Incentivizing Exploration with Selective Data Disclosure

Abstract

We propose and design recommendation systems that incentivize efficient exploration. Agents arrive sequentially, choose actions and receive rewards, drawn from fixed but unknown action-specific distributions. The recommendation system presents each agent with actions and rewards from a subsequence of past agents, chosen ex ante. Thus, the agents engage in sequential social learning, moderated by these subsequences. We asymptotically attain optimal regret rate for exploration, using a flexible frequentist behavioral model and mitigating rationality and commitment assumptions inherent in prior work. We suggest three components of effective recommendation systems: independent focus groups, group aggregators, and interlaced information structures.

The ACM-EC 2020 conference publication corresponds to the Feb'20 version. Section 7 ("robustness") and Section 8 (the numerical study) were added in, resp., Dec'20 and Nov'24. New discussions (Section 3.2.1 and Appendix B) were added in April'26, as well as a partial reframing of the motivating story to emphasize transparency and deemphasize commitment