Modeling Deletion Requests in Machine Unlearning

2026-10-03

AI Summary

This study investigates how adaptive and collective deletion requests differentially influence model behavior when users exercise the right to be forgotten. To this end, this work pioneers the incorporation of user behavior modeling into machine unlearning by formally defining the adaptivity and collectivity of deletion behaviors. Leveraging stochastic optimization theory, it analyzes their underlying mechanisms and proposes a proactive deletion strategy based on data valuation to substantially alter model behavior. Theoretically, this research elucidates the fundamental distinctions between individual and group deletions. Empirically, experiments on computer vision tasks validate that diverse user behavior patterns exert significantly differentiated impacts on model performance, effectively decoupling and isolating the independent effects of adaptivity and collectivity.

Abstract

Machine unlearning is seen as a promising approach to enable users to exercise the"right to erasure"in the context of AI models. We ask how users might influence the behavior of models when exercising this right. We define two types of behaviors that users might adopt when requesting the deletion of their data: adaptivity and collectivity. Drawing connections between the goals of users in this context and results in stochastic optimization, we demonstrate theoretical gaps between the potential effects of groups of users who do and do not display these behaviors. We then show how techniques from data valuation might be used to design deletion requesters that can significantly alter model behavior in realistic settings. In experiments on computer vision tasks, we demonstrate the differential effects of different models of user behavior and attempt to isolate the impacts of adaptivity and collectivity.

Problem

• Machine Unlearning
• Right to Erasure
• Deletion Requests
• Adaptivity
• Collectivity

Innovation

• Machine Unlearning
• Deletion Requests
• Data Valuation
• Adaptivity and Collectivity
• Stochastic Optimization