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.

We propose a self-play approach to generating strategies for playing in multi-player games, where strategies are represented as computer code. We use large language models (LLMs) to generate pieces of code to play in the game, which we refer to as generated bots. We engage the LLM generated bots in competitions, designed to generate increasingly stronger strategies. We follow game theoretic principles in organizing these tournaments, and use a Policy Space Response Oracle (PSRO) approach. We start with an initial set of LLM generated bots, and continue in rounds for adding new bots into the population. Each round adds a bot to the population by asking the LLM to produce code for playing against a bot representing the Nash equilibrium mixture over the current population. Our analysis shows that even a few rounds are sufficient to produces strong bots for playing the game. Our demo shows the process for the game of Checkers. We allow users to select initial bots in the population, run the process, inspect how the bots evolve over time, and play against the generated bots.

AI research agents are demonstrating great potential to accelerate scientific progress by automating the design, implementation, and training of machine learning models. We focus on methods for improving agents' performance on MLE-bench, a challenging benchmark where agents compete in Kaggle competitions to solve real-world machine learning problems. We formalize AI research agents as search policies that navigate a space of candidate solutions, iteratively modifying them using operators. By designing and systematically varying different operator sets and search policies (Greedy, MCTS, Evolutionary), we show that their interplay is critical for achieving high performance. Our best pairing of search strategy and operator set achieves a state-of-the-art result on MLE-bench lite, increasing the success rate of achieving a Kaggle medal from 39.6% to 47.7%. Our investigation underscores the importance of jointly considering the search strategy, operator design, and evaluation methodology in advancing automated machine learning.

Rapidly improving large language models (LLMs) have the potential to assist in scientific progress. One critical skill in this endeavor is the ability to faithfully reproduce existing work. To evaluate the capability of AI agents to reproduce complex code in an active research area, we introduce the Automated LLM Speedrunning Benchmark, leveraging the research community's contributions to the $\textit{NanoGPT speedrun}$, a competition to train a GPT-2 model in the shortest time. Each of the 19 speedrun tasks provides the agent with the previous record's training script, optionally paired with one of three hint formats, ranging from pseudocode to paper-like descriptions of the new record's improvements. Records execute quickly by design and speedrun improvements encompass diverse code-level changes, ranging from high-level algorithmic advancements to hardware-aware optimizations. These features make the benchmark both accessible and realistic for the frontier problem of improving LLM training. We find that recent frontier reasoning LLMs combined with SoTA scaffolds struggle to reimplement already-known innovations in our benchmark, even when given detailed hints. Our benchmark thus provides a simple, non-saturated measure of an LLM's ability to automate scientific reproduction, a necessary (but not sufficient) skill for an autonomous research agent.

Recent work has shown the potential benefit of selective prediction systems that can learn to defer to a human when the predictions of the AI are unreliable, particularly to improve the reliability of AI systems in high-stakes applications like healthcare or conservation. However, most prior work assumes that human behavior remains unchanged when they solve a prediction task as part of a human-AI team as opposed to by themselves. We show that this is not the case by performing experiments to quantify human-AI interaction in the context of selective prediction. In particular, we study the impact of communicating different types of information to humans about the AI system's decision to defer. Using real-world conservation data and a selective prediction system that improves expected accuracy over that of the human or AI system working individually, we show that this messaging has a significant impact on the accuracy of human judgements. Our results study two components of the messaging strategy: 1) Whether humans are informed about the prediction of the AI system and 2) Whether they are informed about the decision of the selective prediction system to defer. By manipulating these messaging components, we show that it is possible to significantly boost human performance by informing the human of the decision to defer, but not revealing the prediction of the AI. We therefore show that it is vital to consider how the decision to defer is communicated to a human when designing selective prediction systems, and that the composite accuracy of a human-AI team must be carefully evaluated using a human-in-the-loop framework.

In many multi-agent settings, participants can form teams to achieve collective outcomes that may far surpass their individual capabilities. Measuring the relative contributions of agents and allocating them shares of the reward that promote long-lasting cooperation are difficult tasks. Cooperative game theory offers solution concepts identifying distribution schemes, such as the Shapley value, that fairly reflect the contribution of individuals to the performance of the team or the Core, which reduces the incentive of agents to abandon their team. Applications of such methods include identifying influential features and sharing the costs of joint ventures or team formation. Unfortunately, using these solutions requires tackling a computational barrier as they are hard to compute, even in restricted settings. In this work, we show how cooperative game-theoretic solutions can be distilled into a learned model by training neural networks to propose fair and stable payoff allocations. We show that our approach creates models that can generalize to games far from the training distribution and can predict solutions for more players than observed during training. An important application of our framework is Explainable AI: our approach can be used to speed-up Shapley value computations on many instances.

Adversarial training, a special case of multi-objective optimization, is an increasingly prevalent machine learning technique: some of its most notable applications include GAN-based generative modeling and self-play techniques in reinforcement learning which have been applied to complex games such as Go or Poker. In practice, a \emph{single} pair of networks is typically trained in order to find an approximate equilibrium of a highly nonconcave-nonconvex adversarial problem. However, while a classic result in game theory states such an equilibrium exists in concave-convex games, there is no analogous guarantee if the payoff is nonconcave-nonconvex. Our main contribution is to provide an approximate minimax theorem for a large class of games where the players pick neural networks including WGAN, StarCraft II and Blotto Game. Our findings rely on the fact that despite being nonconcave-nonconvex with respect to the neural networks parameters, these games are concave-convex with respect to the actual models (e.g., functions or distributions) represented by these neural networks.

We propose a system for conducting an auction over locations in a continuous space. It enables participants to express their preferences over possible choices of location in the space, selecting the location that maximizes the total utility of all agents. We prevent agents from tricking the system into selecting a location that improves their individual utility at the expense of others by using a pricing rule that gives agents no incentive to misreport their true preferences. The system queries participants for their utility in many random locations, then trains a neural network to approximate the preference function of each participant. The parameters of these neural network models are transmitted and processed by the auction mechanism, which composes these into differentiable models that are optimized through gradient ascent to compute the final chosen location and charged prices.

Roma Patel, Marta Garnelo, Ian Gemp, Chris Dyer, Yoram Bachrach. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

A core question in multi-agent systems is understanding the motivations for an agent's actions based on their behavior. Inverse reinforcement learning provides a framework for extracting utility functions from observed agent behavior, casting the problem as finding domain parameters which induce such a behavior from rational decision makers. We show how to efficiently and scalably extend inverse reinforcement learning to multi-agent settings, by reducing the multi-agent problem to N single-agent problems while still satisfying rationality conditions such as strong rationality. However, we observe that rewards learned naively tend to lack insightful structure, which causes them to produce undesirable behavior when optimized in games with different players from those encountered during training. We further investigate conditions under which rewards or utility functions can be precisely identified, on problem domains such as normal-form and Markov games, as well as auctions, where we show we can learn reward functions that properly generalize to new settings.

Recent advances in deep reinforcement learning (RL) have led to considerable progress in many 2-player zero-sum games, such as Go, Poker and Starcraft. The purely adversarial nature of such games allows for conceptually simple and principled application of RL methods. However real-world settings are many-agent, and agent interactions are complex mixtures of common-interest and competitive aspects. We consider Diplomacy, a 7-player board game designed to accentuate dilemmas resulting from many-agent interactions. It also features a large combinatorial action space and simultaneous moves, which are challenging for RL algorithms. We propose a simple yet effective approximate best response operator, designed to handle large combinatorial action spaces and simultaneous moves. We also introduce a family of policy iteration methods that approximate fictitious play. With these methods, we successfully apply RL to Diplomacy: we show that our agents convincingly outperform the previous state-of-the-art, and game theoretic equilibrium analysis shows that the new process yields consistent improvements.

This article proposes a negotiation game, based on the weighted voting paradigm in cooperative game theory, where agents need to form coalitions and agree on how to share the gains. Despite the prevalence of weighted voting in the real world, there has been little work studying people’s behavior in such settings. This work addresses this gap by combining game-theoretic solution concepts with machine learning models for predicting human behavior in such domains. We present a five-player online version of a weighted voting game in which people negotiate to create coalitions. We provide an equilibrium analysis of this game and collect hundreds of instances of people’s play in the game. We show that a machine learning model with features based on solution concepts from cooperative game theory (in particular, an extension of the Deegan-Packel Index) provide a good prediction of people’s decisions to join coalitions in the game. We designed an agent that uses the prediction model to make offers to people in this game and was able to outperform other people in an extensive empirical study. These results demonstrate the benefit of incorporating concepts from cooperative game theory in the design of agents that interact with people in group decision-making settings.

Zero-sum games such as chess and poker are, abstractly, functions that evaluate pairs of agents, for example labeling them ‘winner’ and ‘loser’. If the game is approximately transitive, then self-play generates sequences of agents of increasing strength. However, nontransitive games, such as rock-paper-scissors, can exhibit strategic cycles, and there is no longer a clear objective – we want agents to increase in strength, but against whom is unclear. In this paper, we introduce a geometric framework for formulating agent objectives in zero-sum games, in order to construct adaptive sequences of objectives that yield open-ended learning. The framework allows us to reason about population performance in nontransitive games, and enables the development of a new algorithm (rectified Nash response, PSRO_rN) that uses game-theoretic niching to construct diverse populations of effective agents, producing a stronger set of agents than existing algorithms. We apply PSRO_rN to two highly nontransitive resource allocation games and find that PSRO_rN consistently outperforms the existing alternatives.

We study the problem of emergent communication, in which language arises because speakers and listeners must communicate information in order to solve tasks. In temporally extended reinforcement learning domains, it has proved hard to learn such communication without centralized training of agents, due in part to a difficult joint exploration problem. We introduce inductive biases for positive signalling and positive listening, which ease this problem. In a simple one-step environment, we demonstrate how these biases ease the learning problem. We also apply our methods to a more extended environment, showing that agents with these inductive biases achieve better performance, and analyse the resulting communications protocols.

A key issue in cooperative game theory is coalitional stability, usually captured by the notion of the core---the set of outcomes that are resistant to group deviations. However, some coalitional games have empty cores, and any outcome in such a game is unstable. We investigate the possibility of stabilizing a coalitional game by using subsidies. We consider scenarios where an external party that is interested in having the players work together offers a supplemental payment to the grand coalition, or, more generally, a particular coalition structure. This payment is conditional on players not deviating from this coalition structure, and may be divided among the players in any way they wish. We define the cost of stability as the minimum external payment that stabilizes the game. We provide tight bounds on the cost of stability, both for games where the coalitional values are nonnegative (profit-sharing games) and for games where the coalitional values are nonpositive (cost-sharing games), under natural assumptions on the characteristic function, such as superadditivity, anonymity, or both. We also investigate the relationship between the cost of stability and several variants of the least core. Finally, we study the computational complexity of problems related to the cost of stability, with a focus on weighted voting games.

We examine the surveying problem, where we attempt to predict how a target user is likely to respond to questions by iteratively querying that user, collaboratively based on the responses of a sample set of users. We focus on an active learning approach, where the next question we select to ask the user depends on their responses to the previous questions. We propose a method for solving the problem based on a Bayesian dimensionality reduction technique. We empirically evaluate our method, contrasting it to benchmark approaches based on augmented linear regression, and show that it achieves much better predictive performance, and is much more robust when there is missing data.

We propose a new attention mechanism for neural based question answering, which depends on varying granularities of the input. Previous work focused on augmenting recurrent neural networks for question answering systems with simple attention mechanisms which are a function of the similarity between a question embedding and an answer embeddings across time. We extend this by making the attention mechanism dependent on a global embedding of the answer attained using a separate network. We evaluate our system on InsuranceQA, a large question answering dataset. Our model outperforms current state-of-the-art results on InsuranceQA. Further, we examine which sections of text our attention mechanism focuses on, and explore its performance across different parameter settings.

We propose a novel supervised approach for text tagging and multi-label text classification based on a multi-head encoder-decoder neural network architecture. Our method predicts which subset of possible tags best matches an input text. It efficiently spends computational resources, exploiting dependencies between tags by encoding an input text into a compact representation which is then passed to multiple decoder classifier heads. We test our architecture on a Twitter hashtag prediction task, comparing it to a baseline model with multiple feedforward networks and a baseline model with multiple recurrent neural networks with GRU cells. We show that our approach achieves a significantly better performance than baselines with an equivalent number of parameters.

Despite the prevalence of weighted voting in the real world, there has been relatively little work studying real people's behavior in such settings. This paper proposes a new negotiation game, based on the weighted voting paradigm in cooperative games, where players need to form coalitions and agree on how to share the gains. We show that solution concepts from cooperative game theory (in particular, an extension of the Deegan-Packel Index) provide a good prediction of people's decisions to join a given coalition. With this insight in mind, we design an agent that combines predictive analytics with decision theory to make offers to people in the game. We show that the agent was able to obtain higher shares from coalitions than did people playing other people, without reducing the acceptance rate of its offers. These results demonstrate the potential of incorporating concepts from cooperative game theory in the design of negotiating agents.

We demonstrate a system for predicting gaming related properties from Twitter accounts. Our system predicts various traits of users based on the tweets publicly available in their profiles. Such inferred traits include degrees of tech-savviness and knowledge on computer games, actual gaming performance, preferred platform, degree of originality, humor and influence on others. Our system is based on machine learning models trained on crowd-sourced data. It allows people to select Twitter accounts of their fellow gamers, examine the trait predictions made by our system, and the main drivers of these predictions. We present empirical results on the performance of our system based on its accuracy on our crowd-sourced dataset.

We propose a system for determining properties of mathematical functions given an image of their graph representation. We demonstrate our approach for two-dimensional graphs (curves of single variable functions) and three-dimensional graphs (surfaces of two variable functions), studying the properties of convexity and symmetry. Our method uses a Convolutional Neural Network which classifies functions according to these properties, without using any hand-crafted features. We propose algorithms for randomly constructing functions with convexity or symmetry properties, and use the images generated by these algorithms to train our network. Our system achieves a high accuracy on this task, even for functions where humans find it difficult to determine the function's properties from its image.

We consider the task of predicting various traits of a person given an image of their face. We aim to estimate traits such as gender, ethnicity and age, as well as more subjective traits as the emotion a person expresses or whether they are humorous or attractive. Due to the recent surge of research on Deep Convolutional Neural Networks (CNNs), we begin by using a CNN architecture, and corroborate that CNNs are promising for facial attribute prediction. To further improve performance, we propose a novel approach that incorporates facial landmark information for input images as an additional channel, helping the CNN learn face-specific features so that the landmarks across various training images hold correspondence. We empirically analyze the performance of our proposed method, showing consistent improvement over the baselines across traits. We demonstrate our system on a sizeable Face Attributes Dataset (FAD), comprising of roughly 200,000 labels, for 10 most sought-after traits, for over 10,000 facial images.

We examine communications in a social network to study user emotional contrast -the propensity of users to express different emotions than those expressed by their neighbors.Our analysis is based on a large Twitter dataset, consisting of the tweets of 123,513 users from the USA and Canada.Focusing on Ekman's basic emotions, we analyze differences between the emotional tone expressed by these users and their neighbors of different types, and correlate these differences with perceived user demographics.We demonstrate that many perceived demographic traits correlate with the emotional contrast between users and their neighbors.Unlike other approaches on inferring user attributes that rely solely on user communications, we explore the network structure and show that it is possible to accurately predict a range of perceived demographic traits based solely on the emotions emanating from users and their neighbors.

The Generalized Second Price (GSP) auction has appealing properties when ads are simple (text based and identical in size), but does not generalize to richer ad settings, whereas truthful mechanisms such as VCG do. However, a straight switch from GSP to VCG incurs significant revenue loss for the search engine. We introduce a transitional mechanism which encourages advertisers to update their bids to their valuations, while mitigating revenue loss. In this setting, it is easier to propose first a payment function rather than an allocation function, so we give a general framework which guarantees incentive compatibility by requiring that the payment functions satisfy two specific properties. Finally, we analyze the revenue impacts of our mechanism on a sample of Bing data.

We demonstrate an approach to predict latent personal attributes including user demographics, online personality, emotions and sentiments from texts published on Twitter. We rely on machine learning and natural language processing techniques to learn models from user communications. We first examine individual tweets to detect emotions and opinions emanating from them, and then analyze all the tweets published by a user to infer latent traits of that individual. We consider various user properties including age, gender, income, education, relationship status, optimism and life satisfaction. We focus on Ekman’s six emotions: anger, joy, surprise, fear, disgust and sadness. Our work can help social network users to understand how others may perceive them based on how they communicate in social media, in addition to its evident applications in online sales and marketing, targeted advertising, large scale polling and healthcare analytics.

Online social networks have changed the ways in which people communicate and interact, and have also impacted the business landscape. One recent trend is firms using online social networks as a part of the job hiring process. Firms scrutinize potential employees using their social network profiles, sometimes even seeking access to restricted parts of the profile, for example by demanding applicants to hand over their passwords. We explore the key criteria and profile components that affect perceptions about a user. Our results are based on datasets consisting of reports of participants who actually took part in a task of evaluating candidates. Participants volunteered their Facebook profiles and CVs, to be examined by other participants who provided a detailed report about their job-suitability. We find that in screening based on social network profiles, a profile owner's education and demographic traits correlate with their job-suitability rating. Many profile components, including textual posts, pictures, likes, and even the friend list, relate to an applicant's perceived job-suitability. Further, diverse criteria play a role in forming job-suitability perceptions, including education and skills, personality, offensive content, physical appearance, interests and age, gender, family status or other demographic traits. Thus screening based on social networking websites is very different from CV based screening, where we find that the dominant criterion is education and skills, with personality being a remote second.

We examine the relation between the emotions users express on social networks and their perceived areas of interests, based on a sample of Twitter users. Our methodology relies on training machine learning models to classify the emotions expressed in tweets, according to Ekman's six high-level emotions. We then used raters, sourced from Amazon's Mechanical Turk, to examine several Twitter profiles and to determine whether the profile owner is interested in various areas, including sports, movies, technology and computing, politics, news, economics, science, arts, health and religion. We find that the propensity of a user to express various emotions correlates with their perceived degree of interest in various areas. We present several models that use the emotional distribution of a Twitter user, as reflected by their tweets, to predict whether they are interested or disinterested in a topic or to determine their degree of interest in a topic.

Svitlana Volkova, Benjamin Van Durme, David Yarowsky, Yoram Bachrach. Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Tutorial Abstracts. 2015.

A prominent approach in collaborative filtering based recommender systems is using dimensionality reduction (matrix factorization) techniques to map users and items into low-dimensional vectors. In such systems, a higher inner product between a user vector and an item vector indicates that the item better suits the user's preference. Traditionally, retrieving the most suitable items is done by scoring and sorting all items. Real world online recommender systems must adhere to strict response-time constraints, so when the number of items is large, scoring all items is intractable.

We examine trade-offs among stakeholders in ad auctions. Our metrics are the revenue for the utility of the auctioneer, the number of clicks for the utility of the users and the welfare for the utility of the advertisers. We show how to optimize linear combinations of the stakeholder utilities, showing that these can be tackled through a GSP auction with a per-click reserve price. We then examine constrained optimization of stakeholder utilities.

Individual differences in personality affect users’ online activities as much as they do in the offline world. This work, based on a sample of over a third of a million users, examines how users’ behaviour in the online environment, captured by their website choices and Facebook profile features, relates to their personality, as measured by the standard Five Factor Model personality questionnaire. Results show that there are psychologically meaningful links between users’ personalities, their website preferences and Facebook profile features. We show how website audiences differ in terms of their personality, present the relationships between personality and Facebook profile features, and show how an individual’s personality can be predicted from Facebook profile features. We conclude that predicting a user’s personality profile can be applied to personalize content, optimize search results, and improve online advertising.

Representation languages for coalitional games are a key research area in algorithmic game theory. There is an inherent tradeoff between how general a language is, allowing it to capture more elaborate games, and how hard it is computationally to optimize and solve such games. One prominent such language is the simple yet expressive Weighted Graph Games (WGGs) representation (Deng and Papadimitriou, 1994), which maintains knowledge about synergies between agents in the form of an edge weighted graph. We consider the problem of finding the optimal coalition structure in WGGs. The agents in such games are vertices in a graph, and the value of a coalition is the sum of the weights of the edges present between coalition members. The optimal coalition structure is a partition of the agents to coalitions, that maximizes the sum of utilities obtained by the coalitions. We show that finding the optimal coalition structure is not only hard for general graphs, but is also intractable for restricted families such as planar graphs which are amenable for many other combinatorial problems. We then provide algorithms with constant factor approximations for planar, minor-free and bounded degree graphs.

Object localization is an image annotation task which consists of finding the location of a target object in an image. It is common to crowdsource annotation tasks and aggregate responses to estimate the true annotation. While for other kinds of annotations consensus is simple and powerful, it cannot be applied to object localization as effectively due to the task's rich answer space and inherent noise in responses. We propose a probabilistic graphical model to localize objects in images based on responses from the crowd. We improve upon natural aggregation methods such as the mean and the median by simultaneously estimating the difficulty level of each question and skill level of every participant. We empirically evaluate our model on crowdsourced data and show that our method outperforms simple aggregators both in estimating the true locations and in ranking participants by their ability. We also propose a simple adaptive sourcing scheme that works well for very sparse datasets.

Agents are asked to rank two objects in a setting where effort is costly and agents differ in quality (which is the probability that they can identify the correct, ground truth, ranking). We study simple output-agreement mechanisms that pay an agent in the case she agrees with the report of another, and potentially penalizes for disagreement through a negative payment. Assuming access to a quality oracle, able to determine whether an agent's quality is above a given threshold, we design a payment scheme that aligns incentives so that agents whose quality is above this threshold participate and invest effort. Precluding negative payments leads the expected cost of this quality-oracle mechanism to increase by a factor of 2 to 5 relative to allowing both positive and negative payments. Dropping the assumption about access to a quality oracle, we further show that negative payments can be used to make agents with quality lower than the quality threshold choose to not to participate, while those above continue to participate and invest effort. Through the appropriate choice of payments, any design threshold can be achieved. This self-selection mechanism has the same expected cost as the cost-minimal quality-oracle mechanism, and thus when using the self-selection mechanism, perfect screening comes for free.

We consider how selfish agents are likely to share revenues derived from maintaining connectivity between important network servers. We model a network where a failure of one node may disrupt communication between other nodes as a cooperative game called the vertex Connectivity Game (CG). In this game, each agent owns a vertex, and controls all the edges going to and from that vertex. A coalition of agents wins if it fully connects a certain subset of vertices in the graph, called the primary vertices. Power indices measure an agent's ability to affect the outcome of the game. We show that in our domain, such indices can be used to both determine the fair share of the revenues an agent is entitled to, and identify significant possible points of failure affecting the reliability of communication in the network. We show that in general graphs, calculating the Shapley and Banzhaf power indices is #P-complete, but suggest a polynomial algorithm for calculating them in trees. We also investigate finding stable payoff divisions of the revenues in CGs, captured by the game theoretic solution of the core, and its relaxations, the epsilon-core and least core. We show a polynomial algorithm for computing the core of a CG, but show that testing whether an imputation is in the epsilon-core is coNP-complete. Finally, we show that for trees, it is possible to test for epsilon-core imputations in polynomial time.

We examine designs for crowdsourcing contests, where participants compete for rewards given to superior solutions of a task. We theoretically analyze tradeoffs between the expectation and variance of the principal's utility (i.e. the best solution's quality), and empirically test our theoretical predictions using a controlled experiment on Amazon Mechanical Turk. Our evaluation method is also crowdsourcing based and relies on the peer prediction mechanism. Our theoretical analysis shows an expectation-variance tradeoff of the principal's utility in such contests through a Pareto efficient frontier. In particular, we show that the simple contest with 2 authors and the 2-pair contest have good theoretical properties. In contrast, our empirical results show that the 2-pair contest is the superior design among all designs tested, achieving the highest expectation and lowest variance of the principal's utility.

We propose a natural model for agent failures in congestion games. In our model, each of the agents may fail to participate in the game, introducing uncertainty regarding the set of active agents. We examine how such uncertainty may change the Nash equilibria (NE) of the game. We prove that although the perturbed game induced by the failure model is not always a congestion game, it still admits at least one pure Nash equilibrium. Then, we turn to examine the effect of failures on the maximal social cost in any NE of the perturbed game. We show that in the limit case where failure probability is negligible new equilibria never emerge, and that the social cost may decrease but it never increases. For the case of non-negligible failure probabilities, we provide a full characterization of the maximal impact of failures on the social cost under worst-case equilibrium outcomes.

Weighted voting is a classic model of cooperation among agents in decision-making domains. In such games, each player has a weight, and a coalition of players wins the game if its total weight meets or exceeds a given quota. A player's power in such games is usually not directly proportional to his weight, and is measured by a power index, the most prominent among which are the Shapley-Shubik index and the Banzhaf index.In this paper, we investigate by how much a player can change his power, as measured by the Shapley-Shubik index or the Banzhaf index, by means of a false-name manipulation, i.e., splitting his weight among two or more identities. For both indices, we provide upper and lower bounds on the effect of weight-splitting. We then show that checking whether a beneficial split exists is NP-hard, and discuss efficient algorithms for restricted cases of this problem, as well as randomized algorithms for the general case. We also provide an experimental evaluation of these algorithms. Finally, we examine related forms of manipulative behavior, such as annexation, where a player subsumes other players, or merging, where several players unite into one. We characterize the computational complexity of such manipulations and provide limits on their effects. For the Banzhaf index, we describe a new paradox, which we term the Annexation Non-monotonicity Paradox.

We study the computational complexity of the counting version of the Possible-Winner problem for elections. In the Possible-Winner problem we are given a profile of voters, each with a partial preference order, and ask if there are linear extensions of the votes such that a designated candidate wins. We also analyze a special case of Possible-Winner, the Manipulation problem. We provide polynomial-time algorithms for counting manipulations in a class of scoring protocols and in several other voting rules. We show #P-hardness of the counting variant of Possible-Winner for plurality and veto and give a simple yet general and practically useful randomized algorithm for a variant of Possible-Winner for all voting rules for which a winner can be computed in polynomial time.

We consider optimizing the coalition structure in Coalitional Skill Games (CSGs), a succinct representation of coalitional games. In CSGs, the value of a coalition depends on the tasks its members can achieve. The tasks require various skills to complete them, and agents may have different skill sets. The optimal coalition structure is a partition of the agents to coalitions, that maximizes the sum of utilities obtained by the coalitions. We show that CSGs can represent any characteristic function, and consider optimal coalition structure generation in this representation. We provide hardness results, showing that in general CSGs, as well as in very restricted versions of them, computing the optimal coalition structure is hard. On the positive side, we show that the problem can be reformulated as constraint satisfaction on a hyper graph, and present an algorithm that finds the optimal coalition structure in polynomial time for instances with bounded tree-width and number of tasks.