Incentivizing Honesty among Competitors in Collaborative Learning and Optimization

Collaborative learning techniques have the potential to enable training machine learning models that are superior to models trained on a single entity's data. However, in many cases, potential participants in such collaborative schemes are competitors on a downstream task, such as firms that each aim to attract customers by providing the best recommendations. This can incentivize dishonest updates that damage other participants' models, potentially undermining the benefits of collaboration. In this work, we formulate a game that models such interactions and study two learning tasks within this framework: single-round mean estimation and multi-round SGD on strongly convex objectives. For a natural class of player actions, we show that rational clients are incentivized to strongly manipulate their updates, preventing learning. We then propose mechanisms that incentivize honest communication and ensure learning quality comparable to full cooperation. Lastly, we empirically demonstrate the effectiveness of our incentive scheme on a standard non-convex federated learning benchmark. Our work shows that explicitly modeling the incentives and actions of dishonest clients, rather than assuming them malicious, can enable strong robustness guarantees for collaborative learning. Clients attacking the training are typically modeled as Byzantine (Blanchard et al., 2017; Yin et al., 2018; Alistarh et al., 2018) , adversarially seeking to sabotage training by deviating from the FL protocol in a worst-case manner. The goal of Byzantine-robust learning is to achieve guarantees in that setting. Similar models have been studied in a statistical context, where data is stored at a single location and so communication is not a concern (Qiao & Valiant, 2018; Konstantinov et al., 2020) . In contrast to these works, we model manipulation as a consequence of competitive incentives rather than maliciousness and analyze client behaviour using game theory (Osborne & Rubinstein, 1994) . Rather than focusing on robustness to manipulations, we aim to prevent manipulation altogether. Peer prediction mechanisms Our mechanisms for inducing honesty are closely related to peer prediction that aims to incentivize honest ratings on online platforms. In their seminal paper,