- The paper demonstrates that the MNW rule and its equivalents uniquely satisfy the axioms of EF1, strategyproofness, neutrality, minimal completeness, and IDU in allocating indivisible goods with binary valuations.
- It shows that under binary settings, utilitarian, leximin, and strictly concave additive welfarist rules converge, enabling efficient, strategyproof, and resource-monotone implementations.
- For the two-agent case, enhanced characterizations using non-redundancy and resource-monotonicity confirm tie-breaking consistency and underscore the necessity of each axiom.
Characterizations of Fair Division with Binary Valuations
Problem Setting and Key Concepts
The paper investigates the axiomatic characterization of allocation rules—specifically, the maximum Nash welfare (MNW) rule, leximin rule, and strictly concave additive welfarist rules—for the fair division of indivisible goods among agents with binary additive valuations. In this context, each agent either approves or disapproves each good (assigning a value of 1 or 0, respectively), and utilities are additive across bundles. The study aims to determine the minimal and necessary set of axioms that uniquely identify MNW (and its equivalents) as the allocation rule in this restricted valuation domain.
Key desiderata for fair division rules in this domain include:
- Envy-freeness up to one good (EF1): Any envy an agent holds toward another can be eliminated by removing a single good from the latter's bundle.
- Strategyproofness: Truthful reporting of preferences is a dominant strategy for all agents.
- Neutrality: Allocative outcomes are invariant under relabeling of goods.
- Minimal completeness: All valued goods are allocated, and unvalued goods are left unallocated.
- Invariance under disapproving unassigned goods (IDU): If an agent changes her valuation for a good she does not receive, the outcome should remain unchanged.
- Non-redundancy (in two-agent settings): Every allocated good is assigned to an agent who values it.
- Resource-monotonicity (in two-agent settings): Adding goods cannot reduce any agent's utility.
Under the binary valuation setting, the MNW and leximin rules, as well as all strictly concave additive welfarist rules, coincide—maximizing the number of agents with strictly positive utility and, subject to that, maximizing the product or lexicographically maximizing the vector of agents' utilities.
Main Characterization for Arbitrary Number of Agents
The core result establishes that MNW (and equivalent rules) is uniquely characterized, for any number of agents, as the sole allocation rule satisfying EF1, strategyproofness, neutrality, minimal completeness, and IDU. The proof demonstrates that every one of these axioms is necessary—removing any single axiom admits allocation rules that are not MNW.
Furthermore, the equivalence of welfarist rules under binary valuations is exploited: utilitarian, leximin, and strictly concave additive welfarist rules (such as p-means for p<1) collapse to the same outcome. Notably, within this domain, many of the drawbacks of MNW for general additive valuations—such as computational intractability, lack of strategyproofness, and violation of resource-monotonicity—are eliminated: the rule becomes strategyproof, resource-monotone, and admits a polynomial-time implementation.
The proof hinges on ruling out allocations that admit “critical paths" (as defined by Halpern et al.), showing that any PO and minimally complete rule violating MNW must enable some agent to benefit via manipulation or violate EF1, neutrality, or IDU.
Two-Agent Specialization and Alternative Characterization
For two agents, the paper provides a strictly stronger characterization. It is shown that MNW is the unique allocation rule satisfying EF1, strategyproofness, neutrality, minimal completeness, non-redundancy, and resource-monotonicity. IDU is replaced by the combination of non-redundancy and resource-monotonicity. The proof uses an inductive argument on the number of goods and leverages a symmetry argument via characteristic tuples (summarizing the profile by the number of uniquely and commonly approved goods).
A salient consequence is a tie-breaking constraint: consistency in tie-breaking is required across all instances with the same number of valued goods, regardless of the total number of goods, when there are multiple MNW optima (this occurs, e.g., when the total number of valued goods is odd and can be split in two different ways).
Independence and Necessity of Axioms
The independence of the axioms is rigorously established: in both the general (any n) and two-agent settings, for any one axiom omitted, there exists an allocation rule that satisfies all the others but does not select MNW. This is substantiated via constructed counterexamples—the appendix provides full details and verifies all cases.
Implications
The characterizations provided here are significant for both practical and theoretical reasons:
Practical Implications
- Auditability and transparency: The unique axiomatic characterization clarifies which fairness and strategyproofness assurances are provided by MNW-type rules in approval (binary) settings.
- Mechanism selection: In computational social choice applications where agents supply approval ballots (e.g., participatory budgeting, resource allocation in networks), this result pinpoints the circumstances under which MNW (and only MNW) is justified.
- Tie-breaking discipline: For implementation and system design, the careful analysis of tie-breaking offers operational guidance on agent-favoring biases and the importance of consistency in allocations.
Theoretical Implications
- Contrast with the additive domain: While the additive setting admits many rules satisfying EF1 and PO, these stronger characterizations demonstrably “pin down” MNW-only in the binary case.
- Axiomatic landscape: This paper fills a notable gap—characterizing one of the most important fair division rules through necessary and sufficient conditions in a non-trivial domain.
- Resource-monotonicity and non-redundancy: The alternative axiom set for two agents emphasizes the intricate relationship between efficiency, fairness, and monotonicity properties—suggesting boundaries for what is possible in other domains (e.g., matroid rank, submodular, or general additive valuations).
Future Directions
The authors note that extending this line of work to domains beyond binary additive valuations—especially to full additive domains or more complex combinatorial structures—remains challenging. In particular, many of the positive properties (e.g., strategyproofness and resource-monotonicity) fail for MNW under general additive valuations. Another suggested direction is the axiomatic characterization of other fair division protocols, such as round-robin or envy-cycle-elimination algorithms.
Conclusion
This work delivers a comprehensive axiomatic characterization of the MNW (and equivalent) rule(s) for fair division with binary valuations, establishing the minimal and necessary axioms for unique selection in this domain. The results delineate the boundaries of fairness, efficiency, incentive compatibility, and monotonicity, providing both justification and constraint for the deployment of MNW-type allocation rules in approval-based division scenarios. The analytical tools developed here lay the groundwork for further explorations of the axiomatic foundations of fair division in more general and complex environments.