Learning by doing vs. learning from others in a principal-agent model

We introduce learning in a principal-agent model of output sharing under moral hazard. We use social evolutionary learning to represent social learning and reinforcement, experience-weighted attraction (EWA) and individual evolutionary learning (IEL) to represent individual learning. Learning in the principal-agent model is difficult due to: the stochastic environment; the discontinuity in payoffs at the optimal contract; and the incorrect evaluation of foregone payoffs for IEL and EWA. Social learning is much more successful in adapting to the optimal contract than standard individual learning algorithms. A modified IEL using realized payoffs evaluation performs better but still falls short of social learning.