Search

Fair division is typically framed from a centralized perspective. We study a decentralized variant of fair division inspired by the dynamics observed in community-based targeting, mutual aid networks, and community resource management paradigms. We develop an approach for decentralized fair division and compare it with a centralized approach with respect to fairness and social welfare guarantees. In the context of the existing literature, our decentralized model can be viewed as a relaxation of previous models of sequential exchange in light of impossibility results concerning the inability of those models to achieve desirable outcomes. We find that in settings representative of many real world situations, the two models of resource allocation offer contrasting fairness and social welfare guarantees. In particular, we show that under appropriate conditions, our model of decentralized allocation can ensure high-quality allocative decisions in an efficient fashion.

Graph Neural Networks (GNNs) have been a powerful tool for node classification tasks in complex networks. However, their decision-making processes remain a black-box to users, making it challenging to understand the reasoning behind their predictions. Counterfactual explanations (CFE) have shown promise in enhancing the interpretability of machine learning models. Prior approaches to compute CFE for GNNS often are learning-based approaches that require training additional graphs. In this paper, we propose a semivalue-based, non-learning approach to generate CFE for node classification tasks, eliminating the need for any additional training. Our results reveals that computing Banzhaf values requires lower sample complexity in identifying the counterfactual explanations compared to other popular methods such as computing Shapley values. Our empirical evidence indicates computing Banzhaf values can achieve up to a fourfold speed up compared to Shapley values. We also design a thresholding method for computing Banzhaf values and show theoretical and empirical results on its robustness in noisy environments, making it superior to Shapley values. Furthermore, the thresholded Banzhaf values are shown to enhance efficiency without compromising the quality (i.e., fidelity) in the explanations in three popular graph datasets., Comment: Accepted to WWW 2024

We study the design of caching policies in applications such as serverless computing where there is not a fixed size cache to be filled, but rather there is a cost associated with the time an item stays in the cache. We present a model for such caching policies which captures the trade-off between this cost and the cost of cache misses. We characterize optimal caching policies in general and apply this characterization by deriving a closed form for Hawkes processes. Since optimal policies for Hawkes processes depend on the history of arrivals, we also develop history-independent policies which achieve near-optimal average performance. We evaluate the performances of the optimal policy and approximate polices using simulations and a data trace of Azure Functions, Microsoft's FaaS (Function as a Service) platform for serverless computing., Comment: Presented as a Conference Paper at UAI 2023

Reinforcement learning generalizes multi-armed bandit problems with additional difficulties of a longer planning horizon and unknown transition kernel. We explore a black-box reduction from discounted infinite-horizon tabular reinforcement learning to multi-armed bandits, where, specifically, an independent bandit learner is placed in each state. We show that, under ergodicity and fast mixing assumptions, any slowly changing adversarial bandit algorithm achieving optimal regret in the adversarial bandit setting can also attain optimal expected regret in infinite-horizon discounted Markov decision processes, with respect to the number of rounds $T$. Furthermore, we examine our reduction using a specific instance of the exponential-weight algorithm., Comment: ALT 24

Data and algorithms are essential and complementary parts of a large-scale decision-making process. However, their injudicious use can lead to unforeseen consequences, as has been observed by researchers and activists alike in the recent past. In this paper, we revisit the application of predictive models by the Chicago Department of Public Health to schedule restaurant inspections and prioritize the detection of critical food code violations. We perform the first analysis of the model's fairness to the population served by the restaurants in terms of average time to find a critical violation. We find that the model treats inspections unequally based on the sanitarian who conducted the inspection and that, in turn, there are geographic disparities in the benefits of the model. We examine four alternate methods of model training and two alternative ways of scheduling using the model and find that the latter generate more desirable results. The challenges from this application point to important directions for future work around fairness with collective entities rather than individuals, the use of critical violations as a proxy, and the disconnect between fair classification and fairness in the dynamic scheduling system., Comment: 28 pages, 16 figures

Group fairness definitions such as Demographic Parity and Equal Opportunity make assumptions about the underlying decision-problem that restrict them to classification problems. Prior work has translated these definitions to other machine learning environments, such as unsupervised learning and reinforcement learning, by implementing their closest mathematical equivalent. As a result, there are numerous bespoke interpretations of these definitions. Instead, we provide a generalized set of group fairness definitions that unambiguously extend to all machine learning environments while still retaining their original fairness notions. We derive two fairness principles that enable such a generalized framework. First, our framework measures outcomes in terms of utilities, rather than predictions, and does so for both the decision-algorithm and the individual. Second, our framework considers counterfactual outcomes, rather than just observed outcomes, thus preventing loopholes where fairness criteria are satisfied through self-fulfilling prophecies. We provide concrete examples of how our counterfactual utility fairness framework resolves known fairness issues in classification, clustering, and reinforcement learning problems. We also show that many of the bespoke interpretations of Demographic Parity and Equal Opportunity fit nicely as special cases of our framework.

We study the fair division problem of allocating multiple resources among a set of agents with Leontief preferences that are each required to complete a finite amount of work, which we term "limited demands". We examine the behavior of the classic Dominant Resource Fairness (DRF) mechanism in this setting and show it is fair but only weakly Pareto optimal and inefficient in many natural examples. We propose as an alternative the Least Cost Product (LCP) mechanism, a natural adaptation of Maximum Nash Welfare to this setting. We characterize the structure of allocations of the LCP mechanism in this setting, show that it is Pareto efficient, and that it satisfies the relatively weak fairness property of sharing incentives. While we prove it satisfies the stronger fairness property of (expected) envy freeness in some special cases, we provide a counterexample showing it does not do so in general, a striking contrast to the "unreasonable fairness" of Maximum Nash Welfare in other settings. Simulations suggest, however, that these violations of envy freeness are rare in randomly generated examples., Comment: Presented as a Conference Paper at AAAI 2021

We consider a model where an agent has a repeated decision to make and wishes to maximize their total payoff. Payoffs are influenced by an action taken by the agent, but also an unknown state of the world that evolves over time. Before choosing an action each round, the agent can purchase noisy samples about the state of the world. The agent has a budget to spend on these samples, and has flexibility in deciding how to spread that budget across rounds. We investigate the problem of choosing a sampling algorithm that optimizes total expected payoff. For example: is it better to buy samples steadily over time, or to buy samples in batches? We solve for the optimal policy, and show that it is a natural instantiation of the latter. Under a more general model that includes per-round fixed costs, we prove that a variation on this batching policy is a 2-approximation.

Counterfactual Regret Minimization (CFR) has found success in settings like poker which have both terminal states and perfect recall. We seek to understand how to relax these requirements. As a first step, we introduce a simple algorithm, local no-regret learning (LONR), which uses a Q-learning-like update rule to allow learning without terminal states or perfect recall. We prove its convergence for the basic case of MDPs (and limited extensions of them) and present empirical results showing that it achieves last iterate convergence in a number of settings, most notably NoSDE games, a class of Markov games specifically designed to be challenging to learn where no prior algorithm is known to achieve convergence to a stationary equilibrium even on average., Comment: Presented as conference paper at AAMAS 2020

Recent work shows that we can use partial verification instead of money to implement truthful mechanisms. In this paper we develop tools to answer the following question. Given an allocation rule that can be made truthful with payments, what is the minimal verification needed to make it truthful without them? Our techniques leverage the geometric relationship between the type space and the set of possible allocations., Comment: Extended Version of 'Partial Verification as a Substitute for Money', AAAI 2019

Cloud computing providers face the problem of matching heterogeneous customer workloads to resources that will serve them. This is particularly challenging if customers, who are already running a job on a cluster, scale their resource usage up and down over time. The provider therefore has to continuously decide whether she can add additional workloads to a given cluster or if doing so would impact existing workloads' ability to scale. Currently, this is often done using simple threshold policies to reserve large parts of each cluster, which leads to low efficiency (i.e., low average utilization of the cluster). We propose more sophisticated policies for controlling admission to a cluster and demonstrate that they significantly increase cluster utilization. We first introduce the cluster admission problem and formalize it as a constrained Partially Observable Markov Decision Process (POMDP). As it is infeasible to solve the POMDP optimally, we then systematically design admission policies that estimate moments of each workload's distribution of future resource usage. Via extensive simulations grounded in a trace from Microsoft Azure, we show that our admission policies lead to a substantial improvement over the simple threshold policy. We then show that substantial further gains are possible if high-quality information is available about arriving workloads. Based on this, we propose an information elicitation approach to incentivize users to provide this information and simulate its effects., Comment: This paper is a significantly extended version of work that was published as a 6-page extended abstract in the proceedings of the 14th Workshop on the Economics of Networks, Systems and Computation (NetEcon'19)

The problem of pricing the cloud has attracted much recent attention due to the widespread use of cloud computing and cloud services. From a theoretical perspective, several mechanisms that provide strong efficiency or fairness guarantees and desirable incentive properties have been designed. However, these mechanisms often rely on a rigid model, with several parameters needing to be precisely known in order for the guarantees to hold. In this paper, we consider a stochastic model and show that it is possible to obtain good welfare and revenue guarantees with simple mechanisms that do not make use of the information on some of these parameters. In particular, we prove that a mechanism that sets the same price per time step for jobs of any length achieves at least 50% of the welfare and revenue obtained by a mechanism that can set different prices for jobs of different lengths, and the ratio can be improved if we have more specific knowledge of some parameters. Similarly, a mechanism that sets the same price for all servers even though the servers may receive different kinds of jobs can provide a reasonable welfare and revenue approximation compared to a mechanism that is allowed to set different prices for different servers., Comment: To appear in the 13th Conference on Web and Internet Economics (WINE), 2017. A preliminary version was presented at the 12th Workshop on the Economics of Networks, Systems and Computation (NetEcon), 2017

We study the problem of designing optimal auctions under restrictions on the set of permissible allocations. In addition to allowing us to restrict to deterministic mechanisms, we can also indirectly model non-additive valuations. We prove a strong duality result, extending a result due to Daskalakis et al. [2015], that guarantees the existence of a certificate of optimality for optimal restricted mechanisms. As a corollary of our result, we provide a new characterization of the set of allocations that the optimal mechanism may actually use. To illustrate our result we find and certify optimal mechanisms for four settings where previous frameworks do not apply, and provide new economic intuition about some of the tools that have previously been used to find optimal mechanisms.

We consider the problem of detecting norm violations in open multi-agent systems (MAS). We show how, using ideas from scrip systems, we can design mechanisms where the agents comprising the MAS are incentivised to monitor the actions of other agents for norm violations. The cost of providing the incentives is not borne by the MAS and does not come from fines charged for norm violations (fines may be impossible to levy in a system where agents are free to leave and rejoin again under a different identity). Instead, monitoring incentives come from (scrip) fees for accessing the services provided by the MAS. In some cases, perfect monitoring (and hence enforcement) can be achieved: no norms will be violated in equilibrium. In other cases, we show that, while it is impossible to achieve perfect enforcement, we can get arbitrarily close; we can make the probability of a norm violation in equilibrium arbitrarily small. We show using simulations that our theoretical results hold for multi-agent systems with as few as 1000 agents---the system rapidly converges to the steady-state distribution of scrip tokens necessary to ensure monitoring and then remains close to the steady state., Comment: 24 pages, 7 pdf figures

A property, or statistical functional, is said to be elicitable if it minimizes expected loss for some loss function. The study of which properties are elicitable sheds light on the capabilities and limitations of point estimation and empirical risk minimization. While recent work asks which properties are elicitable, we instead advocate for a more nuanced question: how many dimensions are required to indirectly elicit a given property? This number is called the elicitation complexity of the property. We lay the foundation for a general theory of elicitation complexity, including several basic results about how elicitation complexity behaves, and the complexity of standard properties of interest. Building on this foundation, our main result gives tight complexity bounds for the broad class of Bayes risks. We apply these results to several properties of interest, including variance, entropy, norms, and several classes of financial risk measures. We conclude with discussion and open directions., Comment: This version fixes an error in the condition needed for the main lower bound and adds an application to Range Value at Risk, along with a substantial reorganization of the paper and numerous smaller changes. A previous version appeared in Neural Information Processing Systems 2015

We study the problem of eliciting and aggregating probabilistic information from multiple agents. In order to successfully aggregate the predictions of agents, the principal needs to elicit some notion of confidence from agents, capturing how much experience or knowledge led to their predictions. To formalize this, we consider a principal who wishes to elicit predictions about a random variable from a group of Bayesian agents, each of whom have privately observed some independent samples of the random variable, and hopes to aggregate the predictions as if she had directly observed the samples of all agents. Leveraging techniques from Bayesian statistics, we represent confidence as the number of samples an agent has observed, which is quantified by a hyperparameter from a conjugate family of prior distributions. This then allows us to show that if the principal has access to a few samples, she can achieve her aggregation goal by eliciting predictions from agents using proper scoring rules. In particular, if she has access to one sample, she can successfully aggregate the agents' predictions if and only if every posterior predictive distribution corresponds to a unique value of the hyperparameter. Furthermore, this uniqueness holds for many common distributions of interest. When this uniqueness property does not hold, we construct a novel and intuitive mechanism where a principal with two samples can elicit and optimally aggregate the agents' predictions.

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. We use simulations and analysis of real-world sponsored search auction data to demonstrate the feasible trade-offs, examining the effect of changing the allowed number of ads on the utilities of the stakeholders. We investigate both short term effects, when the players do not have the time to modify their behavior, and long term equilibrium conditions. Finally, we examine a combinatorially richer constrained optimization problem, where there are several possible allowed configurations (templates) of ad formats. This model captures richer ad formats, which allow using the available screen real estate in various ways. We show that two natural generalizations of the GSP auction rules to this domain are poorly behaved, resulting in not having a symmetric Nash equilibrium or having one with poor welfare. We also provide positive results for restricted cases., Comment: 18 pages, 10 figures, ACM Conference on Economics and Computation 2014

In a sponsored search auction, decisions about how to rank ads impose tradeoffs between objectives such as revenue and welfare. In this paper, we examine how these tradeoffs should be made. We begin by arguing that the most natural solution concept to evaluate these tradeoffs is the lowest symmetric Nash equilibrium (SNE). As part of this argument, we generalise the well known connection between the lowest SNE and the VCG outcome. We then propose a new ranking algorithm, loosely based on the revenue-optimal auction, that uses a reserve price to order the ads (not just to filter them) and give conditions under which it raises more revenue than simply applying that reserve price. Finally, we conduct extensive simulations examining the tradeoffs enabled by different ranking algorithms and show that our proposed algorithm enables superior operating points by a variety of metrics., Comment: To appear in Proceedings of the 14th ACM Conference on Electronic Commerce (EC '13)

We present a model of truthful elicitation which generalizes and extends mechanisms, scoring rules, and a number of related settings that do not qualify as one or the other. Our main result is a characterization theorem, yielding characterizations for all of these settings, including a new characterization of scoring rules for non-convex sets of distributions. We generalize this model to eliciting some property of the agent's private information, and provide the first general characterization for this setting. We combine this characterization with duality to give a simple construction to convert between scoring rules and randomized mechanisms. We also show how this characterization gives a new proof of a mechanism design result due to Saks and Yu., Comment: Substantially revised; 47 pages

Scrip, or artificial currency, is a useful tool for designing systems that are robust to selfish behavior by users. However, it also introduces problems for a system designer, such as how the amount of money in the system should be set. In this paper, the effect of varying the total amount of money in a scrip system on efficiency (i.e., social welfare---the total utility of all the agents in the system) is analyzed, and it is shown that by maintaining the appropriate ratio between the total amount of money and the number of agents, efficiency is maximized. This ratio can be found by increasing the money supply to just below the point that the system would experience a "monetary crash," where money is sufficiently devalued that no agent is willing to perform a service. The implications of the presence of altruists, hoarders, sybils, and collusion on the performance of the system are examined. Approaches are discussed to identify the strategies and types of agents., Comment: 24 pages. Forthcoming in Distributed Computing. Premliminary versions of this material appeared in the Proceedings of the 7th and 8th ACM Conferences on Electronic Commerce [14,26] and the Proceedings of the First Conference on Auctions, Market Mechanisms and Their Applications [27]. These are also available as arXiv:0705.4094, arXiv:0705.4110, and arXiv:0903.2278

A game-theoretic model of scrip (artificial currency) systems is analyzed. It is shown that relative entropy can be used to characterize the distribution of agent wealth when all agents use threshold strategies---that is, they volunteer to do work iff they have below a threshold amount of money. Monotonicity of agents' best-reply functions is used to show that scrip systems have pure strategy equilibria where all agents use threshold strategies. An algorithm is given that can compute such an equilibrium and the resulting distribution of wealth., Comment: 32 pages. Forthcoming in ACM Transactions on Economics and Computation. Preliminary versions of this material appeared in the Proceedings of the 7th and 8th ACM Conferences on Electronic Commerce [14, 26] and the Proceedings of the First Conference on Auctions, Market Mechanisms and Their Applications [27]. These are also available as arXiv:0705.4094, arXiv:0705.4110, and arXiv:0903.2278

We study a model of congestible resources, where pricing and scheduling are intertwined. Motivated by the problem of pricing cloud instances, we model a cloud computing service as linked $GI/GI/\cdot$ queuing systems where the provider chooses to offer a fixed pricing service, a dynamic market based service, or a hybrid of both, where jobs can be preempted in the market-based service. Users (jobs), who are heterogeneous in both the value they place on service and their cost for waiting, then choose between the services offered. Combining insights from auction theory with queuing theory we are able to characterize user equilibrium behavior, and show its insensitivity to the precise market design mechanism used. We then provide theoretical and simulation based evidence suggesting that a fixed price typically, though not always, generates a higher expected revenue than the hybrid system for the provider., Comment: 26 pages and 7 figures. An early version of this work appeared in the 7th Workshop on the Economics of Networks, Systems, and Computation (NetEcon) 2012

Recent work has constructed economic mechanisms that are both truthful and differentially private. In these mechanisms, privacy is treated separately from the truthfulness; it is not incorporated in players' utility functions (and doing so has been shown to lead to non-truthfulness in some cases). In this work, we propose a new, general way of modelling privacy in players' utility functions. Specifically, we only assume that if an outcome $o$ has the property that any report of player $i$ would have led to $o$ with approximately the same probability, then $o$ has small privacy cost to player $i$. We give three mechanisms that are truthful with respect to our modelling of privacy: for an election between two candidates, for a discrete version of the facility location problem, and for a general social choice problem with discrete utilities (via a VCG-like mechanism). As the number $n$ of players increases, the social welfare achieved by our mechanisms approaches optimal (as a fraction of $n$).

We consider the zero-error capacity of deletion channels. Specifically, we consider the setting where we choose a codebook ${\cal C}$ consisting of strings of $n$ bits, and our model of the channel corresponds to an adversary who may delete up to $pn$ of these bits for a constant $p$. Our goal is to decode correctly without error regardless of the actions of the adversary. We consider what values of $p$ allow non-zero capacity in this setting. We suggest multiple approaches, one of which makes use of the natural connection between this problem and the problem of finding the expected length of the longest common subsequence of two random sequences., Comment: 9 pages, 0 figures

Consider a matching problem on a graph where disjoint sets of vertices are privately owned by self-interested agents. An edge between a pair of vertices indicates compatibility and allows the vertices to match. We seek a mechanism to maximize the number of matches despite self-interest, with agents that each want to maximize the number of their own vertices that match. Each agent can choose to hide some of its vertices, and then privately match the hidden vertices with any of its own vertices that go unmatched by the mechanism. A prominent application of this model is to kidney exchange, where agents correspond to hospitals and vertices to donor-patient pairs. Here hospitals may game an exchange by holding back pairs and harm social welfare. In this paper we seek to design mechanisms that are strategyproof, in the sense that agents cannot benefit from hiding vertices, and approximately maximize efficiency, i.e., produce a matching that is close in cardinality to the maximum cardinality matching. Our main result is the design and analysis of the eponymous Mix-and-Match mechanism; we show that this randomized mechanism is strategyproof and provides a 2-approximation. Lower bounds establish that the mechanism is near optimal., Comment: 10 pages, 5 figures. Appeared in Proceedings of 11th ACM Conference on Electronic Commerce, Pages 305-314

In large systems, it is important for agents to learn to act effectively, but sophisticated multi-agent learning algorithms generally do not scale. An alternative approach is to find restricted classes of games where simple, efficient algorithms converge. It is shown that stage learning efficiently converges to Nash equilibria in large anonymous games if best-reply dynamics converge. Two features are identified that improve convergence. First, rather than making learning more difficult, more agents are actually beneficial in many settings. Second, providing agents with statistical information about the behavior of others can significantly reduce the number of observations needed., Comment: 8 pages, 2 figures. To Appear in Proceedings of the Eighth International Conference on Autonomous Agents and Multiagent Systems

Game-theoretic analyses of distributed and peer-to-peer systems typically use the Nash equilibrium solution concept, but this explicitly excludes the possibility of strategic behavior involving more than one agent. We examine the effects of two types of strategic behavior involving more than one agent, sybils and collusion, in the context of scrip systems where agents provide each other with service in exchange for scrip. Sybils make an agent more likely to be chosen to provide service, which generally makes it harder for agents without sybils to earn money and decreases social welfare. Surprisingly, in certain circumstances it is possible for sybils to make all agents better off. While collusion is generally bad, in the context of scrip systems it actually tends to make all agents better off, not merely those who collude. These results also provide insight into the effects of allowing agents to advertise and loan money. While many extensions of Nash equilibrium have been proposed that address collusion and other issues relevant to distributed and peer-to-peer systems, our results show that none of them adequately address the issues raised by sybils and collusion in scrip systems., Comment: 20 pages, 5 figures. To appear in the Proceedings of The First Conference on Auctions, Market Mechanisms and Their Applications (AMMA '09)

Many protocols for distributed and peer-to-peer systems have the feature that nodes will stop providing service for others once they have received a certain amount of service. Examples include BitTorent's unchoking policy, BAR Gossip's balanced exchanges, and threshold strategies in scrip systems. An attacker can exploit this by providing service in a targeted way to prevent chosen nodes from providing service. While such attacks cannot be prevented, we discuss techniques that can be used to limit the damage they do. These techniques presume that a certain number of processes will follow the recommended protocol, even if they could do better by ``gaming'' the system.

We discuss the design of efficient scrip systems and develop tools for empirically analyzing them. For those interested in the empirical study of scrip systems, we demonstrate how characteristics of agents in a system can be inferred from the equilibrium distribution of money. From the perspective of a system designer, we examine the effect of the money supply on social welfare and show that social welfare is maximized by increasing the money supply up to the point that the system experiences a ``monetary crash,'' where money is sufficiently devalued that no agent is willing to perform a service. We also examine the implications of the presence of altruists and hoarders on the performance of the system. While a small number of altruists may improve social welfare, too many can also cause the system to experience a monetary crash, which may be bad for social welfare. Hoarders generally decrease social welfare but, surprisingly, they also promote system stability by helping prevent monetary crashes. In addition, we provide new technical tools for analyzing and computing equilibria by showing that our model exhibits strategic complementarities, which implies that there exist equilibria in pure strategies that can be computed efficiently.

A model of providing service in a P2P network is analyzed. It is shown that by adding a scrip system, a mechanism that admits a reasonable Nash equilibrium that reduces free riding can be obtained. The effect of varying the total amount of money (scrip) in the system on efficiency (i.e., social welfare) is analyzed, and it is shown that by maintaining the appropriate ratio between the total amount of money and the number of agents, efficiency is maximized. The work has implications for many online systems, not only P2P networks but also a wide variety of online forums for which scrip systems are popular, but formal analyses have been lacking.