Fair Orientations: Proportionality and Equitability

A new research paper titled 'Fair Orientations: Proportionality and Equitability' by Ankang Sun, Ruijie Wang, and Bo Li has been accepted to the 2026 International Conference on Autonomous Agents and Multiagent Systems (AAMAS). The work addresses a critical gap in algorithmic fairness by extending the study of resource allocation under relevance constraints beyond the commonly studied envy-freeness criterion.

The paper systematically analyzes two other fundamental fairness notions: proportionality (ensuring each agent gets their fair share) and equitability (ensuring agents have equal satisfaction levels). This analysis covers scenarios where items can be goods (desirable), chores (undesirable), or a mixture of both, which mirrors real-world allocation problems more closely than models assuming all items are goods. The researchers provide a comprehensive map of when fair allocations exist under these criteria and establish the computational complexity (whether solutions can be found efficiently) for finding them, filling a significant void in the theoretical computer science and game theory literature.

This research provides essential theoretical foundations for designing fair AI systems, particularly multi-agent systems and automated decision-makers that allocate limited resources—such as computational tasks, data slices, or cloud services—among users with different needs and preferences. By clarifying which fairness goals are computationally achievable under realistic constraints, it guides engineers and policymakers toward implementable and provably fair algorithms.