Test-Time Collective Action: Proxy-Based Perturbations for Correcting Algorithmic Harms
For users of machine learning systems who face subgroup disparities, TTCA provides a corrective intervention without requiring platform cooperation, addressing a practical limitation of existing provider-centric fairness approaches.
The paper introduces Test-Time Collective Action (TTCA), a framework enabling users to correct algorithmic disparities for underserved subgroups by pooling query access to a black-box API, extracting a proxy model, and applying per-class universal perturbations at test time. On CIFAR-10, CIFAR-100, and FairFace, modestly-sized collectives close most of the subgroup accuracy gap, improve worst-group accuracy, equal-opportunity gap, and disparate impact, with pooling being cheaper than individual attacks.
When machine learning systems under-perform for particular subgroups, affected users typically have no way to correct these disparities without relying on platform-level fixes. Existing approaches to algorithmic fairness rely on provider-centric approaches to correct these failures, leaving users with no external lever when faced with harm. Recent work in Algorithmic Collective Action shows that coordinated users can steer an algorithmic system toward a collective goal, but the existing mechanisms require the provider to retrain on the collective's modified data which users may not have control over. We propose Test-Time Collective Action (TTCA), a framework through which a group of users who share query access to the platform, can correct disparities affecting under-served subgroup without participating in the platform's training loop. We implement this through a proxy-based mechanism where the collective pools query access to a black-box API to extract a proxy of the platform, then optimizes a per-class universal perturbation against the proxy. Each member applies this perturbation to their own inputs at submission time, requiring no cooperation from the platform. We empirically evaluate the mechanism on CIFAR-10, CIFAR-100, and FairFace, showing that modestly-sized collectives close most of the subgroup accuracy gap, transfer across architectures (a small proxy can attack a larger platform), and improve worst-group accuracy, equal-opportunity gap, and disparate impact. A query-budget analysis comparing a per-user black-box attack baseline shows that pooling is cheaper than each subgroup member attacking alone. Test-time collective action thus offers corrective intervention to users when platform-side remediation is unavailable or delayed.