Title: Beyond Independent Manipulation: Individual Fairness-aware Strategic Classification with Peer Imitation

Abstract: Strategic classification (SC) examines situations in which agents alter their characteristics to secure advantageous predictions from algorithmic models. While current approaches to fairness-aware SC largely prioritize group-level equity and presume that agents act autonomously, the requirement for individual fairness—ensuring comparable individuals receive comparable results—introduces a new dynamic. Under individual fairness, manipulation strategies become interdependent; an agent’s optimal course of action is contingent upon the outcomes experienced by those in their immediate vicinity. This interdependence creates a disconnect between traditional SC models and fairness-centric decision frameworks, as independent behavioral assumptions fail to capture the reality of strategic manipulation.

To resolve this discrepancy, we propose Individual Fairness-aware Strategic Classification (IFSC), a novel framework that accounts for peer-driven manipulation rooted in individual fairness principles. In this model, agents seek favorable outcomes by imitating nearby peers who have already received positive decisions. IFSC defines strategic behavior as imitation based on similarity toward visible, accepted peers and trains classifiers based on the resulting post-manipulation data distributions. To handle the uncertainty inherent in observing peers, IFSC utilizes a robust learning methodology that incorporates stochastic perturbations during the simulation of manipulation processes. Our experiments, conducted on both synthetic and real-world datasets, show that IFSC enhances consistency in individual fairness and reduces distortions caused by imitation.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC