New algorithm ensures fairness in real-time resource allocation

A team of computer scientists (Christopher En, Yuri Faenza, Andrea Lodi, Gonzalo Muñoz) has tackled the challenge of fair online resource allocation—a problem with applications in refugee resettlement, airline scheduling, and more. Their model incorporates a Lipschitz fairness constraint, requiring that agents arriving in the same batch with similar characteristics receive similar expected outcomes. In the offline setting, they prove that the optimal fair allocation retains at least an Ω(1/γ) fraction of the value of an unfair allocation, where γ is the fairness coefficient—quantifying the price of fairness.

For the online setting, where agents arrive sequentially and resource capacities are limited, the researchers propose an algorithm based on dual mirror descent. This method enforces fairness constraints within each batch while dynamically estimating optimal dual variables, achieving sublinear regret relative to the optimal offline fluid benchmark. The algorithm was validated using real data from the Refugee Economies Programme, demonstrating practical trade-offs between welfare maximization and fairness enforcement. The results, to appear at EC 2026, offer a rigorous framework for balancing efficiency and equity in high-stakes allocation problems.