Fairness for the People, by the People: Minority Collective Action

2025-08-21

AI Summary

To address inherent biases against minority groups in deployed machine learning models, this paper proposes a user-side, decentralized fairness optimization framework: without requiring enterprise involvement or modifications to the training pipeline, minority-group users collectively intervene by strategically relabeling their own input instances. The approach is model-agnostic, compatible with black-box models, and minimally invasive—marking the first effort to empower end users with agency over fairness improvement. We design three approximately optimal relabeling algorithms and evaluate them across multiple real-world datasets. Results show that adjusting labels for only a small fraction of samples significantly reduces inter-group predictive unfairness—e.g., decreasing equal opportunity difference by up to 40%—while preserving overall predictive accuracy, with classification error increasing by less than 0.5%.

Technology Category

• Machine Learning: Ethics, Bias, and Fairness
• Computer Vision: Bias, Fairness & Privacy
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Machine learning models often preserve biases present in training data, leading to unfair treatment of certain minority groups. Despite an array of existing firm-side bias mitigation techniques, they typically incur utility costs and require organizational buy-in. Recognizing that many models rely on user-contributed data, end-users can induce fairness through the framework of Algorithmic Collective Action, where a coordinated minority group strategically relabels its own data to enhance fairness, without altering the firm's training process. We propose three practical, model-agnostic methods to approximate ideal relabeling and validate them on real-world datasets. Our findings show that a subgroup of the minority can substantially reduce unfairness with a small impact on the overall prediction error.

Problem

• Mitigating unfair treatment of minority groups in machine learning models
• Enabling end-users to induce fairness through strategic data relabeling
• Reducing unfairness with minimal impact on overall prediction error

Innovation

• User-contributed data relabeling for fairness
• Model-agnostic methods for bias mitigation
• Minority collective action without process alteration