Long-term Fairness For Real-time Decision Making: A Constrained Online Optimization Approach

Machine learning (ML) has demonstrated remarkable capabilities across manyreal-world systems, from predictive modeling to intelligent automation.However, the widespread integration of machine learning also makes it necessaryto ensure machine learning-driven decision-making systems do not violateethical principles and values of society in which they operate. As ML-drivendecisions proliferate, particularly in cases involving sensitive attributessuch as gender, race, and age, to name a few, the need for equity andimpartiality has emerged as a fundamental concern. In situations demandingreal-time decision-making, fairness objectives become more nuanced and complex:instantaneous fairness to ensure equity in every time slot, and long-termfairness to ensure fairness over a period of time. There is a growing awarenessthat real-world systems that operate over long periods and require fairnessover different timelines. However, existing approaches mainly address dynamiccosts with time-invariant fairness constraints, often disregarding thechallenges posed by time-varying fairness constraints. To bridge this gap, thiswork introduces a framework for ensuring long-term fairness within dynamicdecision-making systems characterized by time-varying fairness constraints. Weformulate the decision problem with fairness constraints over a period as aconstrained online optimization problem. A novel online algorithm, namedLoTFair, is presented that solves the problem 'on the fly'. We prove thatLoTFair can make overall fairness violations negligible while maintaining theperformance over the long run.