Unravelling in Collaborative Learning

TL;DR

The paper addresses adverse selection in collaborative learning when data-quality indices are privately known, risking unravelling where the coalition collapses to the lowest-quality contributor. It models this as a principal-agent game with sampling costs and a data-quality divergence measure, derives a welfare-maximizing scheme under full information, and shows that naive private-information mechanisms trigger unravelling. The main contribution is a transfer-free mechanism based on probabilistic verification that uses estimated agent types to ensure the grand coalition forms a Nash equilibrium with high probability, effectively counteracting unravelling; the authors also discuss practical implementation details in classification tasks. This work has practical significance for stabilizing large-scale collaborative learning systems by leveraging mechanism design to align incentives under information asymmetry, potentially reducing reliance on external transfers.

Abstract

Collaborative learning offers a promising avenue for leveraging decentralized data. However, collaboration in groups of strategic learners is not a given. In this work, we consider strategic agents who wish to train a model together but have sampling distributions of different quality. The collaboration is organized by a benevolent aggregator who gathers samples so as to maximize total welfare, but is unaware of data quality. This setting allows us to shed light on the deleterious effect of adverse selection in collaborative learning. More precisely, we demonstrate that when data quality indices are private, the coalition may undergo a phenomenon known as unravelling, wherein it shrinks up to the point that it becomes empty or solely comprised of the worst agent. We show how this issue can be addressed without making use of external transfers, by proposing a novel method inspired by probabilistic verification. This approach makes the grand coalition a Nash equilibrium with high probability despite information asymmetry, thereby breaking unravelling.