Castiglioni et al. introduce EFS contracts beating fairness price bounds
New algorithmic scheme uses agent-specific subsidies to achieve strict envy-freeness with bounded price of fairness.
A team of computer scientists (Castiglioni, Chen, Li) has introduced Envy-free Contracts with Subsidies (EFS) to tackle the challenge of fairly delegating tasks in multi-agent systems. Traditional envy-free (EF) contracts either suffer from an unbounded price of fairness (PoF) or sacrifice strict fairness. EFS relaxes the model by allowing the principal to offer agent-specific subsidies alongside contracts. The authors prove that EFS contracts restore strict envy-freeness and can outperform standard EF contracts by an arbitrarily large factor. Moreover, they show that EFS admits a tight PoF bound of n^{Θ(n)}—exponential but finite—where n is the number of agents. This is a significant theoretical improvement over unbounded PoF.
On the computational side, the paper establishes that finding the optimal EFS contract is NP-hard in the general case. However, when the number of tasks is fixed (a constant), the researchers provide a polynomial-time algorithm to compute optimal contracts. This suggests practical feasibility for many real-world delegation scenarios with a limited task set. The work advances algorithmic fair contract design, offering a rigorous framework that balances fairness and efficiency. The paper is available on arXiv (2606.25431).
- EFS achieves strict envy-freeness using agent-specific subsidies, overcoming the unbounded price of fairness in previous EF contracts.
- The price of fairness for EFS is bounded by n^{Θ(n)}, a tight exponential bound in the number of agents.
- Optimal EFS contract computation is NP-hard generally, but tractable in polynomial time when the number of tasks is constant.
Why It Matters
This framework enables fair and efficient task delegation in multi-agent systems with guaranteed performance bounds.