Peer-evaluation and selection systems are used when sets of agents evaluate each other in order to select the best k among them. These are commonly used in real-world settings, including academic conferences where those reviewing papers are often the set of submitters. Conferences have attempted to better allocate their reviewing resources by moving to a two-stage mechanism, in which some papers are eliminated after a first stage of review and remaining papers receive additional reviewers. We investigate how two major strategyproof peer selection mechanisms, Partition and ExactDollarPartition, perform when adapted to a two-stage system, in order to try and understand the effect of the two-stage mechanism on which agents get selected. We also examine how the various parameters of the two-stage mechanism influence the outcome. We provide a theoretical basis by showing how a particular setting is influenced by the two stages. However, solving for the general case seems implausible at the moment, and we use extensive simulations of different scenarios and settings to observe which agents benefit and which are harmed by adopting two-stage mechanisms (and we vary this mechanisms parameters as well). We show that the two-stage mechanism's advantage depends the noisiness of reviewer beliefs. Borderline agents benefit most in a low noise environment, while high rank agents benefit more in noisy environments. We show that the effectiveness of these mechanisms is highly dependent on the number of chosen agents, the number of reviews requested from agents, and reviewers' correlation, indicating that organizers need to exercise caution when selecting these parameters for a reviewing process.

Fairness in machine learning is often conceptualized narrowly in comparative, distributional terms. In studying stakeholders' concepts of fairness, we find that this framing is insufficient to capture the full range of issues raised. As an alternative, we propose organizational justice as a framework that subsumes distributional fairness as well as other normative concerns. We conduct a case study of organizational justice relative to personalized recommendation in the context of Kiva Microfunds, a nonprofit micro-lending organization whose mission is to increase financial access for underserved communities across the world. We report on the results of co-design workshops conducted with Kiva employees who are involved in different departments and whose roles often lead them to prioritize normative concerns that are most supportive of the stakeholders with whom they work most closely. We apply organizational justice to understand design trade-offs among different normative goals stakeholders invoke. Based on these goals, we identify a suite of metrics that Kiva employees can use to monitor and assess the recommender system's impact on their organizational justice concerns and to seed discussions within the organization about appropriate configuration and deployment of this system in context.

Computational social choice (COMSOC) studies principled ways to aggregate conflicting individual preferences into collective decisions. In this paper, we call for an increased effort towards Computational Social Choice: Research Development (COMSOC-R D), a problem-driven research agenda that explicitly aims to design, implement, and test collective decision-making systems in the real world. We articulate the defining features of COMSOC-R D, argue for its value, and discuss various roadblocks and possible solutions.

We propose a multi-agent multi-armed bandit (MA-MAB) framework to ensure fair outcomes across agents while maximizing overall system performance. For example, in a ridesharing setting where a central dispatcher assigns drivers to distinct geographic regions, utilitarian welfare (the sum of driver earnings) can be highly skewed—some drivers may receive no rides. We instead measure fairness by Nash social welfare, i.e., the product of individual rewards. A key challenge in this setting is decision-making under limited information about arm rewards (geographic regions). To address this, we introduce a novel probing mechanism that strategically gathers information about selected arms before assignment. In the offline setting, where reward distributions are known, we exploit submodularity to design a greedy probing algorithm with a constant-factor approximation guarantee. In the online setting, we develop a probing-based algorithm that achieves sublinear regret while preserving Nash social welfare. Extensive experiments on synthetic and real-world datasets demonstrate that our approach outperforms baseline methods in both fairness and efficiency.

Peer selection, the evaluation and selection of agents by their peers, is an important problem in the field of computational social choice; with applications to grading in massively online courses (MOOCs) and academic peer review. Current existing algorithmic and empirical work focuses on developing and analyzing novel strategyproof mechanisms, wherein no agent has an incentive to misreport their evaluations. However, the majority of published mechanisms share a flaw: they do not reward agents for any effort expended during the evaluation process. In cases where high quality evaluations are costly to produce this missing incentive fails to align agents with an overall goal of accurate selection. To address this gap we first prove theoretically that incentivizing effort in peer selection requires information beyond a single evaluation. We then propose PeerBTS, a mechanism that combines a peer-prediction lottery, leveraging work on the Robust Bayesian Truth Serum, with any existing peer-selection mechanism to incentivize effort while remaining Bayes-Nash incentive compatible. We find that while an incentive-compatible peer-selection mechanism using agent predictions to incentivize effort is possible it requires adjustments to the assumed problem context and limits other mechanistics properties. We additionally present a series of non-strategic simulations to validate incentives and evaluate the performance of PeerBTS relative to existing strategyproof peer selection mechanisms. Finally, we discuss the results of an initial study on the validity of peer-prediction from a small academic workshop.

We introduce a novel collective decision making problem that captures the ubiquitous issue of ordering food to cater for varied dietary preferences and requirements. Our settings involve agents with diverse dietary requirements over menu options with varied serving sizes. The goal is to select a menu where everyone has enough food they can consume and wastage of food is minimized. We introduce two different consumption models: optimistic and pessimistic. Optimistic consumption assumes a situation when a central planner can optimally allocate the food ordered among the agents to maximize the number of people who get enough to eat. Pessimistic considers the worst case guarantee on consumption when agents fill their own plates in an arbitrary order. Under either consumption model, we seek valid menus (under which all agents are sufficiently fed) of minimum size. Our work provides two sets of characterizations: (1) we characterize valid menus under either consumption model and (2) we characterize the space of instances that admit polynomial-time algorithms to find minimum sized menus. Our results also help us design Integer Linear Programs to find minimum sized menus in general settings. Furthermore, we present polynomial-time algorithms for important special cases. We then consider the worst case discrepancy between the size of minimum sized optimistic and pessimistic menus. We call this the waste of pessimism, captured by the ratio of the minimum sized pessimistic menu to that of the minimum sized optimistic menu. We show tight upper bounds on this ratio. Our results also provide additional insights on the problem of finding a minimum sized maximal matching, which may be of independent interest.

Online reviews provide valuable insights into the perceived quality of facets of a product or service. While aspect-based sentiment analysis has focused on extracting these facets from reviews, there is less work understanding the impact of each aspect on overall perception. This is particularly challenging given correlations among aspects, making it difficult to isolate the effects of each. This paper introduces a methodology based on recent advances in text-based causal analysis, specifically CausalBERT, to disentangle the effect of each factor on overall review ratings. We enhance CausalBERT with three key improvements: temperature scaling for better calibrated treatment assignment estimates; hyperparameter optimization to reduce confound overadjustment; and interpretability methods to characterize discovered confounds. In this work, we treat the textual mentions in reviews as proxies for real-world attributes. We validate our approach on real and semisynthetic data from over 600K reviews of U.S. K-12 schools. We find that the proposed enhancements result in more reliable estimates, and that perception of school administration and performance on benchmarks are significant drivers of overall school ratings.

When working to build machines that have a form of intelligence, it is natural to be inspired by human intelligence. Of course, humans are very different from machines, in their embodiment and myriad other ways. Humans exploit their bodies to experience the world, create an internal model of it, and use this model to reason, learn, and make contextual and informed decisions. Machines lack the same embodiment, but often have access to both more memory and more computing power. Despite these crucial disanalogies, it is still useful to leverage our knowledge of how the human mind reasons and makes decisions to design and build machines that demonstrate behaviors similar to that of a human. In this article, we present a novel AI architecture, Slow and Fast AI (SOFAI), that is inspired by the “thinking fast and slow” cognitive theory of human decision making. SOFAI is a multi-agent architecture that employs both “fast” and “slow” solvers underneath a metacognitive agent that is able to both choose among a set of solvers as well as reflect on and learn from past experience. Experimental results on the behavior of two instances of the SOFAI architecture show that, compared to using just one of the two decision modalities, SOFAI is markedly better in terms of decision quality, resource consumption, and efficiency.

We consider the problem of determining a binary ground truth using advice from a group of independent reviewers (experts) who express their guess about a ground truth correctly with some independent probability (competence). In this setting, when all reviewers are competent (competence greater than one-half), the Condorcet Jury Theorem tells us that adding more reviewers increases the overall accuracy, and if all competences are known, then there exists an optimal weighting of the reviewers. However, in practical settings, reviewers may be noisy or incompetent, i.e., competence below half, and the number of experts may be small, so the asymptotic Condorcet Jury Theorem is not practically relevant. In such cases we explore appointing one or more chairs (judges) who determine the weight of each reviewer for aggregation, creating multiple levels. However, these chairs may be unable to correctly identify the competence of the reviewers they oversee, and therefore unable to compute the optimal weighting. We give conditions when a set of chairs is able to weight the reviewers optimally, and depending on the competence distribution of the agents, give results about when it is better to have more chairs or more reviewers. Through numerical simulations we show that in some cases it is better to have more chairs, but in many cases it is better to have more reviewers.

Fairness in recommender systems is a complex concept, involving multiple definitions, different parties for whom fairness is sought, and various scopes over which fairness might be measured. Researchers seeking fairness-aware systems have derived a variety of solutions, usually highly tailored to specific choices along each of these dimensions, and typically aimed at tackling a single fairness concern, i.e., a single definition for a specific stakeholder group and measurement scope. However, in practical contexts, there are a multiplicity of fairness concerns within a given recommendation application and solutions limited to a single dimension are therefore less useful. We explore a general solution to recommender system fairness using social choice methods to integrate multiple heterogeneous definitions. In this paper, we extend group-fairness results from prior research to provider-side individual fairness, demonstrating in multiple datasets that both individual and group fairness objectives can be integrated and optimized jointly. We identify both synergies and tensions among different objectives with individual fairness correlated with group fairness for some groups and anti-correlated with others.

Artificial intelligence systems, especially those using machine learning, are being deployed in domains from hiring to loan issuance in order to automate these complex decisions. Judging both the effectiveness and fairness of these AI systems, and their human decision making counterpart, is a complex and important topic studied across both computational and social sciences. Within machine learning, a common way to address bias in downstream classifiers is to resample the training data to offset disparities. For example, if hiring rates vary by some protected class, then one may equalize the rate within the training set to alleviate bias in the resulting classifier. While simple and seemingly effective, these methods have typically only been evaluated using data obtained through convenience samples, introducing selection bias and label bias into metrics. Within the social sciences, psychology, public health, and medicine, audit studies, in which fictitious “testers” (e.g., resumes, emails, patient actors) are sent to subjects (e.g., job openings, businesses, doctors) in randomized control trials, provide high quality data that support rigorous estimates of discrimination. In this paper, we investigate how data from audit studies can be used to improve our ability to both train and evaluate automated hiring algorithms. We find that such data reveals cases where the common fairness intervention method of equalizing base rates across classes appears to achieve parity using traditional measures, but in fact has roughly 10

People care about decision outcomes and how decisions get made, both when making decisions and reflecting on decisions. But formalizing the full range of normative concerns that drive decisions is an open challenge. We introduce Axiomatic Choice as a framework for making and evaluating decisions based on formal normative statements about decisions. These statements, or axioms, capture a wide array of desiderata, e.g., ethical constraints, beyond the typical treatment in Social Choice. Using our model of axioms and decisions we define key properties and introduce a taxonomy of axioms which may be of general interest. We then use these properties and our taxonomy to define the Decision-Evaluation Paradox, formalize the concepts of transparency and deception in explaining and justifying decisions, and reveal the limits of existing methods using axioms to make decisions.

There is a critical need for community engagement in the process of adopting artificial intelligence (AI) technologies in public health. Public health practitioners and researchers have historically innovated in areas like vaccination and sanitation but have been slower in adopting emerging technologies such as generative AI. However, with increasingly complex funding, programming, and research requirements, the field now faces a pivotal moment to enhance its agility and responsiveness to evolving health challenges. Participatory methods and community engagement are key components of many current public health programs and research. The field of public health is well positioned to ensure community engagement is part of AI technologies applied to population health issues. Without such engagement, the adoption of these technologies in public health may exclude significant portions of the population, particularly those with the fewest resources, with the potential to exacerbate health inequities. Risks to privacy and perpetuation of bias are more likely to be avoided if AI technologies in public health are designed with knowledge of community engagement, existing health disparities, and strategies for improving equity. This viewpoint proposes a multifaceted approach to ensure safer and more effective integration of AI in public health with the following call to action: (1) include the basics of AI technology in public health training and professional development; (2) use a community engagement approach to co-design AI technologies in public health; and (3) introduce governance and best practice mechanisms that can guide the use of AI in public health to prevent or mitigate potential harms. These actions will support the application of AI to varied public health domains through a framework for more transparent, responsive, and equitable use of this evolving technology, augmenting the work of public health practitioners and researchers to improve health outcomes while minimizing risks and unintended consequences.

Envy-freeness and the relaxation to Envy-freeness up to one item (EF-1) have been used as fairness concepts in the economics, game theory, and social choice literatures since the 1960s, and have recently gained popularity within the recommendation systems communities. In this short position paper we will give an overview of envy-freeness and its use in economics and recommendation systems; and illustrate why envy is not appropriate to measure fairness for use in settings where personalization plays a role.

Algorithmic fairness in the context of personalized recommendation presents significantly different challenges to those commonly encountered in classification tasks. Researchers studying classification have generally considered fairness to be a matter of achieving equality of outcomes between a protected and unprotected group, and built algorithmic interventions on this basis. We argue that fairness in real-world application settings in general, and especially in the context of personalized recommendation, is much more complex and multi-faceted, requiring a more general approach. We propose a model to formalize multistakeholder fairness in recommender systems as a two stage social choice problem. In particular, we express recommendation fairness as a novel combination of an allocation and an aggregation problem, which integrate both fairness concerns and personalized recommendation provisions, and derive new recommendation techniques based on this formulation. Simulations demonstrate the ability of the framework to integrate multiple fairness concerns in a dynamic way.

Large Language Models (LLMs) have become increasingly incorporated into everyday life for many internet users, taking on significant roles as advice givers in the domains of medicine, personal relationships, and even legal matters. The importance of these roles raise questions about how and what responses LLMs make in difficult political and moral domains, especially questions about possible biases. To quantify the nature of potential biases in LLMs, various works have applied Moral Foundations Theory (MFT), a framework that categorizes human moral reasoning into five dimensions: Harm, Fairness, Ingroup Loyalty, Authority, and Purity. Previous research has used the MFT to measure differences in human participants along political, national, and cultural lines. While there has been some analysis of the responses of LLM with respect to political stance in role-playing scenarios, no work so far has directly assessed the moral leanings in the LLM responses, nor have they connected LLM outputs with robust human data. In this paper we analyze the distinctions between LLM MFT responses and existing human research directly, investigating whether commonly available LLM responses demonstrate ideological leanings: either through their inherent responses, straightforward representations of political ideologies, or when responding from the perspectives of constructed human personas. We assess whether LLMs inherently generate responses that align more closely with one political ideology over another, and additionally examine how accurately LLMs can represent ideological perspectives through both explicit prompting and demographic-based role-playing. By systematically analyzing LLM behavior across these conditions and experiments, our study provides insight into the extent of political and demographic dependency in AI-generated responses.

The study of peer selection mechanisms presents a unique opportunity to understand and improve the practice of a group selecting its best members, despite each member of that group wanting to be selected. A prime example of such a setting is academic peer review, for which peer selection offers a variety of improvement directions. We present an annotated reading list covering the foundations of peer selection as well as recent and emerging work within the field.

Synthetic data is a useful resource for algorithmic research. It allows for the evaluation of systems under a range of conditions that might be difficult to achieve in real world settings. In recommender systems, the use of synthetic data is somewhat limited; some work has concentrated on building user-item interaction data at large scale. We believe that fairness-aware recommendation research can benefit from simulated data as it allows the study of protected groups and their interactions without depending on sensitive data that needs privacy protection. In this paper, we propose a novel type of data for fairness-aware recommendation: synthetic recommender system outputs that can be used to study re-ranking algorithms.

Constraining the actions of AI systems is one promising way to ensure that these systems behave in a way that is morally acceptable to humans. But constraints alone come with drawbacks as in many AI systems, they are not flexible. If these constraints are too rigid, they can preclude actions that are actually acceptable in certain, contextual situations. Humans, on the other hand, can often decide when a simple and seemingly inflexible rule should actually be overridden based on the context. In this paper, we empirically investigate the way humans make these contextual moral judgements, with the goal of building AI systems that understand when to follow and when to override constraints. We propose a novel and general preference-based graphical model that captures a modification of standard dual process theories of moral judgment. We then detail the design, implementation, and results of a study of human participants who judge whether it is acceptable to break a well-established rule: no cutting in line. We then develop an instance of our model and compare its performance to that of standard machine learning approaches on the task of predicting the behavior of human participants in the study, showing that our preference-based approach more accurately captures the judgments of human decision-makers. It also provides a flexible method to model the relationship between variables for moral decision-making tasks that can be generalized to other settings.

Algorithmic fairness in recommender systems requires close attention to the needs of a diverse set of stakeholders that may have competing interests. Previous work in this area has often been limited by fixed, single-objective definitions of fairness, built into algorithms or optimization criteria that are applied to a single fairness dimension or, at most, applied identically across dimensions. These narrow conceptualizations limit the ability to adapt fairness-aware solutions to the wide range of stakeholder needs and fairness definitions that arise in practice. Our work approaches recommendation fairness from the standpoint of computational social choice, using a multi-agent framework. In this paper, we explore the properties of different social choice mechanisms and demonstrate the successful integration of multiple, heterogeneous fairness definitions across multiple data sets.