Optimal feedback dynamics against free-riding in collective experimentation

We consider a dynamic model in which a principal decides what information to release about a product of unknown quality (e.g., a vaccine) to incentivize agents to experiment with the product. Assuming that the agents are long-lived and forward-looking, their incentive to wait and see other agents' experiences poses a significant obstacle to social learning. We show that the optimal feedback mechanism to mitigate information free-riding takes a strikingly simple form: the principal recommends adoption as long as she observes no bad news, but only with some probability; once she does not recommend at some point, she stays silent forever after that. Our analysis suggests the optimality of premature termination, which in turn implies that: (i) false positives (termination in the good state) are more acceptable than false negatives (continuation in the bad state); (ii) overly cautious mechanisms that are biased toward termination can be welfare-enhancing.