Performative Personalization Incentivizes Truthfulness in Federated Learning

Kumar Kshitij Patel, Aniket Murhekar

Keywords: Personalization, Bayesian Incentive Compatibility, Federated Learning, Collaborative Learning

TL;DR: Personalization and leave-one-out procedures incentivize truthfulness in collaborations

Abstract: We study collaborative learning with strategic clients who may misreport oracle outputs to steer the learned model. Under simultaneous realizability and leave-one-out identifiability, we show that a leave-one-out consensus mechanism prevents harmful unilateral misreports and identifies the deviator. We also propose a one-shot alternative that uses a minimal cluster-recovering personalization oracle and preserves incentive compatibility without an $M$-fold increase in computation.

Track: Short Paper

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 106