Abstract:We study functions that produce a ranking of $n$ individuals from $n$ such rankings and are impartial in the sense that the position of an individual in the output ranking does not depend on the input ranking submitted by that individual. When $n \geq 4$, two properties concerning the quality of the output in relation to the input can be achieved in addition to impartiality: individual full rank, which requires that each individual can appear in any position of the output ranking; and monotonicity, which requires that an individual cannot move down in the output ranking if it moves up in an input ranking. When $n \geq 5$, monotonicity can be dropped to strengthen individual full rank to weak unanimity, requiring that a ranking submitted by every individual must be chosen as the output ranking. Mechanisms achieving these results can be implemented in polynomial time. Both results are best possible in terms of their dependence on $n$. The second result cannot be strengthened further to a notion of unanimity that requires agreement on pairwise comparisons to be preserved.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to design a ranking aggregation mechanism that can preserve as many other desirable properties as possible while satisfying impartiality, such as individual full rank, weak unanimity, and monotonicity. Specifically: 1. **Impartiality**: It means that an agent's position in the output ranking does not depend on its own submitted input ranking. This prevents the agent from influencing the final result by manipulating its own input ranking. 2. **Individual Full Rank**: Each agent can appear in any position in the output ranking. This implies that no agent is fixed in a particular position. 3. **Weak Unanimity**: When all agents agree on a certain ranking, this ranking will be selected as the output ranking. This is a weaker consistency requirement, which does not require agreement on all pairwise comparisons. 4. **Monotonicity**: If an agent rises in a certain input ranking, and other input rankings remain unchanged, this agent should not decline in the output ranking. The main contributions of the paper include: - Proving that for the case of $n \geq 4$, there exists a mechanism that simultaneously satisfies impartiality, monotonicity, and individual full rank. - Proving that for the case of $n \geq 5$, individual full rank can be strengthened to weak unanimity by giving up monotonicity, and this mechanism can be implemented in polynomial time. These results show that impartiality is a very strict requirement in designing ranking aggregation mechanisms, but it is compatible with some other properties. In particular, when the number of agents is large enough ($n \geq 4$ or $n \geq 5$), mechanisms that are both fair and possess other desirable characteristics can be designed. In addition, the paper also discusses the specific construction methods of these mechanisms and uses techniques in graph theory and combinatorial mathematics to prove the existence and effectiveness of these mechanisms. For example, by reducing the problem to the coloring problem of graphs or hypergraphs, the author can construct mechanisms that satisfy the required conditions. In summary, this paper aims to explore how to maximize the preservation of other desirable ranking aggregation properties while maintaining impartiality, and provides specific construction methods and theoretical proofs.

Impartial Rank Aggregation

Fair Rank Aggregation

Impartial Selection with Additive Guarantees via Iterated Deletion

Impartial Selection Under Combinatorial Constraints

Rank Aggregation via Nuclear Norm Minimization

Individualized rank aggregation using nuclear norm regularization

Rank Aggregation Using Scoring Rules

Diversity in Kemeny Rank Aggregation: A Parameterized Approach

Partial Rankings of Optimizers

Ranking with Fairness Constraints

Improved Bounds for Single-Nomination Impartial Selection

Ranking with Ties based on Noisy Performance Data

Asymptotically Optimal Sequential Design for Rank Aggregation

Rate-Optimal Rank Aggregation with Private Pairwise Rankings

Monotone Retargeting for Unsupervised Rank Aggregation with Object Features

Fair Ranking as Fair Division: Impact-Based Individual Fairness in Ranking

Optimizing positional scoring rules for rank aggregation

Sampling Individually-Fair Rankings that are Always Group Fair

Fairness in Ranking: Robustness through Randomization without the Protected Attribute

Deterministic Impartial Selection with Weights

Ranking with Multiple Objectives