Remember, but also, Forget: Bridging Myopic and Perfect Recall Fairness with Past-Discounting

2025-04-01

AI Summary

In multi-agent dynamic resource allocation, conventional fairness metrics suffer from myopia, failing to jointly optimize short-term efficiency and long-term fairness. To address this, we propose a time-discounted fairness framework that aggregates historical utilities via adjustable exponential decay—either additively or averagely—enabling continuous interpolation between instantaneous and full-horizon fairness. This work pioneers the integration of behavioral temporal awareness into algorithmic fairness, yielding a tunable, bounded, and computationally tractable fairness spectrum. We embed this framework into sequential decision-making and fairness-constrained reinforcement learning. Experiments across multiple canonical multi-agent settings demonstrate that our approach significantly outperforms both instantaneous-fairness and full-memory-fairness baselines, achieving superior joint optimization of long-term fairness and short-term efficiency.

Technology Category

• Multiagent Systems: Mechanism Design
• Game Theory and Economic Paradigms: Fair Division
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Dynamic resource allocation in multi-agent settings often requires balancing efficiency with fairness over time--a challenge inadequately addressed by conventional, myopic fairness measures. Motivated by behavioral insights that human judgments of fairness evolve with temporal distance, we introduce a novel framework for temporal fairness that incorporates past-discounting mechanisms. By applying a tunable discount factor to historical utilities, our approach interpolates between instantaneous and perfect-recall fairness, thereby capturing both immediate outcomes and long-term equity considerations. Beyond aligning more closely with human perceptions of fairness, this past-discounting method ensures that the augmented state space remains bounded, significantly improving computational tractability in sequential decision-making settings. We detail the formulation of discounted-recall fairness in both additive and averaged utility contexts, illustrate its benefits through practical examples, and discuss its implications for designing balanced, scalable resource allocation strategies.

Problem

• Balancing efficiency and fairness in dynamic resource allocation
• Introducing temporal fairness with past-discounting mechanisms
• Improving computational tractability in sequential decision-making

Innovation

• Introduces past-discounting for temporal fairness
• Balances immediate and long-term equity considerations
• Ensures bounded state space for tractability