Hanna Wallach

Amy K. Heger, Liz B. Marquis, Mihaela Vorvoreanu et al. · microsoft-research

Data is central to the development and evaluation of machine learning (ML) models. However, the use of problematic or inappropriate datasets can result in harms when the resulting models are deployed. To encourage responsible AI practice through more deliberate reflection on datasets and transparency around the processes by which they are created, researchers and practitioners have begun to advocate for increased data documentation and have proposed several data documentation frameworks. However, there is little research on whether these data documentation frameworks meet the needs of ML practitioners, who both create and consume datasets. To address this gap, we set out to understand ML practitioners' data documentation perceptions, needs, challenges, and desiderata, with the goal of deriving design requirements that can inform future data documentation frameworks. We conducted a series of semi-structured interviews with 14 ML practitioners at a single large, international technology company. We had them answer a list of questions taken from datasheets for datasets (Gebru, 2021). Our findings show that current approaches to data documentation are largely ad hoc and myopic in nature. Participants expressed needs for data documentation frameworks to be adaptable to their contexts, integrated into their existing tools and workflows, and automated wherever possible. Despite the fact that data documentation frameworks are often motivated from the perspective of responsible AI, participants did not make the connection between the questions that they were asked to answer and their responsible AI implications. In addition, participants often had difficulties prioritizing the needs of dataset consumers and providing information that someone unfamiliar with their datasets might need to know. Based on these findings, we derive seven design requirements for future data documentation frameworks.

Jessie J. Smith, Saleema Amershi, Solon Barocas et al. · microsoft-research

Transparency around limitations can improve the scientific rigor of research, help ensure appropriate interpretation of research findings, and make research claims more credible. Despite these benefits, the machine learning (ML) research community lacks well-developed norms around disclosing and discussing limitations. To address this gap, we conduct an iterative design process with 30 ML and ML-adjacent researchers to develop and test REAL ML, a set of guided activities to help ML researchers recognize, explore, and articulate the limitations of their research. Using a three-stage interview and survey study, we identify ML researchers' perceptions of limitations, as well as the challenges they face when recognizing, exploring, and articulating limitations. We develop REAL ML to address some of these practical challenges, and highlight additional cultural challenges that will require broader shifts in community norms to address. We hope our study and REAL ML help move the ML research community toward more active and appropriate engagement with limitations.

Li Lucy, Su Lin Blodgett, Milad Shokouhi et al. · microsoft-research

Fairness-related assumptions about what constitute appropriate NLG system behaviors range from invariance, where systems are expected to behave identically for social groups, to adaptation, where behaviors should instead vary across them. To illuminate tensions around invariance and adaptation, we conduct five case studies, in which we perturb different types of identity-related language features (names, roles, locations, dialect, and style) in NLG system inputs. Through these cases studies, we examine people's expectations of system behaviors, and surface potential caveats of these contrasting yet commonly held assumptions. We find that motivations for adaptation include social norms, cultural differences, feature-specific information, and accommodation; in contrast, motivations for invariance include perspectives that favor prescriptivism, view adaptation as unnecessary or too difficult for NLG systems to do appropriately, and are wary of false assumptions. Our findings highlight open challenges around what constitute "fair" or "good" NLG system behaviors.

Ahmed Magooda, Alec Helyar, Kyle Jackson et al. · microsoft-research

We present a framework for the automated measurement of responsible AI (RAI) metrics for large language models (LLMs) and associated products and services. Our framework for automatically measuring harms from LLMs builds on existing technical and sociotechnical expertise and leverages the capabilities of state-of-the-art LLMs, such as GPT-4. We use this framework to run through several case studies investigating how different LLMs may violate a range of RAI-related principles. The framework may be employed alongside domain-specific sociotechnical expertise to create measurements for new harm areas in the future. By implementing this framework, we aim to enable more advanced harm measurement efforts and further the responsible use of LLMs.

Angelina Wang, Solon Barocas, Kristen Laird et al.

Previous work has largely considered the fairness of image captioning systems through the underspecified lens of "bias." In contrast, we present a set of techniques for measuring five types of representational harms, as well as the resulting measurements obtained for two of the most popular image captioning datasets using a state-of-the-art image captioning system. Our goal was not to audit this image captioning system, but rather to develop normatively grounded measurement techniques, in turn providing an opportunity to reflect on the many challenges involved. We propose multiple measurement techniques for each type of harm. We argue that by doing so, we are better able to capture the multi-faceted nature of each type of harm, in turn improving the (collective) validity of the resulting measurements. Throughout, we discuss the assumptions underlying our measurement approach and point out when they do not hold.

Dhruv Agarwal, Emily Sheng, Chad Atalla et al.

Evaluating generative AI (GenAI) systems is challenging because many targets of evaluation are broad, contested concepts, such as "reasoning," "fairness," or "creativity." When these concepts are left underspecified, it becomes unclear what should be measured or how evaluation results should be interpreted. This problem reflects a missing step: systematization, that is, moving from a broad background concept to an explicit, structured account of the concept in measurable terms. To help address the fact that systematization is cognitively demanding and resource-intensive, we investigate whether AI assistance can support this process. To enable AI-assisted systematization and assess its quality, we introduce a structured representation of a systematized concept, a concept spec, and a validation worksheet. We then develop two AI-assisted systematizers: a direct, zero-shot approach and a multi-agent approach that more closely mirrors manual systematization approaches from existing literature. We use these systematizers to produce concept specs for two concepts -- hate-based rhetoric and digital empathy -- and evaluate resulting concept specs on content validity and information recoverability.

Alexandra Chouldechova, A. Feder Cooper, Solon Barocas et al.

We argue that conclusions drawn about relative system safety or attack method efficacy via AI red teaming are often not supported by evidence provided by attack success rate (ASR) comparisons. We show, through conceptual, theoretical, and empirical contributions, that many conclusions are founded on apples-to-oranges comparisons or low-validity measurements. Our arguments are grounded in asking a simple question: When can attack success rates be meaningfully compared? To answer this question, we draw on ideas from social science measurement theory and inferential statistics, which, taken together, provide a conceptual grounding for understanding when numerical values obtained through the quantification of system attributes can be meaningfully compared. Through this lens, we articulate conditions under which ASRs can and cannot be meaningfully compared. Using jailbreaking as a running example, we provide examples and extensive discussion of apples-to-oranges ASR comparisons and measurement validity challenges.

Laura Weidinger, Inioluwa Deborah Raji, Hanna Wallach et al.

There is an increasing imperative to anticipate and understand the performance and safety of generative AI systems in real-world deployment contexts. However, the current evaluation ecosystem is insufficient: Commonly used static benchmarks face validity challenges, and ad hoc case-by-case audits rarely scale. In this piece, we advocate for maturing an evaluation science for generative AI systems. While generative AI creates unique challenges for system safety engineering and measurement science, the field can draw valuable insights from the development of safety evaluation practices in other fields, including transportation, aerospace, and pharmaceutical engineering. In particular, we present three key lessons: Evaluation metrics must be applicable to real-world performance, metrics must be iteratively refined, and evaluation institutions and norms must be established. Applying these insights, we outline a concrete path toward a more rigorous approach for evaluating generative AI systems.

Luke Guerdan, Solon Barocas, Kenneth Holstein et al.

The LLM-as-a-judge paradigm, in which a judge LLM system replaces human raters in rating the outputs of other generative AI (GenAI) systems, plays a critical role in scaling and standardizing GenAI evaluations. To validate such judge systems, evaluators assess human--judge agreement by first collecting multiple human ratings for each item in a validation corpus, then aggregating the ratings into a single, per-item gold label rating. For many items, however, rating criteria may admit multiple valid interpretations, so a human or LLM rater may deem multiple ratings "reasonable" or "correct." We call this condition rating indeterminacy. Problematically, many rating tasks that contain rating indeterminacy rely on forced-choice elicitation, whereby raters are instructed to select only one rating for each item. In this paper, we introduce a framework for validating LLM-as-a-judge systems under rating indeterminacy. We draw theoretical connections between different measures of judge system performance under different human--judge agreement metrics, and different rating elicitation and aggregation schemes. We demonstrate that differences in how humans and LLMs resolve rating indeterminacy when responding to forced-choice rating instructions can heavily bias LLM-as-a-judge validation. Through extensive experiments involving 11 real-world rating tasks and 9 commercial LLMs, we show that standard validation approaches that rely upon forced-choice ratings select judge systems that are highly suboptimal, performing as much as 31% worse than judge systems selected by our approach that uses multi-label "response set" ratings to account for rating indeterminacy. We conclude with concrete recommendations for more principled approaches to LLM-as-a-judge validation.

Emma Harvey, Emily Sheng, Su Lin Blodgett et al. · microsoft-research

The NLP research community has made publicly available numerous instruments for measuring representational harms caused by large language model (LLM)-based systems. These instruments have taken the form of datasets, metrics, tools, and more. In this paper, we examine the extent to which such instruments meet the needs of practitioners tasked with evaluating LLM-based systems. Via semi-structured interviews with 12 such practitioners, we find that practitioners are often unable to use publicly available instruments for measuring representational harms. We identify two types of challenges. In some cases, instruments are not useful because they do not meaningfully measure what practitioners seek to measure or are otherwise misaligned with practitioner needs. In other cases, instruments - even useful instruments - are not used by practitioners due to practical and institutional barriers impeding their uptake. Drawing on measurement theory and pragmatic measurement, we provide recommendations for addressing these challenges to better meet practitioner needs.

Luke Guerdan, Hanna Wallach, Solon Barocas et al.

Large language model (LLM) evaluations often assume there is a single correct response -- a gold label -- for each item in the evaluation corpus. However, some tasks can be ambiguous -- i.e., they provide insufficient information to identify a unique interpretation -- or vague -- i.e., they do not clearly indicate where to draw the line when making a determination. Both ambiguity and vagueness can cause task indeterminacy -- the condition where some items in the evaluation corpus have more than one correct response. In this paper, we develop a framework for evaluating LLMs under task indeterminacy. Our framework disentangles the relationships between task specification, human ratings, and LLM responses in the LLM evaluation pipeline. Using our framework, we conduct a synthetic experiment showing that evaluations that use the "gold label" assumption underestimate the true performance. We also provide a method for estimating an error-adjusted performance interval given partial knowledge about indeterminate items in the evaluation corpus. We conclude by outlining implications of our work for the research community.

Emily Corvi, Hannah Washington, Stefanie Reed et al.

Representational harms are widely recognized among fairness-related harms caused by generative language systems. However, their definitions are commonly under-specified. We make a theoretical contribution to the specification of representational harms by introducing a framework, grounded in speech act theory (Austin, 1962), that conceptualizes representational harms caused by generative language systems as the perlocutionary effects (i.e., real-world impacts) of particular types of illocutionary acts (i.e., system behaviors). Building on this argument and drawing on relevant literature from linguistic anthropology and sociolinguistics, we provide new definitions of stereotyping, demeaning, and erasure. We then use our framework to develop a granular taxonomy of illocutionary acts that cause representational harms, going beyond the high-level taxonomies presented in previous work. We also discuss the ways that our framework and taxonomy can support the development of valid measurement instruments. Finally, we demonstrate the utility of our framework and taxonomy via a case study that engages with recent conceptual debates about what constitutes a representational harm and how such harms should be measured.

A. Feder Cooper, Christopher A. Choquette-Choo, Miranda Bogen et al. · deepmind

"Machine unlearning" is a popular proposed solution for mitigating the existence of content in an AI model that is problematic for legal or moral reasons, including privacy, copyright, safety, and more. For example, unlearning is often invoked as a solution for removing the effects of specific information from a generative-AI model's parameters, e.g., a particular individual's personal data or the inclusion of copyrighted content in the model's training data. Unlearning is also proposed as a way to prevent a model from generating targeted types of information in its outputs, e.g., generations that closely resemble a particular individual's data or reflect the concept of "Spiderman." Both of these goals--the targeted removal of information from a model and the targeted suppression of information from a model's outputs--present various technical and substantive challenges. We provide a framework for ML researchers and policymakers to think rigorously about these challenges, identifying several mismatches between the goals of unlearning and feasible implementations. These mismatches explain why unlearning is not a general-purpose solution for circumscribing generative-AI model behavior in service of broader positive impact.

Michael Madaio, Lisa Egede, Hariharan Subramonyam et al.

Various tools and practices have been developed to support practitioners in identifying, assessing, and mitigating fairness-related harms caused by AI systems. However, prior research has highlighted gaps between the intended design of these tools and practices and their use within particular contexts, including gaps caused by the role that organizational factors play in shaping fairness work. In this paper, we investigate these gaps for one such practice: disaggregated evaluations of AI systems, intended to uncover performance disparities between demographic groups. By conducting semi-structured interviews and structured workshops with thirty-three AI practitioners from ten teams at three technology companies, we identify practitioners' processes, challenges, and needs for support when designing disaggregated evaluations. We find that practitioners face challenges when choosing performance metrics, identifying the most relevant direct stakeholders and demographic groups on which to focus, and collecting datasets with which to conduct disaggregated evaluations. More generally, we identify impacts on fairness work stemming from a lack of engagement with direct stakeholders or domain experts, business imperatives that prioritize customers over marginalized groups, and the drive to deploy AI systems at scale.

Aaron Schein, Anjali Nagulpally, Hanna Wallach et al.

We present a new non-negative matrix factorization model for $(0,1)$ bounded-support data based on the doubly non-central beta (DNCB) distribution, a generalization of the beta distribution. The expressiveness of the DNCB distribution is particularly useful for modeling DNA methylation datasets, which are typically highly dispersed and multi-modal; however, the model structure is sufficiently general that it can be adapted to many other domains where latent representations of $(0,1)$ bounded-support data are of interest. Although the DNCB distribution lacks a closed-form conjugate prior, several augmentations let us derive an efficient posterior inference algorithm composed entirely of analytic updates. Our model improves out-of-sample predictive performance on both real and synthetic DNA methylation datasets over state-of-the-art methods in bioinformatics. In addition, our model yields meaningful latent representations that accord with existing biological knowledge.

David Alvarez-Melis, Harmanpreet Kaur, Hal Daumé et al.

We take inspiration from the study of human explanation to inform the design and evaluation of interpretability methods in machine learning. First, we survey the literature on human explanation in philosophy, cognitive science, and the social sciences, and propose a list of design principles for machine-generated explanations that are meaningful to humans. Using the concept of weight of evidence from information theory, we develop a method for generating explanations that adhere to these principles. We show that this method can be adapted to handle high-dimensional, multi-class settings, yielding a flexible framework for generating explanations. We demonstrate that these explanations can be estimated accurately from finite samples and are robust to small perturbations of the inputs. We also evaluate our method through a qualitative user study with machine learning practitioners, where we observe that the resulting explanations are usable despite some participants struggling with background concepts like prior class probabilities. Finally, we conclude by surfacing~design~implications for interpretability tools in general.

Solon Barocas, Anhong Guo, Ece Kamar et al.

Disaggregated evaluations of AI systems, in which system performance is assessed and reported separately for different groups of people, are conceptually simple. However, their design involves a variety of choices. Some of these choices influence the results that will be obtained, and thus the conclusions that can be drawn; others influence the impacts -- both beneficial and harmful -- that a disaggregated evaluation will have on people, including the people whose data is used to conduct the evaluation. We argue that a deeper understanding of these choices will enable researchers and practitioners to design careful and conclusive disaggregated evaluations. We also argue that better documentation of these choices, along with the underlying considerations and tradeoffs that have been made, will help others when interpreting an evaluation's results and conclusions.

Su Lin Blodgett, Solon Barocas, Hal Daumé et al.

We survey 146 papers analyzing "bias" in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing "bias" is an inherently normative process. We further find that these papers' proposed quantitative techniques for measuring or mitigating "bias" are poorly matched to their motivations and do not engage with the relevant literature outside of NLP. Based on these findings, we describe the beginnings of a path forward by proposing three recommendations that should guide work analyzing "bias" in NLP systems. These recommendations rest on a greater recognition of the relationships between language and social hierarchies, encouraging researchers and practitioners to articulate their conceptualizations of "bias"---i.e., what kinds of system behaviors are harmful, in what ways, to whom, and why, as well as the normative reasoning underlying these statements---and to center work around the lived experiences of members of communities affected by NLP systems, while interrogating and reimagining the power relations between technologists and such communities.

Adina Williams, Ryan Cotterell, Lawrence Wolf-Sonkin et al.

We use large-scale corpora in six different gendered languages, along with tools from NLP and information theory, to test whether there is a relationship between the grammatical genders of inanimate nouns and the adjectives used to describe those nouns. For all six languages, we find that there is a statistically significant relationship. We also find that there are statistically significant relationships between the grammatical genders of inanimate nouns and the verbs that take those nouns as direct objects, as indirect objects, and as subjects. We defer a deeper investigation of these relationships for future work.

Abigail Z. Jacobs, Hanna Wallach

We propose measurement modeling from the quantitative social sciences as a framework for understanding fairness in computational systems. Computational systems often involve unobservable theoretical constructs, such as socioeconomic status, teacher effectiveness, and risk of recidivism. Such constructs cannot be measured directly and must instead be inferred from measurements of observable properties (and other unobservable theoretical constructs) thought to be related to them -- i.e., operationalized via a measurement model. This process, which necessarily involves making assumptions, introduces the potential for mismatches between the theoretical understanding of the construct purported to be measured and its operationalization. We argue that many of the harms discussed in the literature on fairness in computational systems are direct results of such mismatches. We show how some of these harms could have been anticipated and, in some cases, mitigated if viewed through the lens of measurement modeling. To do this, we contribute fairness-oriented conceptualizations of construct reliability and construct validity that unite traditions from political science, education, and psychology and provide a set of tools for making explicit and testing assumptions about constructs and their operationalizations. We then turn to fairness itself, an essentially contested construct that has different theoretical understandings in different contexts. We argue that this contestedness underlies recent debates about fairness definitions: although these debates appear to be about different operationalizations, they are, in fact, debates about different theoretical understandings of fairness. We show how measurement modeling can provide a framework for getting to the core of these debates.

David Alvarez-Melis, Hal Daumé, Jennifer Wortman Vaughan et al.

Interpretability is an elusive but highly sought-after characteristic of modern machine learning methods. Recent work has focused on interpretability via $\textit{explanations}$, which justify individual model predictions. In this work, we take a step towards reconciling machine explanations with those that humans produce and prefer by taking inspiration from the study of explanation in philosophy, cognitive science, and the social sciences. We identify key aspects in which these human explanations differ from current machine explanations, distill them into a list of desiderata, and formalize them into a framework via the notion of $\textit{weight of evidence}$ from information theory. Finally, we instantiate this framework in two simple applications and show it produces intuitive and comprehensible explanations.

Adina Williams, Ryan Cotterell, Lawrence Wolf-Sonkin et al.

Many of the world's languages employ grammatical gender on the lexeme. For example, in Spanish, the word for 'house' (casa) is feminine, whereas the word for 'paper' (papel) is masculine. To a speaker of a genderless language, this assignment seems to exist with neither rhyme nor reason. But is the assignment of inanimate nouns to grammatical genders truly arbitrary? We present the first large-scale investigation of the arbitrariness of noun-gender assignments. To that end, we use canonical correlation analysis to correlate the grammatical gender of inanimate nouns with an externally grounded definition of their lexical semantics. We find that 18 languages exhibit a significant correlation between grammatical gender and lexical semantics.

Aaron Schein, Scott W. Linderman, Mingyuan Zhou et al.

This paper presents the Poisson-randomized gamma dynamical system (PRGDS), a model for sequentially observed count tensors that encodes a strong inductive bias toward sparsity and burstiness. The PRGDS is based on a new motif in Bayesian latent variable modeling, an alternating chain of discrete Poisson and continuous gamma latent states that is analytically convenient and computationally tractable. This motif yields closed-form complete conditionals for all variables by way of the Bessel distribution and a novel discrete distribution that we call the shifted confluent hypergeometric distribution. We draw connections to closely related models and compare the PRGDS to these models in studies of real-world count data sets of text, international events, and neural spike trains. We find that a sparse variant of the PRGDS, which allows the continuous gamma latent states to take values of exactly zero, often obtains better predictive performance than other models and is uniquely capable of inferring latent structures that are highly localized in time.

Anhong Guo, Ece Kamar, Jennifer Wortman Vaughan et al.

AI technologies have the potential to dramatically impact the lives of people with disabilities (PWD). Indeed, improving the lives of PWD is a motivator for many state-of-the-art AI systems, such as automated speech recognition tools that can caption videos for people who are deaf and hard of hearing, or language prediction algorithms that can augment communication for people with speech or cognitive disabilities. However, widely deployed AI systems may not work properly for PWD, or worse, may actively discriminate against them. These considerations regarding fairness in AI for PWD have thus far received little attention. In this position paper, we identify potential areas of concern regarding how several AI technology categories may impact particular disability constituencies if care is not taken in their design, development, and testing. We intend for this risk assessment of how various classes of AI might interact with various classes of disability to provide a roadmap for future research that is needed to gather data, test these hypotheses, and build more inclusive algorithms.

Alexander Hoyle, Wolf-Sonkin, Hanna Wallach et al.

Studying the ways in which language is gendered has long been an area of interest in sociolinguistics. Studies have explored, for example, the speech of male and female characters in film and the language used to describe male and female politicians. In this paper, we aim not to merely study this phenomenon qualitatively, but instead to quantify the degree to which the language used to describe men and women is different and, moreover, different in a positive or negative way. To that end, we introduce a generative latent-variable model that jointly represents adjective (or verb) choice, with its sentiment, given the natural gender of a head (or dependent) noun. We find that there are significant differences between descriptions of male and female nouns and that these differences align with common gender stereotypes: Positive adjectives used to describe women are more often related to their bodies than adjectives used to describe men.

Ran Zmigrod, Sabrina J. Mielke, Hanna Wallach et al.

Gender stereotypes are manifest in most of the world's languages and are consequently propagated or amplified by NLP systems. Although research has focused on mitigating gender stereotypes in English, the approaches that are commonly employed produce ungrammatical sentences in morphologically rich languages. We present a novel approach for converting between masculine-inflected and feminine-inflected sentences in such languages. For Spanish and Hebrew, our approach achieves F1 scores of 82% and 73% at the level of tags and accuracies of 90% and 87% at the level of forms. By evaluating our approach using four different languages, we show that, on average, it reduces gender stereotyping by a factor of 2.5 without any sacrifice to grammaticality.

Alexey Romanov, Maria De-Arteaga, Hanna Wallach et al.

There is a growing body of work that proposes methods for mitigating bias in machine learning systems. These methods typically rely on access to protected attributes such as race, gender, or age. However, this raises two significant challenges: (1) protected attributes may not be available or it may not be legal to use them, and (2) it is often desirable to simultaneously consider multiple protected attributes, as well as their intersections. In the context of mitigating bias in occupation classification, we propose a method for discouraging correlation between the predicted probability of an individual's true occupation and a word embedding of their name. This method leverages the societal biases that are encoded in word embeddings, eliminating the need for access to protected attributes. Crucially, it only requires access to individuals' names at training time and not at deployment time. We evaluate two variations of our proposed method using a large-scale dataset of online biographies. We find that both variations simultaneously reduce race and gender biases, with almost no reduction in the classifier's overall true positive rate.

Alexander Hoyle, Lawrence Wolf-Sonkin, Hanna Wallach et al.

When assigning quantitative labels to a dataset, different methodologies may rely on different scales. In particular, when assigning polarities to words in a sentiment lexicon, annotators may use binary, categorical, or continuous labels. Naturally, it is of interest to unify these labels from disparate scales to both achieve maximal coverage over words and to create a single, more robust sentiment lexicon while retaining scale coherence. We introduce a generative model of sentiment lexica to combine disparate scales into a common latent representation. We realize this model with a novel multi-view variational autoencoder (VAE), called SentiVAE. We evaluate our approach via a downstream text classification task involving nine English-Language sentiment analysis datasets; our representation outperforms six individual sentiment lexica, as well as a straightforward combination thereof.

Maria De-Arteaga, Alexey Romanov, Hanna Wallach et al.

We present a large-scale study of gender bias in occupation classification, a task where the use of machine learning may lead to negative outcomes on peoples' lives. We analyze the potential allocation harms that can result from semantic representation bias. To do so, we study the impact on occupation classification of including explicit gender indicators---such as first names and pronouns---in different semantic representations of online biographies. Additionally, we quantify the bias that remains when these indicators are "scrubbed," and describe proxy behavior that occurs in the absence of explicit gender indicators. As we demonstrate, differences in true positive rates between genders are correlated with existing gender imbalances in occupations, which may compound these imbalances.

Kenneth Holstein, Jennifer Wortman Vaughan, Hal Daumé et al.

The potential for machine learning (ML) systems to amplify social inequities and unfairness is receiving increasing popular and academic attention. A surge of recent work has focused on the development of algorithmic tools to assess and mitigate such unfairness. If these tools are to have a positive impact on industry practice, however, it is crucial that their design be informed by an understanding of real-world needs. Through 35 semi-structured interviews and an anonymous survey of 267 ML practitioners, we conduct the first systematic investigation of commercial product teams' challenges and needs for support in developing fairer ML systems. We identify areas of alignment and disconnect between the challenges faced by industry practitioners and solutions proposed in the fair ML research literature. Based on these findings, we highlight directions for future ML and HCI research that will better address industry practitioners' needs.

Timnit Gebru, Jamie Morgenstern, Briana Vecchione et al.

The machine learning community currently has no standardized process for documenting datasets, which can lead to severe consequences in high-stakes domains. To address this gap, we propose datasheets for datasets. In the electronics industry, every component, no matter how simple or complex, is accompanied with a datasheet that describes its operating characteristics, test results, recommended uses, and other information. By analogy, we propose that every dataset be accompanied with a datasheet that documents its motivation, composition, collection process, recommended uses, and so on. Datasheets for datasets will facilitate better communication between dataset creators and dataset consumers, and encourage the machine learning community to prioritize transparency and accountability.

Aaron Schein, Zhiwei Steven Wu, Alexandra Schofield et al.

We present a general method for privacy-preserving Bayesian inference in Poisson factorization, a broad class of models that includes some of the most widely used models in the social sciences. Our method satisfies limited precision local privacy, a generalization of local differential privacy, which we introduce to formulate privacy guarantees appropriate for sparse count data. We develop an MCMC algorithm that approximates the locally private posterior over model parameters given data that has been locally privatized by the geometric mechanism (Ghosh et al., 2012). Our solution is based on two insights: 1) a novel reinterpretation of the geometric mechanism in terms of the Skellam distribution (Skellam, 1946) and 2) a general theorem that relates the Skellam to the Bessel distribution (Yuan & Kalbfleisch, 2000). We demonstrate our method in two case studies on real-world email data in which we show that our method consistently outperforms the commonly-used naive approach, obtaining higher quality topics in text and more accurate link prediction in networks. On some tasks, our privacy-preserving method even outperforms non-private inference which conditions on the true data.

Alekh Agarwal, Alina Beygelzimer, Miroslav Dudík et al.

We present a systematic approach for achieving fairness in a binary classification setting. While we focus on two well-known quantitative definitions of fairness, our approach encompasses many other previously studied definitions as special cases. The key idea is to reduce fair classification to a sequence of cost-sensitive classification problems, whose solutions yield a randomized classifier with the lowest (empirical) error subject to the desired constraints. We introduce two reductions that work for any representation of the cost-sensitive classifier and compare favorably to prior baselines on a variety of data sets, while overcoming several of their disadvantages.

Forough Poursabzi-Sangdeh, Daniel G. Goldstein, Jake M. Hofman et al.

With machine learning models being increasingly used to aid decision making even in high-stakes domains, there has been a growing interest in developing interpretable models. Although many supposedly interpretable models have been proposed, there have been relatively few experimental studies investigating whether these models achieve their intended effects, such as making people more closely follow a model's predictions when it is beneficial for them to do so or enabling them to detect when a model has made a mistake. We present a sequence of pre-registered experiments (N=3,800) in which we showed participants functionally identical models that varied only in two factors commonly thought to make machine learning models more or less interpretable: the number of features and the transparency of the model (i.e., whether the model internals are clear or black box). Predictably, participants who saw a clear model with few features could better simulate the model's predictions. However, we did not find that participants more closely followed its predictions. Furthermore, showing participants a clear model meant that they were less able to detect and correct for the model's sizable mistakes, seemingly due to information overload. These counterintuitive findings emphasize the importance of testing over intuition when developing interpretable models.

Rishabh Mehrotra, Ashton Anderson, Fernando Diaz et al.

Many online services, such as search engines, social media platforms, and digital marketplaces, are advertised as being available to any user, regardless of their age, gender, or other demographic factors. However, there are growing concerns that these services may systematically underserve some groups of users. In this paper, we present a framework for internally auditing such services for differences in user satisfaction across demographic groups, using search engines as a case study. We first explain the pitfalls of naïvely comparing the behavioral metrics that are commonly used to evaluate search engines. We then propose three methods for measuring latent differences in user satisfaction from observed differences in evaluation metrics. To develop these methods, we drew on ideas from the causal inference literature and the multilevel modeling literature. Our framework is broadly applicable to other online services, and provides general insight into interpreting their evaluation metrics.

Aaron Schein, Mingyuan Zhou, Hanna Wallach

We introduce a new dynamical system for sequentially observed multivariate count data. This model is based on the gamma--Poisson construction---a natural choice for count data---and relies on a novel Bayesian nonparametric prior that ties and shrinks the model parameters, thus avoiding overfitting. We present an efficient MCMC inference algorithm that advances recent work on augmentation schemes for inference in negative binomial models. Finally, we demonstrate the model's inductive bias using a variety of real-world data sets, showing that it exhibits superior predictive performance over other models and infers highly interpretable latent structure.

Giacomo Zanella, Brenda Betancourt, Hanna Wallach et al.

Most generative models for clustering implicitly assume that the number of data points in each cluster grows linearly with the total number of data points. Finite mixture models, Dirichlet process mixture models, and Pitman--Yor process mixture models make this assumption, as do all other infinitely exchangeable clustering models. However, for some applications, this assumption is inappropriate. For example, when performing entity resolution, the size of each cluster should be unrelated to the size of the data set, and each cluster should contain a negligible fraction of the total number of data points. These applications require models that yield clusters whose sizes grow sublinearly with the size of the data set. We address this requirement by defining the microclustering property and introducing a new class of models that can exhibit this property. We compare models within this class to two commonly used clustering models using four entity-resolution data sets.

Rahul Goel, Sandeep Soni, Naman Goyal et al.

Language change is a complex social phenomenon, revealing pathways of communication and sociocultural influence. But, while language change has long been a topic of study in sociolinguistics, traditional linguistic research methods rely on circumstantial evidence, estimating the direction of change from differences between older and younger speakers. In this paper, we use a data set of several million Twitter users to track language changes in progress. First, we show that language change can be viewed as a form of social influence: we observe complex contagion for phonetic spellings and "netspeak" abbreviations (e.g., lol), but not for older dialect markers from spoken language. Next, we test whether specific types of social network connections are more influential than others, using a parametric Hawkes process model. We find that tie strength plays an important role: densely embedded social ties are significantly better conduits of linguistic influence. Geographic locality appears to play a more limited role: we find relatively little evidence to support the hypothesis that individuals are more influenced by geographically local social ties, even in their usage of geographical dialect markers.

Aaron Schein, Mingyuan Zhou, David M. Blei et al.

We introduce Bayesian Poisson Tucker decomposition (BPTD) for modeling country--country interaction event data. These data consist of interaction events of the form "country $i$ took action $a$ toward country $j$ at time $t$." BPTD discovers overlapping country--community memberships, including the number of latent communities. In addition, it discovers directed community--community interaction networks that are specific to "topics" of action types and temporal "regimes." We show that BPTD yields an efficient MCMC inference algorithm and achieves better predictive performance than related models. We also demonstrate that it discovers interpretable latent structure that agrees with our knowledge of international relations.

Jeffrey Miller, Brenda Betancourt, Abbas Zaidi et al.

Most generative models for clustering implicitly assume that the number of data points in each cluster grows linearly with the total number of data points. Finite mixture models, Dirichlet process mixture models, and Pitman--Yor process mixture models make this assumption, as do all other infinitely exchangeable clustering models. However, for some tasks, this assumption is undesirable. For example, when performing entity resolution, the size of each cluster is often unrelated to the size of the data set. Consequently, each cluster contains a negligible fraction of the total number of data points. Such tasks therefore require models that yield clusters whose sizes grow sublinearly with the size of the data set. We address this requirement by defining the \emph{microclustering property} and introducing a new model that exhibits this property. We compare this model to several commonly used clustering models by checking model fit using real and simulated data sets.

Aaron Schein, John Paisley, David M. Blei et al.

We present a Bayesian tensor factorization model for inferring latent group structures from dynamic pairwise interaction patterns. For decades, political scientists have collected and analyzed records of the form "country $i$ took action $a$ toward country $j$ at time $t$"---known as dyadic events---in order to form and test theories of international relations. We represent these event data as a tensor of counts and develop Bayesian Poisson tensor factorization to infer a low-dimensional, interpretable representation of their salient patterns. We demonstrate that our model's predictive performance is better than that of standard non-negative tensor factorization methods. We also provide a comparison of our variational updates to their maximum likelihood counterparts. In doing so, we identify a better way to form point estimates of the latent factors than that typically used in Bayesian Poisson matrix factorization. Finally, we showcase our model as an exploratory analysis tool for political scientists. We show that the inferred latent factor matrices capture interpretable multilateral relations that both conform to and inform our knowledge of international affairs.

Fangjian Guo, Charles Blundell, Hanna Wallach et al.

We present the Bayesian Echo Chamber, a new Bayesian generative model for social interaction data. By modeling the evolution of people's language usage over time, this model discovers latent influence relationships between them. Unlike previous work on inferring influence, which has primarily focused on simple temporal dynamics evidenced via turn-taking behavior, our model captures more nuanced influence relationships, evidenced via linguistic accommodation patterns in interaction content. The model, which is based on a discrete analog of the multivariate Hawkes process, permits a fully Bayesian inference algorithm. We validate our model's ability to discover latent influence patterns using transcripts of arguments heard by the US Supreme Court and the movie "12 Angry Men." We showcase our model's capabilities by using it to infer latent influence patterns from Federal Open Market Committee meeting transcripts, demonstrating state-of-the-art performance at uncovering social dynamics in group discussions.

Aaron Schein, Juston Moore, Hanna Wallach

Correlations between anomalous activity patterns can yield pertinent information about complex social processes: a significant deviation from normal behavior, exhibited simultaneously by multiple pairs of actors, provides evidence for some underlying relationship involving those pairs---i.e., a multilateral relation. We introduce a new nonparametric Bayesian latent variable model that explicitly captures correlations between anomalous interaction counts and uses these shared deviations from normal activity patterns to identify and characterize multilateral relations. We showcase our model's capabilities using the newly curated Global Database of Events, Location, and Tone, a dataset that has seen considerable interest in the social sciences and the popular press, but which has is largely unexplored by the machine learning community. We provide a detailed analysis of the latent structure inferred by our model and show that the multilateral relations correspond to major international events and long-term international relationships. These findings lead us to recommend our model for any data-driven analysis of interaction networks where dynamic interactions over the edges provide evidence for latent social structure.