Robust Strategic Classification under Decision-Dependent Cost Uncertainty
2026-06-29 • Machine Learning
Machine LearningComputer Science and Game Theory
AI summaryⓘ
The authors study how people try to 'game' algorithms by changing their input to get better results, which can be costly for both sides. Unlike earlier work that assumed these costs stay the same regardless of decisions, the authors show that costs actually depend on past algorithm choices and change over time. They propose a new two-stage model that accounts for this evolving cost and uncertainty. Their approach helps reduce the chances of people exploiting the system repeatedly by considering how today's decisions affect future manipulation costs.
strategic classificationalgorithmic gamingmanipulation costrobust optimizationdecision-dependent uncertaintytwo-stage optimizationmachine learningadversarial behavioralgorithmic decision systemspolicy-dependence
Authors
Sura Alhanouti, Güzin Bayraksan, Parinaz Naghizadeh
Abstract
Humans facing algorithmic decision systems have been found to ``game'' them by altering their input data (at a cost to them) in order to favorably change the algorithmic outcomes they receive (at a cost to the algorithm). The growing literature on strategic classification seeks to develop robust machine learning algorithms that account for, and reduce, unwanted strategic behavior. A limitation of these existing works is that they assume the cost of strategic behavior to be fixed and independent of the classifier's decision. In practice, however, manipulation costs evolve and depend on past algorithmic decisions: today's decisions influence tomorrow's costs. This paper proposes and analyzes a two-stage robust optimization framework with a decision-dependent uncertainty set to capture such dependencies. We highlight that awareness of policy-dependent costs not only reduces uncertainty, but also better curtails gaming of the algorithmic system over time.