Search

The release of ChatGPT in November 2022 sparked an explosion of interest in post-training and an avalanche of new preference optimization (PO) methods. These methods claim superior alignment by virtue of better correspondence with human pairwise preferences, often measured by LLM-judges. In this work, we attempt to answer the following question -- do LLM-judge preferences translate to progress on other, more concrete metrics for alignment, and if not, why not? We define a concrete metric for alignment, and introduce SOS-Bench (Substance Outweighs Style Benchmark), which is to the best of our knowledge the largest standardized, reproducible LLM meta-benchmark to date. We find that (1) LLM-judge preferences do not correlate with concrete measures of safety, world knowledge, and instruction following; (2) LLM-judges have powerful implicit biases, prioritizing style over factuality and safety; and (3) the supervised fine-tuning (SFT) stage of post-training, and not the PO stage, has the greatest impact on alignment, with data scaling and prompt diversity as the driving factors. Our codebase and complete results can be found at https://github.com/penfever/sos-bench.

Clustering is a fundamental problem in machine learning and operations research. Therefore, given the fact that fairness considerations have become of paramount importance in algorithm design, fairness in clustering has received significant attention from the research community. The literature on fair clustering has resulted in a collection of interesting fairness notions and elaborate algorithms. In this paper, we take a critical view of fair clustering, identifying a collection of ignored issues such as the lack of a clear utility characterization and the difficulty in accounting for the downstream effects of a fair clustering algorithm in machine learning settings. In some cases, we demonstrate examples where the application of a fair clustering algorithm can have significant negative impacts on social welfare. We end by identifying a collection of steps that would lead towards more impactful research in fair clustering.

We study the canonical fair clustering problem where each cluster is constrained to have close to population-level representation of each group. Despite significant attention, the salient issue of having incomplete knowledge about the group membership of each point has been superficially addressed. In this paper, we consider a setting where the assigned group memberships are noisy. We introduce a simple noise model that requires a small number of parameters to be given by the decision maker. We then present an algorithm for fair clustering with provable \emph{robustness} guarantees. Our framework enables the decision maker to trade off between the robustness and the clustering quality. Unlike previous work, our algorithms are backed by worst-case theoretical guarantees. Finally, we empirically verify the performance of our algorithm on real world datasets and show its superior performance over existing baselines.

Despite the advanced capabilities of contemporary machine learning (ML) models, they remain vulnerable to adversarial and backdoor attacks. This vulnerability is particularly concerning in real-world deployments, where compromised models may exhibit unpredictable behavior in critical scenarios. Such risks are heightened by the prevalent practice of collecting massive, internet-sourced datasets for pre-training multimodal models, as these datasets may harbor backdoors. Various techniques have been proposed to mitigate the effects of backdooring in these models such as CleanCLIP which is the current state-of-the-art approach. In this work, we demonstrate that the efficacy of CleanCLIP in mitigating backdoors is highly dependent on the particular objective used during model pre-training. We observe that stronger pre-training objectives correlate with harder to remove backdoors behaviors. We show this by training multimodal models on two large datasets consisting of 3 million (CC3M) and 6 million (CC6M) datapoints, under various pre-training objectives, followed by poison removal using CleanCLIP. We find that CleanCLIP is ineffective when stronger pre-training objectives are used, even with extensive hyperparameter tuning. Our findings underscore critical considerations for ML practitioners who pre-train models using large-scale web-curated data and are concerned about potential backdoor threats. Notably, our results suggest that simpler pre-training objectives are more amenable to effective backdoor removal. This insight is pivotal for practitioners seeking to balance the trade-offs between using stronger pre-training objectives and security against backdoor attacks., Comment: Accepted for oral presentation at BUGS workshop @ NeurIPS 2023 (https://neurips2023-bugs.github.io/)

Most reinforcement learning methods rely heavily on dense, well-normalized environment rewards. DreamerV3 recently introduced a model-based method with a number of tricks that mitigate these limitations, achieving state-of-the-art on a wide range of benchmarks with a single set of hyperparameters. This result sparked discussion about the generality of the tricks, since they appear to be applicable to other reinforcement learning algorithms. Our work applies DreamerV3's tricks to PPO and is the first such empirical study outside of the original work. Surprisingly, we find that the tricks presented do not transfer as general improvements to PPO. We use a high quality PPO reference implementation and present extensive ablation studies totaling over 10,000 A100 hours on the Arcade Learning Environment and the DeepMind Control Suite. Though our experiments demonstrate that these tricks do not generally outperform PPO, we identify cases where they succeed and offer insight into the relationship between the implementation tricks. In particular, PPO with these tricks performs comparably to PPO on Atari games with reward clipping and significantly outperforms PPO without reward clipping., Comment: Accepted to NeurIPS 2023

Recommender systems play an essential role in the choices people make in domains such as entertainment, shopping, food, news, employment, and education. The machine learning models underlying these recommender systems are often enormously large and black-box in nature for users, content providers, and system developers alike. It is often crucial for all stakeholders to understand the model's rationale behind making certain predictions and recommendations. This is especially true for the content providers whose livelihoods depend on the recommender system. Drawing motivation from the practitioners' need, in this work, we propose a recourse framework for recommender systems, targeted towards the content providers. Algorithmic recourse in the recommendation setting is a set of actions that, if executed, would modify the recommendations (or ranking) of an item in the desired manner. A recourse suggests actions of the form: "if a feature changes X to Y, then the ranking of that item for a set of users will change to Z." Furthermore, we demonstrate that RecRec is highly effective in generating valid, sparse, and actionable recourses through an empirical evaluation of recommender systems trained on three real-world datasets. To the best of our knowledge, this work is the first to conceptualize and empirically test a generalized framework for generating recourses for recommender systems., Comment: Accepted as a short paper at CIKM 2023

Representations learned by pre-training a neural network on a large dataset are increasingly used successfully to perform a variety of downstream tasks. In this work, we take a closer look at how features are encoded in such pre-trained representations. We find that learned representations in a given layer exhibit a degree of diffuse redundancy, ie, any randomly chosen subset of neurons in the layer that is larger than a threshold size shares a large degree of similarity with the full layer and is able to perform similarly as the whole layer on a variety of downstream tasks. For example, a linear probe trained on $20\%$ of randomly picked neurons from the penultimate layer of a ResNet50 pre-trained on ImageNet1k achieves an accuracy within $5\%$ of a linear probe trained on the full layer of neurons for downstream CIFAR10 classification. We conduct experiments on different neural architectures (including CNNs and Transformers) pre-trained on both ImageNet1k and ImageNet21k and evaluate a variety of downstream tasks taken from the VTAB benchmark. We find that the loss and dataset used during pre-training largely govern the degree of diffuse redundancy and the "critical mass" of neurons needed often depends on the downstream task, suggesting that there is a task-inherent redundancy-performance Pareto frontier. Our findings shed light on the nature of representations learned by pre-trained deep neural networks and suggest that entire layers might not be necessary to perform many downstream tasks. We investigate the potential for exploiting this redundancy to achieve efficient generalization for downstream tasks and also draw caution to certain possible unintended consequences. Our code is available at \url{https://github.com/nvedant07/diffused-redundancy}., Comment: NeurIPS 2023

The remarkable attention which fair clustering has received in the last few years has resulted in a significant number of different notions of fairness. Despite the fact that these notions are well-justified, they are often motivated and studied in a disjoint manner where one fairness desideratum is considered exclusively in isolation from the others. This leaves the understanding of the relations between different fairness notions as an important open problem in fair clustering. In this paper, we take the first step in this direction. Specifically, we consider the two most prominent demographic representation fairness notions in clustering: (1) Group Fairness (GF), where the different demographic groups are supposed to have close to population-level representation in each cluster and (2) Diversity in Center Selection (DS), where the selected centers are supposed to have close to population-level representation of each group. We show that given a constant approximation algorithm for one constraint (GF or DS only) we can obtain a constant approximation solution that satisfies both constraints simultaneously. Interestingly, we prove that any given solution that satisfies the GF constraint can always be post-processed at a bounded degradation to the clustering cost to additionally satisfy the DS constraint while the reverse is not true. Furthermore, we show that both GF and DS are incompatible (having an empty feasibility set in the worst case) with a collection of other distance-based fairness notions. Finally, we carry experiments to validate our theoretical findings.

This workshop Report Out focuses on the foundational elements of trustworthy AI and OR technology, and how to ensure all AI and OR systems implement these elements in their system designs. Four sessions on various topics within Trustworthy AI were held, these being Fairness, Explainable AI/Causality, Robustness/Privacy, and Human Alignment and Human-Computer Interaction. Following discussions of each of these topics, workshop participants also brainstormed challenge problems which require the collaboration of AI and OR researchers and will result in the integration of basic techniques from both fields to eventually benefit societal needs.

As neural networks increasingly make critical decisions in high-stakes settings, monitoring and explaining their behavior in an understandable and trustworthy manner is a necessity. One commonly used type of explainer is post hoc feature attribution, a family of methods for giving each feature in an input a score corresponding to its influence on a model's output. A major limitation of this family of explainers in practice is that they can disagree on which features are more important than others. Our contribution in this paper is a method of training models with this disagreement problem in mind. We do this by introducing a Post hoc Explainer Agreement Regularization (PEAR) loss term alongside the standard term corresponding to accuracy, an additional term that measures the difference in feature attribution between a pair of explainers. We observe on three datasets that we can train a model with this loss term to improve explanation consensus on unseen data, and see improved consensus between explainers other than those used in the loss term. We examine the trade-off between improved consensus and model performance. And finally, we study the influence our method has on feature attribution explanations.

Deployed artificial intelligence (AI) often impacts humans, and there is no one-size-fits-all metric to evaluate these tools. Human-centered evaluation of AI-based systems combines quantitative and qualitative analysis and human input. It has been explored to some depth in the explainable AI (XAI) and human-computer interaction (HCI) communities. Gaps remain, but the basic understanding that humans interact with AI and accompanying explanations, and that humans' needs -- complete with their cognitive biases and quirks -- should be held front and center, is accepted by the community. In this paper, we draw parallels between the relatively mature field of XAI and the rapidly evolving research boom around large language models (LLMs). Accepted evaluative metrics for LLMs are not human-centered. We argue that many of the same paths tread by the XAI community over the past decade will be retread when discussing LLMs. Specifically, we argue that humans' tendencies -- again, complete with their cognitive biases and quirks -- should rest front and center when evaluating deployed LLMs. We outline three developed focus areas of human-centered evaluation of XAI: mental models, use case utility, and cognitive engagement, and we highlight the importance of exploring each of these concepts for LLMs. Our goal is to jumpstart human-centered LLM evaluation., Comment: Accepted to CHI 2023 workshop on Generative AI and HCI

Motivated by mitigating potentially harmful impacts of technologies, the AI community has formulated and accepted mathematical definitions for certain pillars of accountability: e.g. privacy, fairness, and model transparency. Yet, we argue this is fundamentally misguided because these definitions are imperfect, siloed constructions of the human values they hope to proxy, while giving the guise that those values are sufficiently embedded in our technologies. Under popularized methods, tensions arise when practitioners attempt to achieve each pillar of fairness, privacy, and transparency in isolation or simultaneously. In this position paper, we push for redirection. We argue that the AI community needs to consider all the consequences of choosing certain formulations of these pillars -- not just the technical incompatibilities, but also the effects within the context of deployment. We point towards sociotechnical research for frameworks for the latter, but push for broader efforts into implementing these in practice., Comment: Contributed Talk, NeurIPS 2022 Workshop on Algorithmic Fairness through the Lens of Causality and Privacy; To be published in 2023 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)

Restless multi-armed bandits are often used to model budget-constrained resource allocation tasks where receipt of the resource is associated with an increased probability of a favorable state transition. Prior work assumes that individual arms only benefit if they receive the resource directly. However, many allocation tasks occur within communities and can be characterized by positive externalities that allow arms to derive partial benefit when their neighbor(s) receive the resource. We thus introduce networked restless bandits, a novel multi-armed bandit setting in which arms are both restless and embedded within a directed graph. We then present Greta, a graph-aware, Whittle index-based heuristic algorithm that can be used to efficiently construct a constrained reward-maximizing action vector at each timestep. Our empirical results demonstrate that Greta outperforms comparison policies across a range of hyperparameter values and graph topologies., Comment: Accepted to AAAI 2023

Reinforcement Learning (RL) algorithms have been successfully applied to real world situations like illegal smuggling, poaching, deforestation, climate change, airport security, etc. These scenarios can be framed as Stackelberg security games (SSGs) where defenders and attackers compete to control target resources. The algorithm's competency is assessed by which agent is controlling the targets. This review investigates modeling of SSGs in RL with a focus on possible improvements of target representations in RL algorithms., Comment: Appears in Proceedings of AAAI FSS-22 Symposium "Lessons Learned for Autonomous Assessment of Machine Abilities (LLAAMA)"

Facial analysis systems have been deployed by large companies and critiqued by scholars and activists for the past decade. Many existing algorithmic audits examine the performance of these systems on later stage elements of facial analysis systems like facial recognition and age, emotion, or perceived gender prediction; however, a core component to these systems has been vastly understudied from a fairness perspective: face detection, sometimes called face localization. Since face detection is a pre-requisite step in facial analysis systems, the bias we observe in face detection will flow downstream to the other components like facial recognition and emotion prediction. Additionally, no prior work has focused on the robustness of these systems under various perturbations and corruptions, which leaves open the question of how various people are impacted by these phenomena. We present the first of its kind detailed benchmark of face detection systems, specifically examining the robustness to noise of commercial and academic models. We use both standard and recently released academic facial datasets to quantitatively analyze trends in face detection robustness. Across all the datasets and systems, we generally find that photos of individuals who are $\textit{masculine presenting}$, $\textit{older}$, of $\textit{darker skin type}$, or have $\textit{dim lighting}$ are more susceptible to errors than their counterparts in other identities., Comment: NeurIPS Datasets & Benchmarks Track 2022

Recommender systems influence many of our interactions in the digital world -- impacting how we shop for clothes, sorting what we see when browsing YouTube or TikTok, and determining which restaurants and hotels we are shown when using hospitality platforms. Modern recommender systems are large, opaque models trained on a mixture of proprietary and open-source datasets. Naturally, issues of trust arise on both the developer and user side: is the system working correctly, and why did a user receive (or not receive) a particular recommendation? Providing an explanation alongside a recommendation alleviates some of these concerns. The status quo for auxiliary recommender system feedback is either user-specific explanations (e.g., "users who bought item B also bought item A") or item-specific explanations (e.g., "we are recommending item A because you watched/bought item B"). However, users bring personalized context into their search experience, valuing an item as a function of that item's attributes and their own personal preferences. In this work, we propose RecXplainer, a novel method for generating fine-grained explanations based on a user's preferences over the attributes of recommended items. We evaluate RecXplainer on five real-world and large-scale recommendation datasets using five different kinds of recommender systems to demonstrate the efficacy of RecXplainer in capturing users' preferences over item attributes and using them to explain recommendations. We also compare RecXplainer to five baselines and show RecXplainer's exceptional performance on ten metrics., Comment: Awarded the Best Student Paper at TEA Workshop at NeurIPS 2022

Green Security Games with real-time information (GSG-I) add the real-time information about the agents' movement to the typical GSG formulation. Prior works on GSG-I have used deep reinforcement learning (DRL) to learn the best policy for the agent in such an environment without any need to store the huge number of state representations for GSG-I. However, the decision-making process of DRL methods is largely opaque, which results in a lack of trust in their predictions. To tackle this issue, we present an interpretable DRL method for GSG-I that generates visualization to explain the decisions taken by the DRL algorithm. We also show that this approach performs better and works well with a simpler training regimen compared to the existing method.

Face recognition systems are widely deployed in safety-critical applications, including law enforcement, yet they exhibit bias across a range of socio-demographic dimensions, such as gender and race. Conventional wisdom dictates that model biases arise from biased training data. As a consequence, previous works on bias mitigation largely focused on pre-processing the training data, adding penalties to prevent bias from effecting the model during training, or post-processing predictions to debias them, yet these approaches have shown limited success on hard problems such as face recognition. In our work, we discover that biases are actually inherent to neural network architectures themselves. Following this reframing, we conduct the first neural architecture search for fairness, jointly with a search for hyperparameters. Our search outputs a suite of models which Pareto-dominate all other high-performance architectures and existing bias mitigation methods in terms of accuracy and fairness, often by large margins, on the two most widely used datasets for face identification, CelebA and VGGFace2. Furthermore, these models generalize to other datasets and sensitive attributes. We release our code, models and raw data files at https://github.com/dooleys/FR-NAS.

A major challenge in studying robustness in deep learning is defining the set of ``meaningless'' perturbations to which a given Neural Network (NN) should be invariant. Most work on robustness implicitly uses a human as the reference model to define such perturbations. Our work offers a new view on robustness by using another reference NN to define the set of perturbations a given NN should be invariant to, thus generalizing the reliance on a reference ``human NN'' to any NN. This makes measuring robustness equivalent to measuring the extent to which two NNs share invariances, for which we propose a measure called STIR. STIR re-purposes existing representation similarity measures to make them suitable for measuring shared invariances. Using our measure, we are able to gain insights into how shared invariances vary with changes in weight initialization, architecture, loss functions, and training dataset. Our implementation is available at: \url{https://github.com/nvedant07/STIR}., Comment: Accepted for oral presentation at ICML 2022

While other areas of machine learning have seen more and more automation, designing a high-performing recommender system still requires a high level of human effort. Furthermore, recent work has shown that modern recommender system algorithms do not always improve over well-tuned baselines. A natural follow-up question is, "how do we choose the right algorithm for a new dataset and performance metric?" In this work, we start by giving the first large-scale study of recommender system approaches by comparing 18 algorithms and 100 sets of hyperparameters across 85 datasets and 315 metrics. We find that the best algorithms and hyperparameters are highly dependent on the dataset and performance metric, however, there are also strong correlations between the performance of each algorithm and various meta-features of the datasets. Motivated by these findings, we create RecZilla, a meta-learning approach to recommender systems that uses a model to predict the best algorithm and hyperparameters for new, unseen datasets. By using far more meta-training data than prior work, RecZilla is able to substantially reduce the level of human involvement when faced with a new recommender system application. We not only release our code and pretrained RecZilla models, but also all of our raw experimental results, so that practitioners can train a RecZilla model for their desired performance metric: https://github.com/naszilla/reczilla., Comment: NeurIPS 2022

Numerous algorithms have been produced for the fundamental problem of clustering under many different notions of fairness. Perhaps the most common family of notions currently studied is group fairness, in which proportional group representation is ensured in every cluster. We extend this direction by considering the downstream application of clustering and how group fairness should be ensured for such a setting. Specifically, we consider a common setting in which a decision-maker runs a clustering algorithm, inspects the center of each cluster, and decides an appropriate outcome (label) for its corresponding cluster. In hiring for example, there could be two outcomes, positive (hire) or negative (reject), and each cluster would be assigned one of these two outcomes. To ensure group fairness in such a setting, we would desire proportional group representation in every label but not necessarily in every cluster as is done in group fair clustering. We provide algorithms for such problems and show that in contrast to their NP-hard counterparts in group fair clustering, they permit efficient solutions. We also consider a well-motivated alternative setting where the decision-maker is free to assign labels to the clusters regardless of the centers' positions in the metric space. We show that this setting exhibits interesting transitions from computationally hard to easy according to additional constraints on the problem. Moreover, when the constraint parameters take on natural values we show a randomized algorithm for this setting that always achieves an optimal clustering and satisfies the fairness constraints in expectation. Finally, we run experiments on real world datasets that validate the effectiveness of our algorithms., Comment: Accepted to KDD 2022

Visualizing optimization landscapes has led to many fundamental insights in numeric optimization, and novel improvements to optimization techniques. However, visualizations of the objective that reinforcement learning optimizes (the "reward surface") have only ever been generated for a small number of narrow contexts. This work presents reward surfaces and related visualizations of 27 of the most widely used reinforcement learning environments in Gym for the first time. We also explore reward surfaces in the policy gradient direction and show for the first time that many popular reinforcement learning environments have frequent "cliffs" (sudden large drops in expected return). We demonstrate that A2C often "dives off" these cliffs into low reward regions of the parameter space while PPO avoids them, confirming a popular intuition for PPO's improved performance over previous methods. We additionally introduce a highly extensible library that allows researchers to easily generate these visualizations in the future. Our findings provide new intuition to explain the successes and failures of modern RL methods, and our visualizations concretely characterize several failure modes of reinforcement learning agents in novel ways., Comment: Accepted at ICML 2022 Camera-Ready Version

While the stable marriage problem and its variants model a vast range of matching markets, they fail to capture complex agent relationships, such as the affiliation of applicants and employers in an interview marketplace. To model this problem, the existing literature on matching with externalities permits agents to provide complete and total rankings over matchings based off of both their own and their affiliates' matches. This complete ordering restriction is unrealistic, and further the model may have an empty core. To address this, we introduce the Dichotomous Affiliate Stable Matching (DASM) Problem, where agents' preferences indicate dichotomous acceptance or rejection of another agent in the marketplace, both for themselves and their affiliates. We also assume the agent's preferences over entire matchings are determined by a general weighted valuation function of their (and their affiliates') matches. Our results are threefold: (1) we use a human study to show that real-world matching rankings follow our assumed valuation function; (2) we prove that there always exists a stable solution by providing an efficient, easily-implementable algorithm that finds such a solution; and (3) we experimentally validate the efficiency of our algorithm versus a linear-programming-based approach., Comment: 19 pages, 2 figures

As facial recognition systems are deployed more widely, scholars and activists have studied their biases and harms. Audits are commonly used to accomplish this and compare the algorithmic facial recognition systems' performance against datasets with various metadata labels about the subjects of the images. Seminal works have found discrepancies in performance by gender expression, age, perceived race, skin type, etc. These studies and audits often examine algorithms which fall into two categories: academic models or commercial models. We present a detailed comparison between academic and commercial face detection systems, specifically examining robustness to noise. We find that state-of-the-art academic face detection models exhibit demographic disparities in their noise robustness, specifically by having statistically significant decreased performance on older individuals and those who present their gender in a masculine manner. When we compare the size of these disparities to that of commercial models, we conclude that commercial models - in contrast to their relatively larger development budget and industry-level fairness commitments - are always as biased or more biased than an academic model., Comment: This preprint and arXiv:2108.12508 were combined and a more rigorous analysis added to result in the NeurIPS Datasets & Benchmark 2022 paper arXiv:2211.15937

Online bipartite-matching platforms are ubiquitous and find applications in important areas such as crowdsourcing and ridesharing. In the most general form, the platform consists of three entities: two sides to be matched and a platform operator that decides the matching. The design of algorithms for such platforms has traditionally focused on the operator's (expected) profit. Since fairness has become an important consideration that was ignored in the existing algorithms a collection of online matching algorithms have been developed that give a fair treatment guarantee for one side of the market at the expense of a drop in the operator's profit. In this paper, we generalize the existing work to offer fair treatment guarantees to both sides of the market simultaneously, at a calculated worst case drop to operator profit. We consider group and individual Rawlsian fairness criteria. Moreover, our algorithms have theoretical guarantees and have adjustable parameters that can be tuned as desired to balance the trade-off between the utilities of the three sides. We also derive hardness results that give clear upper bounds over the performance of any algorithm., Comment: Accepted to AAAI 2023

An evaluation criterion for safe and trustworthy deep learning is how well the invariances captured by representations of deep neural networks (DNNs) are shared with humans. We identify challenges in measuring these invariances. Prior works used gradient-based methods to generate identically represented inputs (IRIs), ie, inputs which have identical representations (on a given layer) of a neural network, and thus capture invariances of a given network. One necessary criterion for a network's invariances to align with human perception is for its IRIs look 'similar' to humans. Prior works, however, have mixed takeaways; some argue that later layers of DNNs do not learn human-like invariances (\cite{jenelle2019metamers}) yet others seem to indicate otherwise (\cite{mahendran2014understanding}). We argue that the loss function used to generate IRIs can heavily affect takeaways about invariances of the network and is the primary reason for these conflicting findings. We propose an adversarial regularizer on the IRI generation loss that finds IRIs that make any model appear to have very little shared invariance with humans. Based on this evidence, we argue that there is scope for improving models to have human-like invariances, and further, to have meaningful comparisons between models one should use IRIs generated using the regularizer-free loss. We then conduct an in-depth investigation of how different components (eg architectures, training losses, data augmentations) of the deep learning pipeline contribute to learning models that have good alignment with humans. We find that architectures with residual connections trained using a (self-supervised) contrastive loss with $\ell_p$ ball adversarial data augmentation tend to learn invariances that are most aligned with humans. Code: \url{github.com/nvedant07/Human-NN-Alignment}., Comment: AAAI 2023

Much recent research has uncovered and discussed serious concerns of bias in facial analysis technologies, finding performance disparities between groups of people based on perceived gender, skin type, lighting condition, etc. These audits are immensely important and successful at measuring algorithmic bias but have two major challenges: the audits (1) use facial recognition datasets which lack quality metadata, like LFW and CelebA, and (2) do not compare their observed algorithmic bias to the biases of their human alternatives. In this paper, we release improvements to the LFW and CelebA datasets which will enable future researchers to obtain measurements of algorithmic bias that are not tainted by major flaws in the dataset (e.g. identical images appearing in both the gallery and test set). We also use these new data to develop a series of challenging facial identification and verification questions that we administered to various algorithms and a large, balanced sample of human reviewers. We find that both computer models and human survey participants perform significantly better at the verification task, generally obtain lower accuracy rates on dark-skinned or female subjects for both tasks, and obtain higher accuracy rates when their demographics match that of the question. Computer models are observed to achieve a higher level of accuracy than the survey participants on both tasks and exhibit bias to similar degrees as the human survey participants.

Rideshare and ride-pooling platforms use artificial intelligence-based matching algorithms to pair riders and drivers. However, these platforms can induce inequality either through an unequal income distribution or disparate treatment of riders. We investigate two methods to reduce forms of inequality in ride-pooling platforms: (1) incorporating fairness constraints into the objective function and (2) redistributing income to drivers to reduce income fluctuation and inequality. To evaluate our solutions, we use the New York City taxi data set. For the first method, we find that optimizing for driver-side fairness outperforms state-of-the-art models on the number of riders serviced, both in the worst-off neighborhood and overall, showing that optimizing for fairness can assist profitability in certain circumstances. For the second method, we explore income redistribution as a way to combat income inequality by having drivers keep an $r$ fraction of their income, and contributing the rest to a redistribution pool. For certain values of $r$, most drivers earn near their Shapley value, while still incentivizing drivers to maximize value, thereby avoiding the free-rider problem and reducing income variability. The first method can be extended to many definitions of fairness and the second method provably improves fairness without affecting profitability.

Facial detection and analysis systems have been deployed by large companies and critiqued by scholars and activists for the past decade. Critiques that focus on system performance analyze disparity of the system's output, i.e., how frequently is a face detected for different Fitzpatrick skin types or perceived genders. However, we focus on the robustness of these system outputs under noisy natural perturbations. We present the first of its kind detailed benchmark of the robustness of three such systems: Amazon Rekognition, Microsoft Azure, and Google Cloud Platform. We use both standard and recently released academic facial datasets to quantitatively analyze trends in robustness for each. Across all the datasets and systems, we generally find that photos of individuals who are older, masculine presenting, of darker skin type, or have dim lighting are more susceptible to errors than their counterparts in other identities.

Global demand for donated blood far exceeds supply, and unmet need is greatest in low- and middle-income countries; experts suggest that large-scale coordination is necessary to alleviate demand. Using the Facebook Blood Donation tool, we conduct the first large-scale algorithmic matching of blood donors with donation opportunities. While measuring actual donation rates remains a challenge, we measure donor action (e.g., making a donation appointment) as a proxy for actual donation. We develop automated policies for matching donors with donation opportunities, based on an online matching model. We provide theoretical guarantees for these policies, both regarding the number of expected donations and the equitable treatment of blood recipients. In simulations, a simple matching strategy increases the number of donations by 5-10%; a pilot experiment with real donors shows a 5% relative increase in donor action rate (from 3.7% to 3.9%). When scaled to the global Blood Donation tool user base, this corresponds to an increase of around one hundred thousand users taking action toward donation. Further, observing donor action on a social network can shed light onto donor behavior and response to incentives. Our initial findings align with several observations made in the medical and social science literature regarding donor behavior., Comment: An early version of this paper appeared at EC'20. (https://doi.org/10.1145/3391403.3399458)

As machine learning (ML) systems take a more prominent and central role in contributing to life-impacting decisions, ensuring their trustworthiness and accountability is of utmost importance. Explanations sit at the core of these desirable attributes of a ML system. The emerging field is frequently called ``Explainable AI (XAI)'' or ``Explainable ML.'' The goal of explainable ML is to intuitively explain the predictions of a ML system, while adhering to the needs to various stakeholders. Many explanation techniques were developed with contributions from both academia and industry. However, there are several existing challenges that have not garnered enough interest and serve as roadblocks to widespread adoption of explainable ML. In this short paper, we enumerate challenges in explainable ML from an industry perspective. We hope these challenges will serve as promising future research directions, and would contribute to democratizing explainable ML., Comment: Presented at JOURNE workshop at MLSYS 2021 (https://sites.google.com/view/workshop-journe/home)

Counterfactual explanations (CFEs) are an emerging technique under the umbrella of interpretability of machine learning (ML) models. They provide ``what if'' feedback of the form ``if an input datapoint were $x'$ instead of $x$, then an ML model's output would be $y'$ instead of $y$.'' Counterfactual explainability for ML models has yet to see widespread adoption in industry. In this short paper, we posit reasons for this slow uptake. Leveraging recent work outlining desirable properties of CFEs and our experience running the ML wing of a model monitoring startup, we identify outstanding obstacles hindering CFE deployment in industry., Comment: Presented at CHI HCXAI 2021 workshop

Restless and collapsing bandits are often used to model budget-constrained resource allocation in settings where arms have action-dependent transition probabilities, such as the allocation of health interventions among patients. However, state-of-the-art Whittle-index-based approaches to this planning problem either do not consider fairness among arms, or incentivize fairness without guaranteeing it. We thus introduce ProbFair, a probabilistically fair policy that maximizes total expected reward and satisfies the budget constraint while ensuring a strictly positive lower bound on the probability of being pulled at each timestep. We evaluate our algorithm on a real-world application, where interventions support continuous positive airway pressure (CPAP) therapy adherence among patients, as well as on a broader class of synthetic transition matrices. We find that ProbFair preserves utility while providing fairness guarantees.

Explainable machine learning (ML) has gained traction in recent years due to the increasing adoption of ML-based systems in many sectors. Algorithmic Recourses (ARs) provide "what if" feedback of the form "if an input datapoint were x' instead of x, then an ML-based system's output would be y' instead of y." ARs are attractive due to their actionable feedback, amenability to existing legal frameworks, and fidelity to the underlying ML model. Yet, current AR approaches are single shot -- that is, they assume x can change to x' in a single time period. We propose a novel stochastic-control-based approach that generates sequential ARs, that is, ARs that allow x to move stochastically and sequentially across intermediate states to a final state x'. Our approach is model agnostic and black box. Furthermore, the calculation of ARs is amortized such that once trained, it applies to multiple datapoints without the need for re-optimization. In addition to these primary characteristics, our approach admits optional desiderata such as adherence to the data manifold, respect for causal relations, and sparsity -- identified by past research as desirable properties of ARs. We evaluate our approach using three real-world datasets and show successful generation of sequential ARs that respect other recourse desiderata., Comment: Accepted at AAAI 2022

The design of optimal auctions is a problem of interest in economics, game theory and computer science. Despite decades of effort, strategyproof, revenue-maximizing auction designs are still not known outside of restricted settings. However, recent methods using deep learning have shown some success in approximating optimal auctions, recovering several known solutions and outperforming strong baselines when optimal auctions are not known. In addition to maximizing revenue, auction mechanisms may also seek to encourage socially desirable constraints such as allocation fairness or diversity. However, these philosophical notions neither have standardization nor do they have widely accepted formal definitions. In this paper, we propose PreferenceNet, an extension of existing neural-network-based auction mechanisms to encode constraints using (potentially human-provided) exemplars of desirable allocations. In addition, we introduce a new metric to evaluate an auction allocations' adherence to such socially desirable constraints and demonstrate that our proposed method is competitive with current state-of-the-art neural-network based auction designs. We validate our approach through human subject research and show that we are able to effectively capture real human preferences. Our code is available at https://github.com/neeharperi/PreferenceNet, Comment: This work has been accepted to Neural Information Processing Systems (NeurIPS) 2021. First two authors contributed equally

As demonstrated by Ratliff et al. (2014), inverse optimization can be used to recover the objective function parameters of players in multi-player Nash games. These games involve the optimization problems of multiple players in which the players can affect each other in their objective functions. In generalized Nash equilibrium problems (GNEPs), a player's set of feasible actions is also impacted by the actions taken by other players in the game; see Facchinei and Kanzow (2010) for more background on this problem. One example of such impact comes in the form of joint/"coupled" constraints as referenced by Rosen (1965), Harker (1991), and Facchinei et al. (2007) which involve other players' variables in the constraints of the feasible region. We extend the framework of Ratliff et al. (2014) to find inverse optimization solutions for the class of GNEPs with joint constraints. The resulting formulation is then applied to a simulated multi-player transportation problem on a road network. Also, we provide some theoretical results related to this transportation problem regarding runtime of the extended framework as well as uniqueness and non-uniqueness of solutions to our simulation experiments. We see that our model recovers parameterizations that produce the same flow patterns as the original parameterizations and that this holds true across multiple networks, different assumptions regarding players' perceived costs, and the majority of restrictive capacity settings and the associated numbers of players. Code for the project can be found at: https://github.com/sallen7/IO_GNEP.

As machine learning algorithms have been widely deployed across applications, many concerns have been raised over the fairness of their predictions, especially in high stakes settings (such as facial recognition and medical imaging). To respond to these concerns, the community has proposed and formalized various notions of fairness as well as methods for rectifying unfair behavior. While fairness constraints have been studied extensively for classical models, the effectiveness of methods for imposing fairness on deep neural networks is unclear. In this paper, we observe that these large models overfit to fairness objectives, and produce a range of unintended and undesirable consequences. We conduct our experiments on both facial recognition and automated medical diagnosis datasets using state-of-the-art architectures.

AI systems are often used to make or contribute to important decisions in a growing range of applications, including criminal justice, hiring, and medicine. Since these decisions impact human lives, it is important that the AI systems act in ways which align with human values. Techniques for preference modeling and social choice help researchers learn and aggregate peoples' preferences, which are used to guide AI behavior; thus, it is imperative that these learned preferences are accurate. These techniques often assume that people are willing to express strict preferences over alternatives; which is not true in practice. People are often indecisive, and especially so when their decision has moral implications. The philosophy and psychology literature shows that indecision is a measurable and nuanced behavior -- and that there are several different reasons people are indecisive. This complicates the task of both learning and aggregating preferences, since most of the relevant literature makes restrictive assumptions on the meaning of indecision. We begin to close this gap by formalizing several mathematical \emph{indecision} models based on theories from philosophy, psychology, and economics; these models can be used to describe (indecisive) agent decisions, both when they are allowed to express indecision and when they are not. We test these models using data collected from an online survey where participants choose how to (hypothetically) allocate organs to patients waiting for a transplant., Comment: Accepted at AAAI 2020

In barter exchanges, participants swap goods with one another without exchanging money; exchanges are often facilitated by a central clearinghouse, with the goal of maximizing the aggregate quality (or number) of swaps. Barter exchanges are subject to many forms of uncertainty--in participant preferences, the feasibility and quality of various swaps, and so on. Our work is motivated by kidney exchange, a real-world barter market in which patients in need of a kidney transplant swap their willing living donors, in order to find a better match. Modern exchanges include 2- and 3-way swaps, making the kidney exchange clearing problem NP-hard. Planned transplants often fail for a variety of reasons--if the donor organ is refused by the recipient's medical team, or if the donor and recipient are found to be medically incompatible. Due to 2- and 3-way swaps, failed transplants can "cascade" through an exchange; one US-based exchange estimated that about 85% of planned transplants failed in 2019. Many optimization-based approaches have been designed to avoid these failures; however most exchanges cannot implement these methods due to legal and policy constraints. Instead we consider a setting where exchanges can query the preferences of certain donors and recipients--asking whether they would accept a particular transplant. We characterize this as a two-stage decision problem, in which the exchange program (a) queries a small number of transplants before committing to a matching, and (b) constructs a matching according to fixed policy. We show that selecting these edges is a challenging combinatorial problem, which is non-monotonic and non-submodular, in addition to being NP-hard. We propose both a greedy heuristic and a Monte Carlo tree search, which outperforms previous approaches, using experiments on both synthetic data and real kidney exchange data from the United Network for Organ Sharing., Comment: Appears at NeurIPS 2020

Machine learning plays a role in many deployed decision systems, often in ways that are difficult or impossible to understand by human stakeholders. Explaining, in a human-understandable way, the relationship between the input and output of machine learning models is essential to the development of trustworthy machine learning based systems. A burgeoning body of research seeks to define the goals and methods of explainability in machine learning. In this paper, we seek to review and categorize research on counterfactual explanations, a specific class of explanation that provides a link between what could have happened had input to a model been changed in a particular way. Modern approaches to counterfactual explainability in machine learning draw connections to the established legal doctrine in many countries, making them appealing to fielded systems in high-impact areas such as finance and healthcare. Thus, we design a rubric with desirable properties of counterfactual explanation algorithms and comprehensively evaluate all currently proposed algorithms against that rubric. Our rubric provides easy comparison and comprehension of the advantages and disadvantages of different approaches and serves as an introduction to major research themes in this field. We also identify gaps and discuss promising research directions in the space of counterfactual explainability., Comment: 23 pages (8 pages of references)

In many labor markets, workers and firms are connected via affiliative relationships. A management consulting firm wishes to both accept the best new workers but also place its current affiliated workers at strong firms. Similarly, a research university wishes to hire strong job market candidates while also placing its own candidates at strong peer universities. We model this affiliate matching problem in a generalization of the classic stable marriage setting by permitting firms to state preferences over not just which workers to whom they are matched, but also to which firms their affiliated workers are matched. Based on results from a human survey, we find that participants (acting as firms) give preference to their own affiliate workers in surprising ways that violate some assumptions of the classical stable marriage problem. This motivates a nuanced discussion of how stability could be defined in affiliate matching problems; we give an example of a marketplace which admits a stable match under one natural definition of stability, and does not for that same marketplace under a different, but still natural, definition. We conclude by setting a research agenda toward the creation of a centralized clearing mechanism in this general setting.

Motivated by kidney exchange, we study a stochastic cycle and chain packing problem, where we aim to identify structures in a directed graph to maximize the expectation of matched edge weights. All edges are subject to failure, and the failures can have nonidentical probabilities. To the best of our knowledge, the state-of-the-art approaches are only tractable when failure probabilities are identical. We formulate a relevant non-convex optimization problem and propose a tractable mixed-integer linear programming reformulation to solve it. In addition, we propose a model that integrates both risks and the expected utilities of the matching by incorporating conditional value at risk (CVaR) into the objective function, providing a robust formulation for this problem. Subsequently, we propose a sample-average-approximation (SAA) based approach to solve this problem. We test our approaches on data from the United Network for Organ Sharing (UNOS) and compare against state-of-the-art approaches. Our model provides better performance with the same running time as a leading deterministic approach (PICEF). Our CVaR extensions with an SAA-based method improves the $\alpha \times 100\%$ ($0<\alpha\leqslant 1$) worst-case performance substantially compared to existing models.

As deep neural networks (DNNs) get adopted in an ever-increasing number of applications, explainability has emerged as a crucial desideratum for these models. In many real-world tasks, one of the principal reasons for requiring explainability is to in turn assess prediction robustness, where predictions (i.e., class labels) that do not conform to their respective explanations (e.g., presence or absence of a concept in the input) are deemed to be unreliable. However, most, if not all, prior methods for checking explanation-conformity (e.g., LIME, TCAV, saliency maps) require significant manual intervention, which hinders their large-scale deployability. In this paper, we propose a robustness-assessment framework, at the core of which is the idea of using machine-checkable concepts. Our framework defines a large number of concepts that the DNN explanations could be based on and performs the explanation-conformity check at test time to assess prediction robustness. Both steps are executed in an automated manner without requiring any human intervention and are easily scaled to datasets with a very large number of classes. Experiments on real-world datasets and human surveys show that our framework is able to enhance prediction robustness significantly: the predictions marked to be robust by our framework have significantly higher accuracy and are more robust to adversarial perturbations., Comment: 22 pages, 12 figures, 11 tables

In clustering problems, a central decision-maker is given a complete metric graph over vertices and must provide a clustering of vertices that minimizes some objective function. In fair clustering problems, vertices are endowed with a color (e.g., membership in a group), and the features of a valid clustering might also include the representation of colors in that clustering. Prior work in fair clustering assumes complete knowledge of group membership. In this paper, we generalize prior work by assuming imperfect knowledge of group membership through probabilistic assignments. We present clustering algorithms in this more general setting with approximation ratio guarantees. We also address the problem of "metric membership", where different groups have a notion of order and distance. Experiments are conducted using our proposed algorithms as well as baselines to validate our approach and also surface nuanced concerns when group membership is not known deterministically.

Optimal auctions maximize a seller's expected revenue subject to individual rationality and strategyproofness for the buyers. Myerson's seminal work in 1981 settled the case of auctioning a single item; however, subsequent decades of work have yielded little progress moving beyond a single item, leaving the design of revenue-maximizing auctions as a central open problem in the field of mechanism design. A recent thread of work in "differentiable economics" has used tools from modern deep learning to instead learn good mechanisms. We focus on the RegretNet architecture, which can represent auctions with arbitrary numbers of items and participants; it is trained to be empirically strategyproof, but the property is never exactly verified leaving potential loopholes for market participants to exploit. We propose ways to explicitly verify strategyproofness under a particular valuation profile using techniques from the neural network verification literature. Doing so requires making several modifications to the RegretNet architecture in order to represent it exactly in an integer program. We train our network and produce certificates in several settings, including settings for which the optimal strategyproof mechanism is not known.

The efficient and fair allocation of limited resources is a classical problem in economics and computer science. In kidney exchanges, a central market maker allocates living kidney donors to patients in need of an organ. Patients and donors in kidney exchanges are prioritized using ad-hoc weights decided on by committee and then fed into an allocation algorithm that determines who gets what--and who does not. In this paper, we provide an end-to-end methodology for estimating weights of individual participant profiles in a kidney exchange. We first elicit from human subjects a list of patient attributes they consider acceptable for the purpose of prioritizing patients (e.g., medical characteristics, lifestyle choices, and so on). Then, we ask subjects comparison queries between patient profiles and estimate weights in a principled way from their responses. We show how to use these weights in kidney exchange market clearing algorithms. We then evaluate the impact of the weights in simulations and find that the precise numerical values of the weights we computed matter little, other than the ordering of profiles that they imply. However, compared to not prioritizing patients at all, there is a significant effect, with certain classes of patients being (de)prioritized based on the human-elicited value judgments.

We study the problem of eliciting the preferences of a decision-maker through a moderate number of pairwise comparison queries to make them a high quality recommendation for a specific problem. We are motivated by applications in high stakes domains, such as when choosing a policy for allocating scarce resources to satisfy basic needs (e.g., kidneys for transplantation or housing for those experiencing homelessness) where a consequential recommendation needs to be made from the (partially) elicited preferences. We model uncertainty in the preferences as being set based and} investigate two settings: a) an offline elicitation setting, where all queries are made at once, and b) an online elicitation setting, where queries are selected sequentially over time in an adaptive fashion. We propose robust optimization formulations of these problems which integrate the preference elicitation and recommendation phases with aim to either maximize worst-case utility or minimize worst-case regret, and study their complexity. For the offline case, where active preference elicitation takes the form of a two and half stage robust optimization problem with decision-dependent information discovery, we provide an equivalent reformulation in the form of a mixed-binary linear program which we solve via column-and-constraint generation. For the online setting, where active preference learning takes the form of a multi-stage robust optimization problem with decision-dependent information discovery, we propose a conservative solution approach. Numerical studies on synthetic data demonstrate that our methods outperform state-of-the art approaches from the literature in terms of worst-case rank, regret, and utility. We showcase how our methodology can be used to assist a homeless services agency in choosing a policy for allocating scarce housing resources of different types to people experiencing homelessness., Comment: 74 pages, 10 figures, submitted for publication

Collaborative work often benefits from having teams or organizations with heterogeneous members. In this paper, we present a method to form such diverse teams from people arriving sequentially over time. We define a monotone submodular objective function that combines the diversity and quality of a team and propose an algorithm to maximize the objective while satisfying multiple constraints. This allows us to balance both how diverse the team is and how well it can perform the task at hand. Using crowd experiments, we show that, in practice, the algorithm leads to large gains in team diversity. Using simulations, we show how to quantify the additional cost of forming diverse teams and how to address the problem of simultaneously maximizing diversity for several attributes (e.g., country of origin, gender). Our method has applications in collaborative work ranging from team formation, the assignment of workers to teams in crowdsourcing, and reviewer allocation to journal papers arriving sequentially. Our code is publicly accessible for further research., Comment: Journal of Mechanical Design

Rideshare platforms, when assigning requests to drivers, tend to maximize profit for the system and/or minimize waiting time for riders. Such platforms can exacerbate biases that drivers may have over certain types of requests. We consider the case of peak hours when the demand for rides is more than the supply of drivers. Drivers are well aware of their advantage during the peak hours and can choose to be selective about which rides to accept. Moreover, if in such a scenario, the assignment of requests to drivers (by the platform) is made only to maximize profit and/or minimize wait time for riders, requests of a certain type (e.g. from a non-popular pickup location, or to a non-popular drop-off location) might never be assigned to a driver. Such a system can be highly unfair to riders. However, increasing fairness might come at a cost of the overall profit made by the rideshare platform. To balance these conflicting goals, we present a flexible, non-adaptive algorithm, \lpalg, that allows the platform designer to control the profit and fairness of the system via parameters $\alpha$ and $\beta$ respectively. We model the matching problem as an online bipartite matching where the set of drivers is offline and requests arrive online. Upon the arrival of a request, we use \lpalg to assign it to a driver (the driver might then choose to accept or reject it) or reject the request. We formalize the measures of profit and fairness in our setting and show that by using \lpalg, the competitive ratios for profit and fairness measures would be no worse than $\alpha/e$ and $\beta/e$ respectively. Extensive experimental results on both real-world and synthetic datasets confirm the validity of our theoretical lower bounds. Additionally, they show that $\lpalg$ under some choice of $(\alpha, \beta)$ can beat two natural heuristics, Greedy and Uniform, on \emph{both} fairness and profit., Comment: 8 pages, 4 figures, Accepted at AAAI 2020 & AIES (Oral) 2020

Bias in machine learning has manifested injustice in several areas, such as medicine, hiring, and criminal justice. In response, computer scientists have developed myriad definitions of fairness to correct this bias in fielded algorithms. While some definitions are based on established legal and ethical norms, others are largely mathematical. It is unclear whether the general public agrees with these fairness definitions, and perhaps more importantly, whether they understand these definitions. We take initial steps toward bridging this gap between ML researchers and the public, by addressing the question: does a lay audience understand a basic definition of ML fairness? We develop a metric to measure comprehension of three such definitions--demographic parity, equal opportunity, and equalized odds. We evaluate this metric using an online survey, and investigate the relationship between comprehension and sentiment, demographics, and the definition itself.