| dc.description |
Punishment is an important method for discouraging uncooperative behavior. This work studies the information used when deciding to apply a punishment, and what punishment to apply. We use a novel design for a public goods game in which a player’s actual contribution is a random deviation from their intended contribution, and both the intended and actual contributions are explicitly displayed to all players. This feature lets players detect accidental free riding or accidental high contributing. Multiple types of punishment are studied, including fines, ostracism, and reputation marking. We investigate the effect of a punishment’s efficacy for changing behavior on the continued use of the punishment. We investigate the effect of local norms of punishment. We also investigate the effect of the cost of applying a punishment. Our novel design with automated players allows complete experimental control and thus provides the capability to manipulate these factors directly. Bayesian hierarchical models are used for data analysis. Contrary to some pre-existing literature, punishment decisions are found to be flexible, to be responsive to changing conditions, and to emphasize outcomes over intentions only in specific, narrow circumstances. Moreover, we find that the rarely studied punishments of ostracism and reputation marking are quite different from the more often studied fine in how they are utilized, and thus these and other alternative punishments are essential to study in the future. |
|