As LLM-based agents with user-instructed goals are becoming widely deployed, they increasingly encounter each other in strategic interactions, and face challenges of finding mutually beneficial outcomes. Prior literature has argued that cooperation problems such as the Prisoner's Dilemma are resolvable in settings where agents know they follow very similar decision making patterns, as for example in monocultural AI ecosystems. Following that line of work, this paper introduces the first framework for evaluating LLM decision making when agents are provided with graded similarity signals. Among our findings, we establish that different LLM models vary drastically in how they navigate similarity signals, with some modern models showing consistent behavior across cooperation problems, payoff structures, and prompt framing. Perhaps surprisingly, our experiments also show that the dataset based on which the similarity signal is computed has small to no impact on induced cooperation, and that LLM models systematically self-identify as highly similar when asked to evaluate another model's chain-of-thought reasoning by themselves. Finally, we develop an LLM-behavioral-game-theoretic model that captures some of their reasoning rationale, and show that it can support cooperative outcomes in equilibrium under sufficiently high similarity scores.

Core stability is a natural and well-studied notion for group fairness in multi-winner voting, where the task is to select a committee from a pool of candidates. We study the setting where voters either approve or disapprove of each candidate; here, it remains a major open problem whether a core-stable committee always exists. In this work, we develop an approach based on mixed-integer linear programming for deciding whether and when core-stable committees are guaranteed to exist. In contrast to SAT-based approaches popular in computational social choice, our method can produce proofs for a specific number of candidates independent of the number of voters. In addition to these computational gains, our program lends itself to a novel duality-based reformulation of the core stability problem, from which we obtain new existence results in special cases. Further, we use our framework to reveal previously unknown relationships between core stability and other desirable properties, such as notions of priceability.

Rank aggregation has critical applications for developing AI agents, as well as for evaluating them. However, different methods can give rise to significantly different aggregate rankings, impacting these applications. Indeed, work in social choice and statistics has produced many rank aggregation methods, each with its desirable properties, but also with its limitations. Given this trade-off, how do we decide which aggregation rule to use, _i.e._, what is a good _rule picking rule (RPR)_? In this paper, we design a data-driven RPR that identifies the best method for each dataset without assuming a generative model. The principle behind our RPR is to maximize consistency if the data collection process was repeated. We show that our method satisfies several consistency-related axioms failed by a wide class of natural RPRs. While we prove that the computational problem of maximizing consistency is hard, we provide a sampling-based implementation that is efficient in practice. We run this implementation on known statistical models to experimentally demonstrate its desirable properties, as well as on real-world data where our method provides important insights into how to improve consistency.

It is well known, by the Gibbard-Satterthwaite Theorem, that when there are more than two candidates, any non-dictatorial voting rule can be manipulated by untruthful voters. But how strong is the incentive to manipulate under different voting rules? We suggest measuring the potential advantage of a strategic voter by asking how many copies of their (truthful) vote must be added to the election in order to achieve an outcome as good as their best manipulation. Intuitively, this definition quantifies what a voter can gain by manipulating in comparison to what they would have gained by finding like-minded voters to join the election. The higher the former is, the more incentive a voter will have to manipulate, even when it is computationally costly. Using this framework, we obtain a principled method to measure and compare the manipulation potential for different voting rules. We analyze and report this potential for well-known and broad classes of social choice functions. In particular, we show that the positional scoring rule with the smallest manipulation potential will always be either Borda Count (if the number of voters outweighs the number of candidates) or Plurality (vice versa). Further, we prove that any rule satisfying a weak form of majority consistency (and therefore any Condorcet-consistent rule) cannot outperform Plurality, and that any majoritarian Condorcet rule will perform significantly worse. Consequently, out of the voting rules we analyze, Borda Count stands out as the only one with a manipulation potential that does not grow with the number of voters. By establishing a clear separation between different rules in terms of manipulation potential, our work paves the way for the search for rules that provide voters with minimal incentive to manipulate.

Given a set of items and a set of evaluators who all individually rank them, how do we aggregate these evaluations into a single societal ranking? Work in social choice and statistics has produced many aggregation methods for this problem, each with its desirable properties, but also with its limitations. Further, existing impossibility results rule out designing a single method that achieves every property of interest. Faced with this trade-off between incompatible desiderata, how do we decide which aggregation rule to use, i.e., what is a good rule picking rule? In this paper, we formally address this question by introducing a novel framework for rule picking rules (RPRs). We then design a data-driven RPR that identifies the best aggregation method for each specific setting, without assuming any generative model. The principle behind our RPR is to pick the rule which maximizes the consistency of the output ranking if the data collection process were repeated. We introduce several consistency-related axioms for RPRs and show that our method satisfies them, including those failed by a wide class of natural RPRs. While we prove that the algorithmic problem of maximizing consistency is computationally hard, we provide a sampling-based implementation of our RPR that is efficient in practice. We run this implementation on known statistical models and find that, when possible, our method selects the maximum likelihood estimator of the data. Finally, we show that our RPR can be used in many real-world settings to gain insights about how the rule currently being used can be modified or replaced to substantially improve the consistency of the process. Taken together, our work bridges an important gap between the axiomatic and statistical approaches to rank aggregation, laying a robust theoretical and computational foundation for principled rule picking.

We study two axioms for social choice functions that capture the impact of similar candidates: independence of clones (IoC) and composition consistency (CC). We clarify the relationship between these axioms by observing that CC is strictly more demanding than IoC, and investigate whether common voting rules that are known to be independent of clones (such as STV, Ranked Pairs, Schulze, and Split Cycle) are composition-consistent. While for most of these rules the answer is negative, we identify a variant of Ranked Pairs that satisfies CC. Further, we show how to efficiently modify any (neutral) social choice function so that it satisfies CC, while maintaining its other desirable properties. Our transformation relies on the hierarchical representation of clone structures via PQ-trees. We extend our analysis to social preference functions. Finally, we interpret IoC and CC as measures of robustness against strategic manipulation by candidates, with IoC corresponding to strategy-proofness and CC corresponding to obvious strategy-proofness.

The Independence of Clones (IoC) criterion measures a voting rule's robustness to strategic nomination. Prior literature has established empirically that individuals may still submit costly, distortionary misreports even in strategy-proof (SP) settings, due to failure to recognize the SP property. The intersection of these issues motivates the search for mechanisms that are Obviously Independent of Clones (OIoC): where strategic nomination/exiting of clones obviously has no effect on the outcome. We construct a formal and intuitive definition of a voting rule being OIoC and examine five IoC rules to identify whether they satisfy OIoC.

Imperfect-recall games, in which players may forget previously acquired information, have found many practical applications, ranging from game abstractions to team games and testing AI agents. In this paper, we quantify the utility gain by endowing a player with perfect recall, which we call the value of recall (VoR). While VoR can be unbounded in general, we parameterize it in terms of various game properties, namely the structure of chance nodes and the degree of absentmindedness (the number of successive times a player enters the same information set). Further, we identify several pathologies that arise with VoR, and show how to circumvent them. We also study the complexity of computing VoR, and how to optimally apportion partial recall. Finally, we connect VoR to other previously studied concepts in game theory, including the price of anarchy. We use that connection in conjunction with the celebrated smoothness framework to characterize VoR in a broad class of games.

Φ-equilibria – and the associated notion of Φ-regret – are a powerful and flexible framework at the heart of online learning and game theory, whereby enriching the set of deviations Φ begets stronger notions of rationality. Recently, Daskalakis, Farina, Fishelson, Pipis, and Schneider (STOC '24) – abbreviated as DFFPS – settled the existence of efficient algorithms when Φ contains only linear maps under a general, d-dimensional convex constraint set 𝒳. In this paper, we significantly extend their work by resolving the case where Φ is k-dimensional; degree-ℓ polynomials constitute a canonical such example with k = d^O(ℓ). In particular, positing only oracle access to 𝒳, we obtain two main positive results: i) a poly(n, d, k, log(1/ε))-time algorithm for computing ε-approximate Φ-equilibria in n-player multilinear games, and ii) an efficient online algorithm that incurs average Φ-regret at most ε using poly(d, k)/ε^2 rounds. We also show nearly matching lower bounds in the online learning setting, thereby obtaining for the first time a family of deviations that captures the learnability of Φ-regret. From a technical standpoint, we extend the framework of DFFPS from linear maps to the more challenging case of maps with polynomial dimension. At the heart of our approach is a polynomial-time algorithm for computing an expected fixed point of any ϕ: 𝒳→𝒳 based on the ellipsoid against hope (EAH) algorithm of Papadimitriou and Roughgarden (JACM '08). In particular, our algorithm for computing Φ-equilibria is based on executing EAH in a nested fashion – each step of EAH itself being implemented by invoking a separate call to EAH.

Infinitely repeated games support equilibrium concepts beyond those present in one-shot games (e.g., cooperation in the prisoner's dilemma). Nonetheless, repeated games fail to capture our real-world intuition for settings with many anonymous agents interacting in pairs. Repeated games with restarts, introduced by Berker and Conitzer, address this concern by giving players the option to restart the game with someone new whenever their partner deviates from an agreed-upon sequence of actions. In their work, they studied symmetric games with symmetric strategies. We significantly extend these results, introducing and analyzing more general notions of equilibria in asymmetric games with restarts. We characterize which goal strategies players can be incentivized to play in equilibrium, and we consider the computational problem of finding such sequences of actions with minimal cost for the agents. We show that this problem is NP-hard in general. However, when the goal sequence maximizes social welfare, we give a pseudo-polynomial time algorithm.

Variational inequalities (VIs) encompass many fundamental problems in diverse areas ranging from engineering to economics and machine learning. However, their considerable expressivity comes at the cost of computational intractability. In this paper, we introduce and analyze a natural relaxation -- which we refer to as expected variational inequalities (EVIs) -- where the goal is to find a distribution that satisfies the VI constraint in expectation. By adapting recent techniques from game theory, we show that, unlike VIs, EVIs can be solved in polynomial time under general (nonmonotone) operators. EVIs capture the seminal notion of correlated equilibria, but enjoy a greater reach beyond games. We also employ our framework to capture and generalize several existing disparate results, including from settings such as smooth games, and games with coupled constraints or nonconcave utilities.

Infinitely repeated games can support cooperative outcomes that are not equilibria in the one-shot game. The idea is to make sure that any gains from deviating will be offset by retaliation in future rounds. However, this model of cooperation fails in anonymous settings with many strategic agents that interact in pairs. Here, a player can defect and then avoid penalization by immediately switching partners. In this paper, we focus on a specific set of equilibria that avoids this pitfall. In them, agents follow a designated sequence of actions, and restart if their opponent ever deviates. We show that the socially-optimal sequence of actions consists of an infinitely repeating goal value, preceded by a hazing period. We introduce an equivalence relation on sequences and prove that the computational problem of finding a representative from the optimal equivalence class is (weakly) NP-hard. Nevertheless, we present a pseudo-polynomial time dynamic program for this problem, as well as an integer linear program, and show they are efficient in practice. Lastly, we introduce a fully polynomial-time approximation scheme that outputs a hazing sequence with arbitrarily small approximation ratio.

Federated learning (FL) is an increasingly popular approach in machine learning that enables a set of clients to jointly train a global model without ever sharing their private data, using a central server to aggregate clients' local weight updates. However, previous work has shown that the distributed nature of federated learning makes it susceptible to two major attacks: backdoor attacks, where malicious clients submit large weights that incorrectly change model behavior, and gradient inversion attacks, where a malicious eavesdropper is able to reconstruct the clients' training data by viewing the weight updates sent by clients to the central server. Although various solutions have been proposed in the literature that defend against these two attacks separately, present approaches remain largely incompatible, creating a trade-off between defending against the two types of attacks. This poses a major challenge in deploying FL in privacy-sensitive ML applications. We present SAFHE (Secure Aggregation with Fully Homomorphic Encryption), a novel scheme to defend against both backdoor attacks and gradient inversion attacks. Our secure aggregation method combines the use of fully homomorphic encryption (FHE) and the gradient norm clipping defense to defend against large malicious client updates, by pre-weighting client updates using a function that can be evaluated in the encrypted domain. This allows the server to reject large-magnitude updates without seeing their cleartext values. We demonstrate that Chebyshev approximations of a product of sigmoids work for this purpose, and perform simulations suggesting that such a scheme can defend against backdoor attacks without significantly impacting model accuracy. Additionally, we show that these approximations can be accurately and efficiently computed in the encrypted domain.

In this paper, we provide a comprehensive overview of a recent debate over the quantum versus classical solvability of bounded distance decoding (BDD). Specifically, we review the work of Eldar and Hallgren [EH22], [Hal21] demonstrating a quantum algorithm solving $\lambda_1 2^{-\Omega(\sqrt{k \log q})}$-BDD in polynomial time for lattices of periodicity $q$, finite group rank $k$, and shortest lattice vector length $\lambda_1$. Subsequently, we prove the results of [DvW21a], [DvW21b] with far greater detail and elaboration than in the original work. Namely, we show that there exists a deterministic, classical algorithm achieving the same result.

The Independence of Clones (IoC) criterion for social choice functions (voting rules) measures a function's robustness to strategic nomination. However, prior literature has established empirically that individuals cannot always recognize whether or not a mechanism is strategy-proof and may still submit costly, distortionary misreports even in strategy-proof settings. The intersection of these issues motivates the search for mechanisms which are Obviously Independent of Clones (OIoC): where strategic nomination or strategic exiting of clones obviously have no effect on the outcome of the election. We examine three IoC ranked-choice voting mechanisms and the pre-existing proofs that they are independent of clones: Single Transferable Vote (STV), Ranked Pairs, and the Schulze method. We construct a formal definition of a voting system being Obviously Independent of Clones based on a reduction to a clocked election by considering a bounded agent. Finally, we show that STV and Ranked Pairs are OIoC, whereas we prove an impossibility result for the Schulze method showing that this voting system is not OIoC.