Fair Division Breakthrough for Mixed Goods and Chores Achieved
Researchers solve long-standing fairness problem when items can be good or bad.
A team of computer scientists led by Haris Aziz (UNSW) has published a paper on arXiv tackling a classic fair-division problem with a twist: items can be a good for some agents and a chore for others simultaneously. Their work introduces a 'best-of-both-worlds' framework, where a randomized allocation is exactly fair ex ante (before randomization) while being supported on allocations that are approximately fair ex post (after randomization). The specific fairness notions are envy-freeness (EF) and envy-freeness up to one item (EF1), a well-known relaxation. The main result proves that ex-ante EF and ex-post EF1 can be achieved together for all additive utilities.
The technical contribution is a novel probabilistic Hall-type matrix decomposition that correlates the fractional assignments of goods and chores. The authors resolve this decomposition by combining continuous minimax duality (via Sion's minimax theorem) with carefully designed biased flow networks. This approach is a significant step in algorithmic game theory, as it extends fairness guarantees to scenarios where an item's value can be negative for some participants—a common real-world situation (e.g., work tasks or undesirable assets). The algorithm has potential applications in resource allocation, scheduling, and multi-agent systems where fairness constraints are critical.
- Simultaneously achieves ex-ante envy-freeness (EF) and ex-post EF1 for mixed goods and chores.
- Introduces a probabilistic Hall-type matrix decomposition using Sion's minimax theorem and biased flow networks.
- Extends fair division theory to handle items that are simultaneously good and bad for different agents.
Why It Matters
Enables provably fair allocation of resources with mixed valuations, even when items are undesirable to some.