each fall asleep at the wheel. One hits a tree; she can expect an expensive ticket and a suspended license. Another hits and kills a person; he can face years in prison. Their behavior is identical, but the "luck" of the outcome leads to radically different punishments. Why? The proposed research tests an explanation grounded in the human capacity to learn from punishment, in particular the punishment of accidental outcomes. This is tested using games of chance, such as dice and darts, that produce accidents. When an accident happens, do people learn best when punished based on the outcome, or instead based on their intent? The answer will determine the optimal punishment strategy, potentially explaining actual human punishment behavior. The second case study involves the distinction between harmful commissions and omissions. Harms of commission are typically punished more than harms of omission, as indicated both in psychological research and by criminal law. The proposed research tests whether this derives from constraints on learning; specifically, whether people learn from punishment of commissions more effectively than from punishment of omissions.