Committee-selection problems arise in many contexts and applications, and there has been increasing interest within the social choice research community on identifying which properties are satisfied by different multi-winner voting rules. In this work, we propose a data-driven framework to evaluate how frequently voting rules violate axioms across diverse preference distributions in practice, shifting away from the binary perspective of axiom satisfaction given by worst-case analysis. Using this framework, we analyze the relationship between multi-winner voting rules and their axiomatic performance under several preference distributions, and propose a methodology for systematically minimizing axioms violations. Our results suggest that data-driven approaches to social choice can inform the design of new voting systems and support the continuation of data-driven research in social choice.

In self-consuming generative models that train on their own outputs, alignment with user preferences becomes a recursive rather than one-time process. In this paper, we provide the first formal foundation for analyzing the long-term effects of such recursive retraining on alignment. Under a two-stage curation mechanism based on the Bradley–Terry (BT) model, we model alignment as an interaction between two factions: the Model Owner, who filters which outputs should be learned by the model, and the Public User, who determines which outputs are ultimately shared and retained through interactions with the model. Our analysis reveals three structural convergence regimes: consensus collapse, compromise on shared optima, and asymmetric refinement, depending on the degree of preference alignment. We prove a fundamental impossibility theorem: no recursive BT-based curation mechanism can simultaneously preserve diversity, ensure symmetric influence, and eliminate dependence on initialization. Framing the process as dynamic social choice, we show that alignment is not a static goal but an evolving equilibrium shaped by power asymmetries and path dependence.

Opponent modeling methods typically involve two crucial steps: building a belief distribution over opponents' strategies, and exploiting this opponent model by playing a best response. However, existing approaches typically require domain-specific heurstics to come up with such a model, and algorithms for approximating best responses are hard to scale in large, imperfect information domains. In this work, we introduce a scalable and generic multiagent training regime for opponent modeling using deep game-theoretic reinforcement learning. We first propose Generative Best Respoonse (GenBR), a best response algorithm based on Monte-Carlo Tree Search (MCTS) with a learned deep generative model that samples world states during planning. This new method scales to large imperfect information domains and can be plug and play in a variety of multiagent algorithms. We use this new method under the framework of Policy Space Response Oracles (PSRO), to automate the generation of an offline opponent model via iterative game-theoretic reasoning and population-based training. We propose using solution concepts based on bargaining theory to build up an opponent mixture, which we find identifying profiles that are near the Pareto frontier. Then GenBR keeps updating an online opponent model and reacts against it during gameplay. We conduct behavioral studies where human participants negotiate with our agents in Deal-or-No-Deal, a class of bilateral bargaining games. Search with generative modeling finds stronger policies during both training time and test time, enables online Bayesian co-player prediction, and can produce agents that achieve comparable social welfare and Nash bargaining score negotiating with humans as humans trading among themselves.

AI agents are commonly aligned with "human values" through reinforcement learning from human feedback (RLHF), where a single reward model is learned from aggregated human feedback and used to align an agent's behavior. However, human values are not homogeneous--different people hold distinct and sometimes conflicting values. Aggregating feedback into a single reward model risks disproportionately suppressing minority preferences. To address this, we present a novel reward modeling approach for learning individualized reward models. Our approach uses a language model to guide users through reflective dialogues where they critique agent behavior and construct their preferences. This personalized dialogue history, containing the user's reflections and critiqued examples, is then used as context for another language model that serves as an individualized reward function (what we call a "verbal reward model") for evaluating new trajectories. In studies with 30 participants, our method achieved a 9-12% improvement in accuracy over non-reflective verbal reward models while being more sample efficient than traditional supervised learning methods.

The field of eXplainable artificial intelligence (XAI) has produced a plethora of methods (e.g., saliency-maps) to gain insight into artificial intelligence (AI) models, and has exploded with the rise of deep learning (DL). However, human-participant studies question the efficacy of these methods, particularly when the AI output is wrong. In this study, we collected and analyzed 156 human-generated text and saliency-based explanations collected in a question-answering task (N = 40) and compared them empirically to state-of-the-art XAI explanations (integrated gradients, conservative LRP, and ChatGPT) in a human-participant study (N = 136). Our findings show that participants found human saliency maps to be more helpful in explaining AI answers than machine saliency maps, but performance negatively correlated with trust in the AI model and explanations. This finding hints at the dilemma of AI errors in explanation, where helpful explanations can lead to lower task performance when they support wrong AI predictions.

Citizen-focused democratic processes where participants deliberate on alternatives and then vote to make the final decision are increasingly popular today. While the computational social choice literature has extensively investigated voting rules, there is limited work that explicitly looks at the interplay of the deliberative process and voting. In this paper, we build a deliberation model using established models from the opinion-dynamics literature and study the effect of different deliberation mechanisms on voting outcomes achieved when using well-studied voting rules. Our results show that deliberation generally improves welfare and representation guarantees, but the results are sensitive to how the deliberation process is organized. We also show, experimentally, that simple voting rules, such as approval voting, perform as well as more sophisticated rules such as proportional approval voting or method of equal shares if deliberation is properly supported. This has ramifications on the practical use of such voting rules in citizen-focused democratic processes.

While it has long been recognized that a team of individual learning agents can be greater than the sum of its parts, recent work has shown that larger teams are not necessarily more effective than smaller ones. In this paper, we study why and under which conditions certain team structures promote effective learning for a population of individual learning agents. We show that, depending on the environment, some team structures help agents learn to specialize into specific roles, resulting in more favorable global results. However, large teams create credit assignment challenges that reduce coordination, leading to large teams performing poorly compared to smaller ones. We support our conclusions with both theoretical analysis and empirical results.

In the last decade, there have been significant advances in multi-agent reinforcement learning (MARL) but there are still numerous challenges, such as high sample complexity and slow convergence to stable policies, that need to be overcome before wide-spread deployment is possible. However, many real-world environments already, in practice, deploy sub-optimal or heuristic approaches for generating policies. An interesting question that arises is how to best use such approaches as advisors to help improve reinforcement learning in multi-agent domains. We provide a principled framework for incorporating action recommendations from online sub-optimal advisors in multi-agent settings. We describe the problem of ADvising Multiple Intelligent Reinforcement Agents (ADMIRAL) in nonrestrictive general-sum stochastic game environments and present two novel Q-learning-based algorithms: ADMIRAL - Decision Making (ADMIRAL-DM) and ADMIRAL - Advisor Evaluation (ADMIRAL-AE), which allow us to improve learning by appropriately incorporating advice from an advisor (ADMIRAL-DM), and evaluate the effectiveness of an advisor (ADMIRAL-AE). We analyze the algorithms theoretically and provide fixed point guarantees regarding their learning in general-sum stochastic games. Furthermore, extensive experiments illustrate that these algorithms: can be used in a variety of environments, have performances that compare favourably to other related baselines, can scale to large state-action spaces, and are robust to poor advice from advisors.

While there has been an increasing focus on the use of game theoretic models for autonomous driving, empirical evidence shows that there are still open questions around dealing with the challenges of common knowledge assumptions as well as modeling bounded rationality. To address some of these practical challenges, we develop a framework of generalized dynamic cognitive hierarchy for both modelling naturalistic human driving behavior as well as behavior planning for autonomous vehicles (AV). This framework is built upon a rich model of level-0 behavior through the use of automata strategies, an interpretable notion of bounded rationality through safety and maneuver satisficing, and a robust response for planning. Based on evaluation on two large naturalistic datasets as well as simulation of critical traffic scenarios, we show that i) automata strategies are well suited for level-0 behavior in a dynamic level-k framework, and ii) the proposed robust response to a heterogeneous population of strategic and non-strategic reasoners can be an effective approach for game theoretic planning in AV.

Voting is a crucial methodology for eliciting and combining agents' preferences and information across many applications. Just as there are numerous voting rules exhibiting different properties, we also see many different voting systems. In this paper we investigate how different voting systems perform as a function of the characteristics of the underlying voting population and social network. In particular, we compare direct democracy, liquid democracy, and sortition in a ground truth voting context. Through simulations -- using both real and artificially generated social networks -- we illustrate how voter competency distributions and levels of direct participation affect group accuracy differently in each voting mechanism. Our results can be used to guide the selection of a suitable voting system based on the characteristics of a particular voting setting.

For problems requiring cooperation, many multiagent systems implement solutions among either individual agents or across an entire population towards a common goal. Multiagent teams are primarily studied when in conflict; however, organizational psychology (OP) highlights the benefits of teams among human populations for learning how to coordinate and cooperate. In this paper, we propose a new model of multiagent teams for reinforcement learning (RL) agents inspired by OP and early work on teams in artificial intelligence. We validate our model using complex social dilemmas that are popular in recent multiagent RL and find that agents divided into teams develop cooperative pro-social policies despite incentives to not cooperate. Furthermore, agents are better able to coordinate and learn emergent roles within their teams and achieve higher rewards compared to when the interests of all agents are aligned.

In the last decade, there have been significant advances in multi-agent reinforcement learning (MARL) but there are still numerous challenges, such as high sample complexity and slow convergence to stable policies, that need to be overcome before wide-spread deployment is possible. However, many real-world environments already, in practice, deploy sub-optimal or heuristic approaches for generating policies. An interesting question that arises is how to best use such approaches as advisors to help improve reinforcement learning in multi-agent domains. In this paper, we provide a principled framework for incorporating action recommendations from online suboptimal advisors in multi-agent settings. We describe the problem of ADvising Multiple Intelligent Reinforcement Agents (ADMIRAL) in nonrestrictive general-sum stochastic game environments and present two novel Q-learning based algorithms: ADMIRAL - Decision Making (ADMIRAL-DM) and ADMIRAL - Advisor Evaluation (ADMIRAL-AE), which allow us to improve learning by appropriately incorporating advice from an advisor (ADMIRAL-DM), and evaluate the effectiveness of an advisor (ADMIRAL-AE). We analyze the algorithms theoretically and provide fixed point guarantees regarding their learning in general-sum stochastic games. Furthermore, extensive experiments illustrate that these algorithms: can be used in a variety of environments, have performances that compare favourably to other related baselines, can scale to large state-action spaces, and are robust to poor advice from advisors.

We study one-sided matching problems where each agent must be assigned at most one object. In this classic problem it is often assumed that agents specify only ordinal preferences over objects and the goal is to return a matching that satisfies some desirable property such as Pareto optimality or rank-maximality. However, agents may have cardinal utilities describing their preference intensities and ignoring this can result in welfare loss. We investigate how to elicit additional cardinal information from agents using simple threshold queries and use it in turn to design algorithms that return a matching satisfying a desirable matching property, while also achieving a good approximation to the optimal welfare among all matchings satisfying that property. Overall, our results show how one can improve welfare by even non-adaptively asking agents for just one bit of extra information per object.

Social choice theory provides a principled framework for the aggregation of individuals' preferences in support of group decision‐making and recommendation. Much of this work, however, either assumes that individuals' subjective preferences (and thus, their votes) are correctly specified by the individuals themselves, or alternatively that the votes of individuals are noisy estimates of some underlying ground truth over rankings of alternatives. We argue that neither model appropriately addresses some of the issues which arise in the context of group‐recommendation domains where individuals have subjective preferences but for some reason (eg, the high cognitive burden, concerns about privacy, etc.) may instead vote using a noisy estimate of their subjective preference rankings. In this paper, we propose a general probabilistic framework for modeling noisy subjective preferences, and explore the accuracy and reliability of four well‐studied voting rules under various noise models. Our results demonstrate that there is no single reliable method amongst the examined methods. Specifically, we observe the change in noise distribution can flip one method from being the most reliable to the least.

Artificial intelligence (AI) assistants for clinical decision making show increasing promise in medicine. However, medical assessments can be contentious, leading to expert disagreement. This raises the question of how AI assistants should be designed to handle the classification of ambiguous cases. Our study compared two AI assistants that provide classification labels for medical time series data along with quantitative uncertainty estimates: conventional vs. ambiguity-aware. We simulated our ambiguity-aware AI based on real-world expert discussions to highlight cases likely to lead to expert disagreement, and to present arguments for conflicting classification choices. Our results demonstrate that ambiguity-aware AI can alter expert workflows by significantly increasing the proportion of contentious cases reviewed. We also found that the relevance of AI-provided arguments (selected from guidelines either randomly or by experts) affected experts' accuracy at revising AI-suggested labels. Our work contributes a novel perspective on the design of AI for contentious clinical assessments.

Expert disagreement is pervasive in clinical decision making and collective adjudication is a useful approach for resolving divergent assessments. Prior work shows that expert disagreement can arise due to diverse factors including expert background, the quality and presentation of data, and guideline clarity. In this work, we study how these factors predict initial discrepancies in the context of medical time series analysis, examining why certain disagreements persist after adjudication, and how adjudication impacts clinical decisions. Results from a case study with 36 experts and 4,543 adjudicated cases in a sleep stage classification task show that these factors contribute to both initial disagreement and resolvability, each in their own unique way. We provide evidence suggesting that structured adjudication can lead to significant revisions in treatment-relevant clinical parameters. Our work demonstrates how structured adjudication can support consensus and facilitate a deep understanding of expert disagreement in medical data analysis.

Collaborative crowdsourcing tasks allow crowd workers to solve problems that they could not handle alone, but worker motivation in these tasks is not well understood. In this paper, we study how to motivate groups of workers by paying them equitably. To this end, we characterize existing collaborative tasks based on the types of information available to crowd workers. Then, we apply concepts from equity theory to show how fair payments relate to worker motivation, and we propose two theoretically grounded classes of fair payments. Finally, we run two experiments using an audio transcription task on Amazon Mechanical Turk to understand how workers perceive these payments. Our results show that workers recognize fair and unfair payment divisions, but are biased toward payments that reward them more. Additionally, our data suggests that fair payments could lead to a small increase in worker effort. These results inform the design of future collaborative crowdsourcing tasks.

Crowdsourced classification of data typically assumes that objects can be unambiguously classified into categories. In practice, many classification tasks are ambiguous due to various forms of disagreement. Prior work shows that exchanging verbal justifications can significantly improve answer accuracy over aggregation techniques. In this work, we study how worker deliberation affects resolvability and accuracy using case studies with both an objective and a subjective task. Results show that case resolvability depends on various factors, including the level and reasons for the initial disagreement, as well as the amount and quality of deliberation activities. Our work reinforces the finding that deliberation can increase answer accuracy and the importance of verbal discussion in this process. We contribute a new public data set on worker deliberation for text classification tasks, and discuss considerations for the design of deliberation workflows for classification.

We study the classic cake cutting problem from a mechanism design perspective, in particular focusing on deterministic mechanisms that are strategyproof and fair. We begin by looking at mechanisms that are non-wasteful and primarily show that for even the restricted class of piecewise constant valuations there exists no direct-revelation mechanism that is strategyproof and even approximately proportional. Subsequently, we remove the non-wasteful constraint and show another impossibility result stating that there is no strategyproof and approximately proportional direct-revelation mechanism that outputs contiguous allocations, again, for even the restricted class of piecewise constant valuations. In addition to the above results, we also present some negative results when considering an approximate notion of strategyproofness, show a connection between direct-revelation mechanisms and mechanisms in the Robertson-Webb model when agents have piecewise constant valuations, and finally also present a (minor) modification to the well-known Even-Paz algorithm that has better incentive-compatible properties for the cases when there are two or three agents.

Understanding when and how computational complexity can be used to protect elections against different manipulative actions has been a highly active research area over the past two decades. A recent body of work, however, has shown that many of the NP-hardness shields, previously obtained, vanish when the electorate has single-peaked or nearly single-peaked preferences. In light of these results, we investigate whether it is possible to reimpose NP-hardness shields for such electorates by allowing the voters to specify partial preferences instead of insisting they cast complete ballots. In particular, we show that in single-peaked and nearly single-peaked electorates, if voters are allowed to submit top-truncated ballots, then the complexity of manipulation and bribery for many voting rules increases from being in P to being NP-complete.

A confluence of technical, economic and political forces are revolutionizing the energy sector. Policy-makers, who decide on incentives and penalties for possible courses of actions, play a critical role in determining which outcomes arise. However, designing appropriate energy policies is a complex and challenging task. Our vision is to provide tools and methodologies for policy makers so that they can leverage the power of big data to make evidence-based decisions. In this paper we present an approach we call big-data mechanism design which combines a mechanism design framework with stakeholder surveys and data to allow policy-makers to gauge the costs and benefits of potential policy decisions.We illustrate the effectiveness of this approach in a concrete application domain: the peaksaver PLUS program in Ontario, Canada.

Deciding the outcome of an election when voters have provided only partial orderings over their preferences requires voting rules that accommodate missing data. While existing techniques, including considerable recent work, address missingness through circumvention, we propose the novel application of conventional machine learning techniques to predict the missing components of ballots via latent patterns in the information that voters are able to provide. We show that suitable predictive features can be extracted from the data, and demonstrate the high performance of our new framework on the ballots from many real world elections, including comparisons with existing techniques for voting with partial orderings. Our technique offers a new and interesting conceptualization of the problem, with stronger connections to machine learning than conventional social choice techniques.

We consider the problem of repeatedly matching a set of alternatives to a set of agents with dynamic ordinal preferences. Despite a recent focus on designing one-shot matching mechanisms in the absence of monetary transfers, little study has been done on strategic behavior of agents in sequential assignment problems. We formulate a generic dynamic matching problem via a sequential stochastic matching process. We design a mechanism based on random serial dictatorship (RSD) that, given any history of preferences and matching decisions, guarantees global stochastic strategyproofness while satisfying desirable local properties. We further investigate the notion of envyfreeness in such sequential settings.

We consider the problem of repeatedly matching a set of alternatives to a set of agents in the absence of monetary transfer. We propose a generic framework for evaluating sequential matching mechanisms with dynamic preferences, and show that unlike single-shot settings, the random serial dictatorship mechanism is manipulable.

Wildland fires (or wildfires) occur on all continents except for Antarctica. These fires threaten communities, change ecosystems, destroy vast quantities of natural resources and the cost estimates of the damage done annually is in the billions of dollars. Controlling wildland fires is resource-intensive and there are numerous examples where the resource demand has outstripped resource availability. Trends in changing climates, fire occurrence and the expansion of the wildland-urban interface all point to increased resource shortages in the future. One approach for coping with these shortages has been the sharing of resources across different wildland-fire agencies. This introduces new issues as agencies have to balance their own needs and risk-management with their desire to help fellow agencies in need. Using ideas from the field of multiagent systems, we conduct the first analysis of strategic issues arising in resource-sharing for wildland-fire control. We also argue that the wildland-fire domain has numerous features that make it attractive to researchers in artificial intelligence and computational sustainability.

In this paper, we propose a novel incentive mechanism for promoting honesty in electronic marketplaces that is based on trust modeling. In our mechanism, buyers model other buyers and select the most trustworthy ones as their neighbors to form a social network which can be used to ask advice about sellers. In addition, however, sellers model the reputation of buyers based on the social network. Reputable buyers provide truthful ratings for sellers, and are likely to be neighbors of many other buyers. Sellers will provide more attractive products to reputable buyer to build their own reputation. We theoretically prove that a marketplace operating with our mechanism leads to greater profit both for honest buyers and honest sellers. We emphasize the value of our approach through a series of illustrative examples and in direct contrast to other frameworks for addressing agent trustworthiness. In all, we offer an effective approach for the design of e‐marketplaces that is attractive to users, through its promotion of honesty.

In this paper we introduce and analyze social distance games, a family of non-transferable utility coalitional games where an agent's utility is a measure of closeness to the other members of the coalition. We study both social welfare maximisation and stability in these games using a graph theoretic perspective. We use the stability gap to investigate the welfare of stable coalition structures, and propose two new solution concepts with improved welfare guarantees. We argue that social distance games are both interesting in themselves, as well as in the context of social networks.

Recently Argumentation Mechanism Design (ArgMD) was introduced as a paradigm for studying argumentation using game-theoretic techniques. To date, this framework has been used to study under what conditions a direct mechanism based on Dung's grounded semantics is strategy-proof (i.e. truth-enforcing) when knowledge of arguments is private to self-interested agents. In this paper, we study Dung's preferred semantics in order to understand under what conditions it is possible to design strategy-proof mechanisms. This is challenging since, unlike with the grounded semantics, there may be multiple preferred extensions, forcing a mechanism to select one. We show that this gives rise to interesting strategic behaviour, and we show that in general it is not possible to have a strategy-proof mechanism that selects amongst the preferred extensions in a non-biased manner. We also investigaet refinements of preferred semantics which induce unique outcomes, namely the skeptical-preferred and ideal semantics.

In many market settings, agents do not know their preferences a priori. Instead, they may have to solve computationally complex optimization problems, query databases, or perform expensive searches in order to determine their values for di#erent outcomes. For such settings, we have introduced the deliberation equilibrium as the game-theoretic solution concept where the agents' deliberation actions are modeled as part of their strategies.

Coalition formation is a key topic in multiagent systems. One would prefer a coalition structure that maximizes the sum of the values of the coalitions, but often the number of coalition structures is too large to allow for exhaustive search for the optimal one. We present experimental results for three anytime algorithms that search the space of coalition structures. We show that, in the average case, all three algorithms do much better than the recently established theoretical worst case results in Sandholm et al. (1999a). We also show that no one algorithm is dominant. Each algorithm's performance is influenced by the particular instance distribution, with each algorithm outperforming the others for different instances. We present a possible explanation for the behaviour of the algorithms and support our hypothesis with data collected from a controlled experimental run.