How to design incentive systems based on the multiple signals from agent's behavior

摘要

Agency theory has been often employed to discuss the incentive problems. Holmstrom noted that if there is a signal from the agent's behavior, the principal had better design IS considering this signal. In this paper, a model including intrinsicmotivation which is given little considerations in the former models is developed. From the analysis of it, following results are obtained, that is, 1. When the agent is highly intrinsically motivated, the principal has to take not only the accuratesignal but also the noisy signal into account in evaluating the agent's behavior. 2. When the agent is not intrinsically motivated, the principal has to evaluate the agent's behavior based on only the signal which has few noise.