Manish Raghavan

Rohan Alur, Loren Laine, Darrick K. Li et al. · mit

High-stakes prediction tasks (e.g., patient diagnosis) are often handled by trained human experts. A common source of concern about automation in these settings is that experts may exercise intuition that is difficult to model and/or have access to information (e.g., conversations with a patient) that is simply unavailable to a would-be algorithm. This raises a natural question whether human experts add value which could not be captured by an algorithmic predictor. We develop a statistical framework under which we can pose this question as a natural hypothesis test. Indeed, as our framework highlights, detecting human expertise is more subtle than simply comparing the accuracy of expert predictions to those made by a particular learning algorithm. Instead, we propose a simple procedure which tests whether expert predictions are statistically independent from the outcomes of interest after conditioning on the available inputs (`features'). A rejection of our test thus suggests that human experts may add value to any algorithm trained on the available data, and has direct implications for whether human-AI `complementarity' is achievable in a given prediction task. We highlight the utility of our procedure using admissions data collected from the emergency department of a large academic hospital system, where we show that physicians' admit/discharge decisions for patients with acute gastrointestinal bleeding (AGIB) appear to be incorporating information that is not available to a standard algorithmic screening tool. This is despite the fact that the screening tool is arguably more accurate than physicians' discretionary decisions, highlighting that -- even absent normative concerns about accountability or interpretability -- accuracy is insufficient to justify algorithmic automation.

Chris Hays, Zachary Schutzman, Manish Raghavan et al. · mit

Accurate bot detection is necessary for the safety and integrity of online platforms. It is also crucial for research on the influence of bots in elections, the spread of misinformation, and financial market manipulation. Platforms deploy infrastructure to flag or remove automated accounts, but their tools and data are not publicly available. Thus, the public must rely on third-party bot detection. These tools employ machine learning and often achieve near perfect performance for classification on existing datasets, suggesting bot detection is accurate, reliable and fit for use in downstream applications. We provide evidence that this is not the case and show that high performance is attributable to limitations in dataset collection and labeling rather than sophistication of the tools. Specifically, we show that simple decision rules -- shallow decision trees trained on a small number of features -- achieve near-state-of-the-art performance on most available datasets and that bot detection datasets, even when combined together, do not generalize well to out-of-sample datasets. Our findings reveal that predictions are highly dependent on each dataset's collection and labeling procedures rather than fundamental differences between bots and humans. These results have important implications for both transparency in sampling and labeling procedures and potential biases in research using existing bot detection tools for pre-processing.

Chris Hays, Ben Laufer, Solon Barocas et al.

Recent scholarship has argued that firms building data-driven decision systems in high-stakes domains like employment, credit, and housing should search for "less discriminatory algorithms" (LDAs) (Black et al., 2024). That is, for a given decision problem, firms considering deploying a model should make a good-faith effort to find equally performant models with lower disparate impact across social groups. Evidence from the literature on model multiplicity shows that randomness in training pipelines can lead to multiple models with the same performance, but meaningful variations in disparate impact. This suggests that developers can find LDAs simply by randomly retraining models. Firms cannot continue retraining forever, though, which raises the question: What constitutes a good-faith effort? In this paper, we formalize LDA search via model multiplicity as an optimal stopping problem, where a model developer with limited information wants to produce strong evidence that they have sufficiently explored the space of models. Our primary contribution is an adaptive stopping algorithm that yields a high-probability upper bound on the gains achievable from a continued search, allowing the developer to certify (e.g., to a court) that their search was sufficient. We provide a framework under which developers can impose stronger assumptions about the distribution of models, yielding correspondingly stronger bounds. We validate the method on real-world credit, employment and housing datasets.

Chris Hays, Rachel Li, Bailey Flanigan et al.

AI arenas, which rank generative models from pairwise preferences of users, are a popular method for measuring the relative performance of models in the course of their organic use. Because rankings are computed from noisy preferences, there is a concern that model producers can exploit this randomness by submitting many models (e.g., multiple variants of essentially the same model) and thereby artificially improve the rank of their top models. This can lead to degradations in the quality, and therefore the usefulness, of the ranking. In this paper, we begin by establishing, both theoretically and in simulations calibrated to data from the platform Arena (formerly LMArena, Chatbot Arena), conditions under which producers can benefit from submitting clones when their goal is to be ranked highly. We then propose a new mechanism for ranking models from pairwise comparisons, called You-Rank-We-Rank (YRWR). It requires that producers submit rankings over their own models and uses these rankings to correct statistical estimates of model quality. We prove that this mechanism is approximately clone-robust, in the sense that a producer cannot improve their rank much by doing anything other than submitting each of their unique models exactly once. Moreover, to the extent that model producers are able to correctly rank their own models, YRWR improves overall ranking accuracy. In further simulations, we show that indeed the mechanism is approximately clone-robust and quantify improvements to ranking accuracy, even under producer misranking.

Nathanael Jo, Manish Raghavan

Generative AI is quickly becoming an integral part of people's everyday workflows. Early evidence has shown that while generative AI can increase individual-level productivity, it does so at the cost of collective diversity, potentially narrowing the set of ideas and perspectives produced. Our research stands in contrast to this concern: through a pre-registered randomized control trial, we show that incentives mediate AI's homogenizing force in a creative writing task where participants can use AI interactively. Participants rewarded for originality relative to peers produce collectively more diverse writing than those rewarded for quality alone. This divergence is driven not by abandoning AI, but by how participants use it: those incentivized for originality incorporate fewer AI suggestions verbatim, relying on the model more selectively for brainstorming, proofreading, and targeted edits. Our results reveal that the effects of generative AI depend not only on the technology itself, but also the behavioral strategies and incentive structures surrounding its use.

Charlotte Park, Kate Donahue, Manish Raghavan

Generative AI models differ from traditional machine learning tools in that they allow users to provide as much or as little information as they choose in their inputs. This flexibility often leads users to omit certain details, relying on the models to infer and fill in under-specified information based on distributional knowledge of user preferences. Such inferences may privilege majority viewpoints and disadvantage users with atypical preferences, raising concerns about fairness. Unlike more traditional recommender systems, LLMs can explicitly solicit more information from users through natural language. However, while directly eliciting user preferences could increase personalization and mitigate inequality, excessive querying places a burden on users who value efficiency. We develop a stylized model of user-LLM interaction and develop an objective that captures tradeoff between user burden and preference representation. Building on the observation that individual preferences are often correlated, we analyze how AI systems should balance inference and elicitation, characterizing the optimal amount of information to solicit before content generation. Ultimately, we show that information elicitation can mitigate the systematic biases of preference inference, enabling the design of generative tools that better incorporate diverse user perspectives while maintaining efficiency. We complement this theoretical analysis with an empirical evaluation illustrating the model's predictions and exploring their practical implications.

Manish Raghavan

Recent evidence, both in the lab and in the wild, suggests that the use of generative artificial intelligence reduces the diversity of content produced. The use of the same or similar AI models appears to lead to more homogeneous behavior. Our work begins with the observation that there is a force pushing in the opposite direction: compe- tition. When producers compete with one another (e.g., for customers or attention), they are incentivized to create novel or unique content. We explore the impact com- petition has on both content diversity and overall social welfare. Through a formal game-theoretic model, we show that competitive markets select for diverse AI models, mitigating monoculture. We further show that a generative AI model that performs well in isolation (i.e., according to a benchmark) may fail to provide value in a compet- itive market. Our results highlight the importance of evaluating generative AI models across the breadth of their output distributions, particularly when they will be deployed in competitive environments. We validate our results empirically by using language models to play Scattergories, a word game in which players are rewarded for answers that are both correct and unique. Overall, our results suggest that homogenization due to generative AI is unlikely to persist in competitive markets, and instead, competition in downstream markets may drive diversification in AI model development

Benjamin Laufer, Manish Raghavan, Solon Barocas

Disparate impact doctrine offers an important legal apparatus for targeting discriminatory data-driven algorithmic decisions. A recent body of work has focused on conceptualizing one particular construct from this doctrine: the less discriminatory alternative, an alternative policy that reduces disparities while meeting the same business needs of a status quo or baseline policy. However, attempts to operationalize this construct in the algorithmic setting must grapple with some thorny challenges and ambiguities. In this paper, we attempt to raise and resolve important questions about less discriminatory algorithms (LDAs). How should we formally define LDAs, and how does this interact with different societal goals they might serve? And how feasible is it for firms or plaintiffs to computationally search for candidate LDAs? We find that formal LDA definitions face fundamental challenges when they attempt to evaluate and compare predictive models in the absence of held-out data. As a result, we argue that LDA definitions cannot be purely quantitative, and must rely on standards of "reasonableness." We then identify both mathematical and computational constraints on firms' ability to efficiently conduct a proactive search for LDAs, but we provide evidence that these limits are "weak" in a formal sense. By defining LDAs formally, we put forward a framework in which both firms and plaintiffs can search for alternative models that comport with societal goals.

Rohan Alur, Manish Raghavan, Devavrat Shah

We introduce a novel framework for incorporating human expertise into algorithmic predictions. Our approach leverages human judgment to distinguish inputs which are algorithmically indistinguishable, or "look the same" to predictive algorithms. We argue that this framing clarifies the problem of human-AI collaboration in prediction tasks, as experts often form judgments by drawing on information which is not encoded in an algorithm's training data. Algorithmic indistinguishability yields a natural test for assessing whether experts incorporate this kind of "side information", and further provides a simple but principled method for selectively incorporating human feedback into algorithmic predictions. We show that this method provably improves the performance of any feasible algorithmic predictor and precisely quantify this improvement. We find empirically that although algorithms often outperform their human counterparts on average, human judgment can improve algorithmic predictions on specific instances (which can be identified ex-ante). In an X-ray classification task, we find that this subset constitutes nearly $30\%$ of the patient population. Our approach provides a natural way of uncovering this heterogeneity and thus enabling effective human-AI collaboration.

Chris Hays, Manish Raghavan

Researchers and practitioners often wish to measure treatment effects in settings where units interact via markets and recommendation systems. In these settings, units are affected by certain shared states, like prices, algorithmic recommendations or social signals. We formalize this structure, calling it shared-state interference, and argue that our formulation captures many relevant applied settings. Our key modeling assumption is that individuals' potential outcomes are independent conditional on the shared state. We then prove an extension of a double machine learning (DML) theorem providing conditions for achieving efficient inference under shared-state interference. We also instantiate our general theorem in several models of interest where it is possible to efficiently estimate the average direct effect (ADE) or global average treatment effect (GATE).

Divya Shanmugam, Shuvom Sadhuka, Manish Raghavan et al.

It remains difficult to evaluate machine learning classifiers in the absence of a large, labeled dataset. While labeled data can be prohibitively expensive or impossible to obtain, unlabeled data is plentiful. Here, we introduce Semi-Supervised Model Evaluation (SSME), a method that uses both labeled and unlabeled data to evaluate machine learning classifiers. SSME is the first evaluation method to take advantage of the fact that: (i) there are frequently multiple classifiers for the same task, (ii) continuous classifier scores are often available for all classes, and (iii) unlabeled data is often far more plentiful than labeled data. The key idea is to use a semi-supervised mixture model to estimate the joint distribution of ground truth labels and classifier predictions. We can then use this model to estimate any metric that is a function of classifier scores and ground truth labels (e.g., accuracy or expected calibration error). We present experiments in four domains where obtaining large labeled datasets is often impractical: (1) healthcare, (2) content moderation, (3) molecular property prediction, and (4) image annotation. Our results demonstrate that SSME estimates performance more accurately than do competing methods, reducing error by 5.1x relative to using labeled data alone and 2.4x relative to the next best competing method. SSME also improves accuracy when evaluating performance across subsets of the test distribution (e.g., specific demographic subgroups) and when evaluating the performance of language models.

Rohan Alur, Loren Laine, Darrick K. Li et al.

We introduce a novel framework for human-AI collaboration in prediction and decision tasks. Our approach leverages human judgment to distinguish inputs which are algorithmically indistinguishable, or "look the same" to any feasible predictive algorithm. We argue that this framing clarifies the problem of human-AI collaboration in prediction and decision tasks, as experts often form judgments by drawing on information which is not encoded in an algorithm's training data. Algorithmic indistinguishability yields a natural test for assessing whether experts incorporate this kind of "side information", and further provides a simple but principled method for selectively incorporating human feedback into algorithmic predictions. We show that this method provably improves the performance of any feasible algorithmic predictor and precisely quantify this improvement. We demonstrate the utility of our framework in a case study of emergency room triage decisions, where we find that although algorithmic risk scores are highly competitive with physicians, there is strong evidence that physician judgments provide signal which could not be replicated by any predictive algorithm. This insight yields a range of natural decision rules which leverage the complementary strengths of human experts and predictive algorithms.

Brian Hedden, Manish Raghavan

Algorithmic decision-making is replacing idiosyncratic human judgment in domains such as hiring, lending, and criminal justice. This shift promises increased consistency, but many scholars worry that it can go too far. They warn of the dangers of algorithmic monoculture, in which all decisions across a domain are made using a single algorithm. We systematically evaluate a range of objections to monoculture, formalizing and rigorously assessing familiar critiques alongside novel ones. These objections concern systematic exclusion, agency and gaming, and information aggregation and exploration. We conclude that monoculture is less problematic than its critics have supposed: commonly cited objections fail, and while other objections have some force, they are not decisive against monoculture in general.

Rohan Alur, Chris Hays, Manish Raghavan et al.

In this technical note, we study the problem of inverse permutation learning in decoder-only transformers. Given a permutation and a string to which that permutation has been applied, the model is tasked with producing the original (``canonical'') string. We argue that this task models a natural robustness property across a variety of reasoning tasks, including long-context retrieval, multiple choice QA and in-context learning. Our primary contribution is an impossibility result: we show that an arbitrary depth, decoder-only transformer cannot learn this task. This result concerns the expressive capacity of decoder-only transformer models and is agnostic to training dynamics or sample complexity. We give a pair of alternative constructions under which inverse permutation learning is feasible. The first of these highlights the fundamental role of the causal attention mask, and reveals a gap between the expressivity of encoder-decoder transformers and the more popular decoder-only architecture. The latter result is more surprising: we show that simply padding the input with ``scratch tokens" yields a construction under which inverse permutation learning is possible. We conjecture that this may suggest an alternative mechanism by which chain-of-thought prompting or, more generally, intermediate ``thinking'' tokens can enable reasoning in large language models, even when these tokens encode no meaningful semantic information (e.g., the results of intermediate computations).

Chloé Bakalar, Renata Barreto, Stevie Bergman et al.

Many technical approaches have been proposed for ensuring that decisions made by machine learning systems are fair, but few of these proposals have been stress-tested in real-world systems. This paper presents an example of one team's approach to the challenge of applying algorithmic fairness approaches to complex production systems within the context of a large technology company. We discuss how we disentangle normative questions of product and policy design (like, "how should the system trade off between different stakeholders' interests and needs?") from empirical questions of system implementation (like, "is the system achieving the desired tradeoff in practice?"). We also present an approach for answering questions of the latter sort, which allows us to measure how machine learning systems and human labelers are making these tradeoffs across different relevant groups. We hope our experience integrating fairness tools and approaches into large-scale and complex production systems will be useful to other practitioners facing similar challenges, and illuminating to academics and researchers looking to better address the needs of practitioners.

Jon Kleinberg, Manish Raghavan

As algorithms are increasingly applied to screen applicants for high-stakes decisions in employment, lending, and other domains, concerns have been raised about the effects of algorithmic monoculture, in which many decision-makers all rely on the same algorithm. This concern invokes analogies to agriculture, where a monocultural system runs the risk of severe harm from unexpected shocks. Here we show that the dangers of algorithmic monoculture run much deeper, in that monocultural convergence on a single algorithm by a group of decision-making agents, even when the algorithm is more accurate for any one agent in isolation, can reduce the overall quality of the decisions being made by the full collection of agents. Unexpected shocks are therefore not needed to expose the risks of monoculture; it can hurt accuracy even under "normal" operations, and even for algorithms that are more accurate when used by only a single decision-maker. Our results rely on minimal assumptions, and involve the development of a probabilistic framework for analyzing systems that use multiple noisy estimates of a set of alternatives.

Jessie Finocchiaro, Roland Maio, Faidra Monachou et al.

Decision-making systems increasingly orchestrate our world: how to intervene on the algorithmic components to build fair and equitable systems is therefore a question of utmost importance; one that is substantially complicated by the context-dependent nature of fairness and discrimination. Modern decision-making systems that involve allocating resources or information to people (e.g., school choice, advertising) incorporate machine-learned predictions in their pipelines, raising concerns about potential strategic behavior or constrained allocation, concerns usually tackled in the context of mechanism design. Although both machine learning and mechanism design have developed frameworks for addressing issues of fairness and equity, in some complex decision-making systems, neither framework is individually sufficient. In this paper, we develop the position that building fair decision-making systems requires overcoming these limitations which, we argue, are inherent to each field. Our ultimate objective is to build an encompassing framework that cohesively bridges the individual frameworks of mechanism design and machine learning. We begin to lay the ground work towards this goal by comparing the perspective each discipline takes on fair decision-making, teasing out the lessons each field has taught and can teach the other, and highlighting application domains that require a strong collaboration between these disciplines.

Manish Raghavan, Aleksandrs Slivkins, Jennifer Wortman Vaughan et al.

Online learning algorithms, widely used to power search and content optimization on the web, must balance exploration and exploitation, potentially sacrificing the experience of current users in order to gain information that will lead to better decisions in the future. While necessary in the worst case, explicit exploration has a number of disadvantages compared to the greedy algorithm that always "exploits" by choosing an action that currently looks optimal. We ask under what conditions inherent diversity in the data makes explicit exploration unnecessary. We build on a recent line of work on the smoothed analysis of the greedy algorithm in the linear contextual bandits model. We improve on prior results to show that a greedy approach almost matches the best possible Bayesian regret rate of any other algorithm on the same problem instance whenever the diversity conditions hold, and that this regret is at most $\tilde O(T^{1/3})$.

Manish Raghavan, Solon Barocas, Jon Kleinberg et al.

There has been rapidly growing interest in the use of algorithms in hiring, especially as a means to address or mitigate bias. Yet, to date, little is known about how these methods are used in practice. How are algorithmic assessments built, validated, and examined for bias? In this work, we document and analyze the claims and practices of companies offering algorithms for employment assessment. In particular, we identify vendors of algorithmic pre-employment assessments (i.e., algorithms to screen candidates), document what they have disclosed about their development and validation procedures, and evaluate their practices, focusing particularly on efforts to detect and mitigate bias. Our analysis considers both technical and legal perspectives. Technically, we consider the various choices vendors make regarding data collection and prediction targets, and explore the risks and trade-offs that these choices pose. We also discuss how algorithmic de-biasing techniques interface with, and create challenges for, antidiscrimination law.

Jon Kleinberg, Manish Raghavan

Algorithms are often used to produce decision-making rules that classify or evaluate individuals. When these individuals have incentives to be classified a certain way, they may behave strategically to influence their outcomes. We develop a model for how strategic agents can invest effort in order to change the outcomes they receive, and we give a tight characterization of when such agents can be incentivized to invest specified forms of effort into improving their outcomes as opposed to "gaming" the classifier. We show that whenever any "reasonable" mechanism can do so, a simple linear mechanism suffices.

Manish Raghavan, Aleksandrs Slivkins, Jennifer Wortman Vaughan et al.

Online learning algorithms, widely used to power search and content optimization on the web, must balance exploration and exploitation, potentially sacrificing the experience of current users for information that will lead to better decisions in the future. Recently, concerns have been raised about whether the process of exploration could be viewed as unfair, placing too much burden on certain individuals or groups. Motivated by these concerns, we initiate the study of the externalities of exploration - the undesirable side effects that the presence of one party may impose on another - under the linear contextual bandits model. We introduce the notion of a group externality, measuring the extent to which the presence of one population of users impacts the rewards of another. We show that this impact can in some cases be negative, and that, in a certain sense, no algorithm can avoid it. We then study externalities at the individual level, interpreting the act of exploration as an externality imposed on the current user of a system by future users. This drives us to ask under what conditions inherent diversity in the data makes explicit exploration unnecessary. We build on a recent line of work on the smoothed analysis of the greedy algorithm that always chooses the action that currently looks optimal, improving on prior results to show that a greedy approach almost matches the best possible Bayesian regret rate of any other algorithm on the same problem instance whenever the diversity conditions hold, and that this regret is at most $\tilde{O}(T^{1/3})$. Returning to group-level effects, we show that under the same conditions, negative group externalities essentially vanish under the greedy algorithm. Together, our results uncover a sharp contrast between the high externalities that exist in the worst case, and the ability to remove all externalities if the data is sufficiently diverse.

Manish Raghavan, Ashton Anderson, Jon Kleinberg

The surge in political information, discourse, and interaction has been one of the most important developments in social media over the past several years. There is rich structure in the interaction among different viewpoints on the ideological spectrum. However, we still have only a limited analytical vocabulary for expressing the ways in which these viewpoints interact. In this paper, we develop network-based methods that operate on the ways in which users share content; we construct \emph{invocation graphs} on Web domains showing the extent to which pages from one domain are invoked by users to reply to posts containing pages from other domains. When we locate the domains on a political spectrum induced from the data, we obtain an embedded graph showing how these interaction links span different distances on the spectrum. The structure of this embedded network, and its evolution over time, helps us derive macro-level insights about how political interaction unfolded through 2016, leading up to the US Presidential election. In particular, we find that the domains invoked in replies spanned increasing distances on the spectrum over the months approaching the election, and that there was clear asymmetry between the left-to-right and right-to-left patterns of linkage.

Jon Kleinberg, Manish Raghavan

Over the past two decades, the notion of implicit bias has come to serve as an important component in our understanding of discrimination in activities such as hiring, promotion, and school admissions. Research on implicit bias posits that when people evaluate others -- for example, in a hiring context -- their unconscious biases about membership in particular groups can have an effect on their decision-making, even when they have no deliberate intention to discriminate against members of these groups. A growing body of experimental work has pointed to the effect that implicit bias can have in producing adverse outcomes. Here we propose a theoretical model for studying the effects of implicit bias on selection decisions, and a way of analyzing possible procedural remedies for implicit bias within this model. A canonical situation represented by our model is a hiring setting: a recruiting committee is trying to choose a set of finalists to interview among the applicants for a job, evaluating these applicants based on their future potential, but their estimates of potential are skewed by implicit bias against members of one group. In this model, we show that measures such as the Rooney Rule, a requirement that at least one of the finalists be chosen from the affected group, can not only improve the representation of this affected group, but also lead to higher payoffs in absolute terms for the organization performing the recruiting. However, identifying the conditions under which such measures can lead to improved payoffs involves subtle trade-offs between the extent of the bias and the underlying distribution of applicant characteristics, leading to novel theoretical questions about order statistics in the presence of probabilistic side information.

Geoff Pleiss, Manish Raghavan, Felix Wu et al.

The machine learning community has become increasingly concerned with the potential for bias and discrimination in predictive models. This has motivated a growing line of work on what it means for a classification procedure to be "fair." In this paper, we investigate the tension between minimizing error disparity across different population groups while maintaining calibrated probability estimates. We show that calibration is compatible only with a single error constraint (i.e. equal false-negatives rates across groups), and show that any algorithm that satisfies this relaxation is no better than randomizing a percentage of predictions for an existing classifier. These unsettling findings, which extend and generalize existing results, are empirically confirmed on several datasets.

Jon Kleinberg, Sendhil Mullainathan, Manish Raghavan

Recent discussion in the public sphere about algorithmic classification has involved tension between competing notions of what it means for a probabilistic classification to be fair to different groups. We formalize three fairness conditions that lie at the heart of these debates, and we prove that except in highly constrained special cases, there is no method that can satisfy these three conditions simultaneously. Moreover, even satisfying all three conditions approximately requires that the data lie in an approximate version of one of the constrained special cases identified by our theorem. These results suggest some of the ways in which key notions of fairness are incompatible with each other, and hence provide a framework for thinking about the trade-offs between them.