Fair Division of Indivisible Chores via Earning Restricted Equilibria

We study fair division of $m$ indivisible chores among $n$ agents with additive preferences. We consider the desirable fairness notions of envy-freeness up to any chore (EFX) and envy-freeness up to $k$ chores (EF$k$), alongside the efficiency notion of Pareto optimality (PO). We present the first constant approximations of these notions, showing the existence of: - 5-EFX allocations, which improves the best-known factor of $O(n^2)$-EFX. - 3-EFX and PO allocations for the special case of bivalued instances, which improves the best-known factor of $O(n)$-EFX without any efficiency guarantees. - 2-EF2 + PO allocations, which improves the best-known factor of EF$m$ and PO. A notable contribution of our work is the introduction of the novel concept of earning-restricted (ER) competitive equilibrium for fractional allocations, which limits agents' earnings from each chore. Technically, our work addresses two main challenges: proving the existence of an ER equilibrium and designing algorithms that leverage ER equilibria to achieve the above results. To tackle the first challenge, we formulate a linear complementarity problem (LCP) formulation that captures all ER equilibria and show that the classic complementary pivot algorithm on the LCP must terminate at an ER equilibrium. For the second challenge, we carefully set the earning limits and use properties of ER equilibria to design sophisticated procedures that involve swapping and merging bundles to meet the desired fairness and efficiency criteria. We expect that the concept of ER equilibrium will be instrumental in deriving further results on related problems.