Anticipating Gaming to Incentivize Improvement: Guiding Agents in (Fair) Strategic Classification

TL;DR

This paper models strategic human responses to machine learning classifiers as a game, analyzing how to design fair algorithms that incentivize genuine improvement over manipulation.

Contribution

It introduces a Stackelberg game framework to study strategic responses and proposes methods to design classifiers that promote genuine improvement and fairness.

Findings

Optimal classifiers can prevent manipulation and encourage improvement.

Different agent response classes are characterized based on costs and efficacy.

Fair strategic policies can align incentives towards genuine skill enhancement.

Abstract

As machine learning algorithms increasingly influence critical decision making in different application areas, understanding human strategic behavior in response to these systems becomes vital. We explore individuals' choice between genuinely improving their qualifications (``improvement'') vs. attempting to deceive the algorithm by manipulating their features (``manipulation'') in response to an algorithmic decision system. We further investigate an algorithm designer's ability to shape these strategic responses, and its fairness implications. Specifically, we formulate these interactions as a Stackelberg game, where a firm deploys a (fair) classifier, and individuals strategically respond. Our model incorporates both different costs and stochastic efficacy for manipulation and improvement. The analysis reveals different potential classes of agent responses, and characterizes optimal classifiers accordingly. Based on these, we identify when and why a (fair) strategic policy can not only prevent manipulation, but also incentivize agents to opt for improvement. Our findings shed light on the intertwined nature of microeconomic and ethical implications of firms' anticipation of strategic behavior when employing ML-driven decision systems.