Incentive-Aware Machine Learning; Robustness, Fairness, Improvement & Causality
Chara Podimata
TL;DR
The paper surveys incentive-aware ML, addressing how strategic input manipulation by individuals affects policy goals across robustness, fairness, and improvement/causality. It synthesizes offline and online learning frameworks, causal models, and mechanisms for partial information, heterogeneous agents, and beyond-rational behavior. Key contributions include identifying fundamental results in strategic classification, designing robust and fair decision rules, and clarifying when adaptations yield genuine improvement versus gaming, along with performative and causal perspectives. The work highlights open challenges, such as agent heterogeneity, information asymmetry, and deploying theory-driven incentives in real-world domains, and proposes directions for linking theoretical insights to concrete application areas.
Abstract
The article explores the emerging domain of incentive-aware machine learning (ML), which focuses on algorithmic decision-making in contexts where individuals can strategically modify their inputs to influence outcomes. It categorizes the research into three perspectives: robustness, aiming to design models resilient to "gaming"; fairness, analyzing the societal impacts of such systems; and improvement/causality, recognizing situations where strategic actions lead to genuine personal or societal improvement. The paper introduces a unified framework encapsulating models for these perspectives, including offline, online, and causal settings, and highlights key challenges such as differentiating between gaming and improvement and addressing heterogeneity among agents. By synthesizing findings from diverse works, we outline theoretical advancements and practical solutions for robust, fair, and causally-informed incentive-aware ML systems.