Addressing Polarization and Unfairness in Performative Prediction
This addresses fairness issues in performative prediction for applications like recommendations, hiring, and lending, offering a novel approach to mitigate polarization and disparities.
The paper tackles the problem of polarization and unfairness in performative prediction, where models influence their own training data, showing that performative stable solutions can cause severe disparities and that conventional fairness interventions often fail, and introduces novel fairness mechanisms that provably ensure both stability and fairness.
In many real-world applications of machine learning such as recommendations, hiring, and lending, deployed models influence the data they are trained on, leading to feedback loops between predictions and data distribution. The performative prediction (PP) framework captures this phenomenon by modeling the data distribution as a function of the deployed model. While prior work has focused on finding performative stable (PS) solutions for robustness, their societal impacts, particularly regarding fairness, remain underexplored. We show that PS solutions can lead to severe polarization and prediction performance disparities, and that conventional fairness interventions in previous works often fail under model-dependent distribution shifts due to failing the PS criteria. To address these challenges in PP, we introduce novel fairness mechanisms that provably ensure both stability and fairness, validated by theoretical analysis and empirical results.