online marketplaces

Abstract: A random assignment of n indivisible objects to n agents is specified by its assignment matrix and implemented by drawing a deterministic assignment from a Birkhoff-von Neumann decomposition. Kawase et al. observed that the choice of decomposition matters for fairness: a matrix that is envy-free in the sense of stochastic dominance (SD-EF) can be decomposed so that some agent envies another with probability close to 1. They call a decomposition Dec-EF if every agent envies every other agent with probability at most 1/2, proved that every SD-EF matrix admits a Dec-EF decomposition when n <= 3 or when there are at most two distinct preferences, and left the general case open. We settle the first open case: every SD-EF matrix with four agents admits a Dec-EF decomposition. The worst case over the SD-EF polytope of a profile is attained at a vertex, and our computer-aided proof enumerates all 26,927 vertices for the 762 profiles up to symmetry in exact arithmetic and certifies each by a rational decomposition. The same method settles five agents with at most four distinct preferences and the probabilistic serial rule for all five-agent profiles, and adversarial search up to seven agents finds no counterexample. For general n, an envy-budget identity shows that 1/2 is the best possible threshold. We prove that every SD-EF matrix with at most two distinct rows admits a Dec-EF decomposition, and that the maximum-entropy decomposition is Dec-EF whenever all agents but two share a preference; the latter proof rests on a new monotonicity lemma for weighted least-squares rankings. In general, natural decompositions fail: greedy Birkhoff-von Neumann can come arbitrarily close to envy probability (n-1)/n, and maximum entropy fails at n = 4 when all preferences differ. Deciding whether an arbitrary random assignment, not necessarily SD-EF, admits a Dec-EF decomposition is strongly NP-complete.

Abstract: In the classical i.i.d. prophet-inequality problem, a single item is allocated to one of $n$ agents who arrive sequentially, with values drawn independently from a known distribution. When an agent arrives, their value is revealed, and the algorithm must immediately allocate the item or continue. The usual objective is utilitarian welfare: the expected value of the recipient. Guarantees for this objective extend to allocating $m$ indivisible items to sequentially arriving agents with i.i.d.\ additive values. Utilitarian welfare, however, ignores how expected utility is distributed across agents. Motivated by a rich literature in fair division, we instead evaluate an online rule by its generalized $p$-mean welfare, which includes utilitarian welfare at $p=1$, Nash welfare (the geometric mean of utilities) at $p=0$, and egalitarian welfare (the minimum utility) as $p\to-\infty$. When the number of items is large, we show that this many-item fair-division problem is captured exactly by a single-item prophet problem evaluated by the $p$-mean of agents' expected utilities. We characterize this single-item problem: every online rule is weakly Pareto dominated by a quantile-threshold rule, and an optimal egalitarian rule equalizes agents' expected utilities. We prove that for every $n$, the online optimum is at least $Γ\approx0.7059$ times the prophet's egalitarian welfare; by monotonicity of generalized means, the same guarantee holds for every $p\le1$. Further, for egalitarian welfare, the optimal ratio converges to $Γ$ as $n\to\infty$. Thus, asymptotically, optimizing egalitarian rather than utilitarian welfare costs only about four percentage points relative to the classical utilitarian ratio of $0.7451$. Finally, when $m=n$, the worst-case competitive ratio converges to zero as $n\to\infty$ for every $p\le0$, showing that the large-item assumption is necessary.

Abstract: Exploration in centralized serial-dictatorship matching bandits must use complete matchings, so learning one player--arm pair can impose regret on others. We study this externality under a known common priority order and Gaussian rewards with unit variance. We show that the matching-level Graves--Lai constraints reduce to finitely many pairwise exploration quotas and, at top-choice-separated instances, yield a polynomial-size marginal linear program. At these instances, the exact attainable set of expected logarithmic regret coefficients is $G(θ)\Xset(θ)$, where $\Xset$ is the feasible matching-allocation set and $G$ maps allocations to player regret. The usual upper-closed Graves--Lai region can be strictly larger despite having the same Pareto-minimal boundary. We further show that identical exploration quotas can induce very different regret through their scheduling. Finally, we construct estimate--solve--track policies, uniformly good on the full row-strict class, that attain every fixed positively weighted optimum without assuming optimizer uniqueness. Every Pareto-minimal point is pointwise attainable, possibly through an instance-calibrated target.