Learning and experimentation in a "principal-agent" framework

The purpose of this work is to introduce learning and experimenting in a principal-agent model, where the agent has some private information which is valuable for the principal. Intuitively this should lead the principal to experiment in order to improve his knowledge. For this to be possible the information must be partially revealed by some observable output correlated with the agent's action; furthermore, the principal has to have some way to influence the agent's choices, or else he would not be able to experiment but only be a passive learner.With these characteristics, experimenting in such a principal-agent model should not be very different from experimenting a new consumer good. The innovative point arises from allowing the agent to react to experimentation. This should occur the agent being rational enough and having an incentive to keep his information private. As will be seen the most rational agent will try to blur the experiments carried by the principal and in some situations the equilibrium level of learning might be significantly impaired.A simple example of a principal-agent model with experimentation will be formulated and analyzed thoroughly. Several interesting results will be shown and the optimal behavior of both principal and agent will be derived and explained.A more general model is then introduced requiring the use of more sophisticated mathematical tools to prove the validity of the previous results. However, in some cases, this could not be attained.U of I OnlyETDs are only available to UIUC Users without author permissio