arXiv subjects

Sendhil Mullainathan

Publications and source records attributed to Sendhil Mullainathan.

At least 19 recordsLinked to original sources

arXiv2025

AI Behavioral Science

We outline a foundation for a new field of ``AI Behavioral Science,'' covering three perspectives. First, as AI becomes ubiquitous and is increasingly proprietary and opaque, it becomes vital to develop techniques for assessing AI behavior. We outline how tools developed to assess people's behaviors by social scientists can be used to assess and infer AI's behaviors biases, tendencies, and heuristics. Second, we also discuss how AI can change the ways in which we learn about human behavior. Beyond its computational power, AI offers new techniques for simulating, inferring, and predicting human behaviors that we outline and discuss. Third, as humans and AI are interacting in increasingly complex and intertwined systems, we need to understand the implications for the resulting economic and political outcomes. We outline issues that are increasingly pressing concerning the future of human-AI interactions and potential changes and disruptions that can ensue.

cs.HC↗

arXiv2025

What Has a Foundation Model Found? Using Inductive Bias to Probe for World Models

Foundation models are premised on the idea that sequence prediction can uncover deeper domain understanding, much like how Kepler's predictions of planetary motion later led to the discovery of Newtonian mechanics. However, evaluating whether these models truly capture deeper structure remains a challenge. We develop a technique for evaluating foundation models that examines how they adapt to synthetic datasets generated from some postulated world model. Our technique measures whether the foundation model's inductive bias aligns with the world model, and so we refer to it as an inductive bias probe. Across multiple domains, we find that foundation models can excel at their training tasks yet fail to develop inductive biases towards the underlying world model when adapted to new tasks. We particularly find that foundation models trained on orbital trajectories consistently fail to apply Newtonian mechanics when adapted to new physics tasks. Further analysis reveals that these models behave as if they develop task-specific heuristics that fail to generalize.

cs.LG↗

arXiv2025

Potemkin Understanding in Large Language Models

Large language models (LLMs) are regularly evaluated using benchmark datasets. But what justifies making inferences about an LLM's capabilities based on its answers to a curated set of questions? This paper first introduces a formal framework to address this question. The key is to note that the benchmarks used to test LLMs -- such as AP exams -- are also those used to test people. However, this raises an implication: these benchmarks are only valid tests if LLMs misunderstand concepts in ways that mirror human misunderstandings. Otherwise, success on benchmarks only demonstrates potemkin understanding: the illusion of understanding driven by answers irreconcilable with how any human would interpret a concept. We present two procedures for quantifying the existence of potemkins: one using a specially designed benchmark in three domains, the other using a general procedure that provides a lower-bound on their prevalence. We find that potemkins are ubiquitous across models, tasks, and domains. We also find that these failures reflect not just incorrect understanding, but deeper internal incoherence in concept representations.

cs.CL↗

arXiv2025

What's Producible May Not Be Reachable: Measuring the Steerability of Generative Models

How should we evaluate the quality of generative models? Many existing metrics focus on a model's producibility, i.e. the quality and breadth of outputs it can generate. However, the actual value from using a generative model stems not just from what it can produce but whether a user with a specific goal can produce an output that satisfies that goal. We refer to this property as steerability. In this paper, we first introduce a mathematical decomposition for quantifying steerability independently from producibility. Steerability is more challenging to evaluate than producibility because it requires knowing a user's goals. We address this issue by creating a benchmark task that relies on one key idea: sample an output from a generative model and ask users to reproduce it. We implement this benchmark in user studies of text-to-image and large language models. Despite the ability of these models to produce high-quality outputs, they all perform poorly on steerability. These results suggest that we need to focus on improving the steerability of generative models. We show such improvements are indeed possible: simple image-based steering mechanisms achieve more than 2x improvement on this benchmark.

cs.LG↗

arXiv2024

Large Language Models: An Applied Econometric Framework

Large language models (LLMs) enable researchers to analyze text at unprecedented scale and minimal cost. Researchers can now revisit old questions and tackle novel ones with rich data. We provide an econometric framework for realizing this potential in two empirical uses. For prediction problems -- forecasting outcomes from text -- valid conclusions require ``no training leakage'' between the LLM's training data and the researcher's sample, which can be enforced through careful model choice and research design. For estimation problems -- automating the measurement of economic concepts for downstream analysis -- valid downstream inference requires combining LLM outputs with a small validation sample to deliver consistent and precise estimates. Absent a validation sample, researchers cannot assess possible errors in LLM outputs, and consequently seemingly innocuous choices (which model, which prompt) can produce dramatically different parameter estimates. When used appropriately, LLMs are powerful tools that can expand the frontier of empirical economics.

econ.EM↗

arXiv2024

Evaluating the World Model Implicit in a Generative Model

Recent work suggests that large language models may implicitly learn world models. How should we assess this possibility? We formalize this question for the case where the underlying reality is governed by a deterministic finite automaton. This includes problems as diverse as simple logical reasoning, geographic navigation, game-playing, and chemistry. We propose new evaluation metrics for world model recovery inspired by the classic Myhill-Nerode theorem from language theory. We illustrate their utility in three domains: game playing, logic puzzles, and navigation. In all domains, the generative models we consider do well on existing diagnostics for assessing world models, but our evaluation metrics reveal their world models to be far less coherent than they appear. Such incoherence creates fragility: using a generative model to solve related but subtly different tasks can lead to failures. Building generative models that meaningfully capture the underlying logic of the domains they model would be immensely valuable; our results suggest new ways to assess how close a given model is to that goal.

cs.CL↗

arXiv2024

Do Large Language Models Perform the Way People Expect? Measuring the Human Generalization Function

What makes large language models (LLMs) impressive is also what makes them hard to evaluate: their diversity of uses. To evaluate these models, we must understand the purposes they will be used for. We consider a setting where these deployment decisions are made by people, and in particular, people's beliefs about where an LLM will perform well. We model such beliefs as the consequence of a human generalization function: having seen what an LLM gets right or wrong, people generalize to where else it might succeed. We collect a dataset of 19K examples of how humans make generalizations across 79 tasks from the MMLU and BIG-Bench benchmarks. We show that the human generalization function can be predicted using NLP methods: people have consistent structured ways to generalize. We then evaluate LLM alignment with the human generalization function. Our results show that -- especially for cases where the cost of mistakes is high -- more capable models (e.g. GPT-4) can do worse on the instances people choose to use them for, exactly because they are not aligned with the human generalization function.

cs.CL↗

arXiv2024

From Predictive Algorithms to Automatic Generation of Anomalies

How can we extract theoretical insights from machine learning algorithms? We take a familiar lesson: researchers often turn their intuitions into theoretical insights by constructing "anomalies" -- specific examples highlighting hypothesized flaws in a theory, such as the Allais paradox and the Kahneman-Tversky choice experiments for expected utility. We develop procedures that replace researchers' intuitions with predictive algorithms: given a predictive algorithm and a theory, our procedures automatically generate anomalies for that theory. We illustrate our procedures with a concrete application: generating anomalies for expected utility theory. Based on a neural network that accurately predicts lottery choices, our procedures recover known anomalies for expected utility theory and discover new ones absent from existing work. In incentivized experiments, subjects violate expected utility theory on these algorithmically generated anomalies at rates similar to the Allais paradox and common ratio effect.

econ.EM↗

arXiv2024

Language Generation in the Limit

Although current large language models are complex, the most basic specifications of the underlying language generation problem itself are simple to state: given a finite set of training samples from an unknown language, produce valid new strings from the language that don't already appear in the training data. Here we ask what we can conclude about language generation using only this specification, without further assumptions. In particular, suppose that an adversary enumerates the strings of an unknown target language L that is known only to come from one of a possibly infinite list of candidates. A computational agent is trying to learn to generate from this language; we say that the agent generates from L in the limit if after some finite point in the enumeration of L, the agent is able to produce new elements that come exclusively from L and that have not yet been presented by the adversary. Our main result is that there is an agent that is able to generate in the limit for every countable list of candidate languages. This contrasts dramatically with negative results due to Gold and Angluin in a well-studied model of language learning where the goal is to identify an unknown language from samples; the difference between these results suggests that identifying a language is a fundamentally different problem than generating from it.

cs.DS↗

arXiv2023

Using large language models to promote health equity

Advances in large language models (LLMs) have driven an explosion of interest about their societal impacts. Much of the discourse around how they will impact social equity has been cautionary or negative, focusing on questions like "how might LLMs be biased and how would we mitigate those biases?" This is a vital discussion: the ways in which AI generally, and LLMs specifically, can entrench biases have been well-documented. But equally vital, and much less discussed, is the more opportunity-focused counterpoint: "what promising applications do LLMs enable that could promote equity?" If LLMs are to enable a more equitable world, it is not enough just to play defense against their biases and failure modes. We must also go on offense, applying them positively to equity-enhancing use cases to increase opportunities for underserved groups and reduce societal discrimination. There are many choices which determine the impact of AI, and a fundamental choice very early in the pipeline is the problems we choose to apply it to. If we focus only later in the pipeline -- making LLMs marginally more fair as they facilitate use cases which intrinsically entrench power -- we will miss an important opportunity to guide them to equitable impacts. Here, we highlight the emerging potential of LLMs to promote equity by presenting four newly possible, promising research directions, while keeping risks and cautionary points in clear view.

cs.CY↗

arXiv2023

The Challenge of Understanding What Users Want: Inconsistent Preferences and Engagement Optimization

Online platforms have a wealth of data, run countless experiments and use industrial-scale algorithms to optimize user experience. Despite this, many users seem to regret the time they spend on these platforms. One possible explanation is misaligned incentives: platforms are not optimizing for user happiness. We suggest the problem runs deeper, transcending the specific incentives of any particular platform, and instead stems from a mistaken foundational assumption: To understand what users want, platforms look at what users do. Yet research has demonstrated, and personal experience affirms, that we often make choices in the moment that are inconsistent with what we actually want. In this work, we develop a model of media consumption where users have inconsistent preferences. We consider a platform which simply wants to maximize user utility, but only observes user engagement. We show how our model of users' preference inconsistencies produces phenomena that are familiar from everyday experience, but difficult to capture in traditional user interaction models. A key ingredient in our model is a formulation for how platforms determine what to show users: they optimize over a large set of potential content (the content manifold) parametrized by underlying features of the content. Whether improving engagement improves user welfare depends on the direction of movement in the content manifold: for certain directions of change, increasing engagement makes users less happy, while in other directions, increasing engagement makes users happier. We characterize the structure of content manifolds for which increasing engagement fails to increase user utility. By linking these effects to abstractions of platform design choices, our model thus creates a theoretical framework and vocabulary in which to explore interactions between design, behavioral science, and social media.

cs.SI↗

arXiv2020

Characterizing the Value of Information in Medical Notes

Machine learning models depend on the quality of input data. As electronic health records are widely adopted, the amount of data in health care is growing, along with complaints about the quality of medical notes. We use two prediction tasks, readmission prediction and in-hospital mortality prediction, to characterize the value of information in medical notes. We show that as a whole, medical notes only provide additional predictive power over structured information in readmission prediction. We further propose a probing framework to select parts of notes that enable more accurate predictions than using all notes, despite that the selected information leads to a distribution shift from the training data ("all notes"). Finally, we demonstrate that models trained on the selected valuable information achieve even better predictive performance, with only 6.8% of all the tokens for readmission prediction.

cs.CL↗

arXiv2020

Quantifying the Causal Effects of Conversational Tendencies

Understanding what leads to effective conversations can aid the design of better computer-mediated communication platforms. In particular, prior observational work has sought to identify behaviors of individuals that correlate to their conversational efficiency. However, translating such correlations to causal interpretations is a necessary step in using them in a prescriptive fashion to guide better designs and policies. In this work, we formally describe the problem of drawing causal links between conversational behaviors and outcomes. We focus on the task of determining a particular type of policy for a text-based crisis counseling platform: how best to allocate counselors based on their behavioral tendencies exhibited in their past conversations. We apply arguments derived from causal inference to underline key challenges that arise in conversational settings where randomized trials are hard to implement. Finally, we show how to circumvent these inference challenges in our particular domain, and illustrate the potential benefits of an allocation policy informed by the resulting prescriptive information.

cs.CL↗

arXiv2020

The Data Station: Combining Data, Compute, and Market Forces

This paper introduces Data Stations, a new data architecture that we are designing to tackle some of the most challenging data problems that we face today: access to sensitive data; data discovery and integration; and governance and compliance. Data Stations depart from modern data lakes in that both data and derived data products, such as machine learning models, are sealed and cannot be directly seen, accessed, or downloaded by anyone. Data Stations do not deliver data to users; instead, users bring questions to data. This inversion of the usual relationship between data and compute mitigates many of the security risks that are otherwise associated with sharing and working with sensitive data. Data Stations are designed following the principle that many data problems require human involvement, and that incentives are the key to obtaining such involvement. To that end, Data Stations implement market designs to create, manage, and coordinate the use of incentives. We explain the motivation for this new kind of platform and its design.

cs.DB↗

arXiv2019

Measuring the Completeness of Theories

We use machine learning to provide a tractable measure of the amount of predictable variation in the data that a theory captures, which we call its "completeness." We apply this measure to three problems: assigning certain equivalents to lotteries, initial play in games, and human generation of random sequences. We discover considerable variation in the completeness of existing models, which sheds light on whether to focus on developing better models with the same features or instead to look for new features that will improve predictions. We also illustrate how and why completeness varies with the experiments considered, which highlights the role played in choosing which experiments to run.

econ.TH↗

arXiv2019

Simplicity Creates Inequity: Implications for Fairness, Stereotypes, and Interpretability

Algorithms are increasingly used to aid, or in some cases supplant, human decision-making, particularly for decisions that hinge on predictions. As a result, two additional features in addition to prediction quality have generated interest: (i) to facilitate human interaction and understanding with these algorithms, we desire prediction functions that are in some fashion simple or interpretable; and (ii) because they influence consequential decisions, we also want them to produce equitable allocations. We develop a formal model to explore the relationship between the demands of simplicity and equity. Although the two concepts appear to be motivated by qualitatively distinct goals, we show a fundamental inconsistency between them. Specifically, we formalize a general framework for producing simple prediction functions, and in this framework we establish two basic results. First, every simple prediction function is strictly improvable: there exists a more complex prediction function that is both strictly more efficient and also strictly more equitable. Put another way, using a simple prediction function both reduces utility for disadvantaged groups and reduces overall welfare relative to other options. Second, we show that simple prediction functions necessarily create incentives to use information about individuals' membership in a disadvantaged group --- incentives that weren't present before simplification, and that work against these individuals. Thus, simplicity transforms disadvantage into bias against the disadvantaged group. Our results are not only about algorithms but about any process that produces simple models, and as such they connect to the psychology of stereotypes and to an earlier economics literature on statistical discrimination.

cs.LG↗

arXiv2019

Direct Uncertainty Prediction for Medical Second Opinions

The issue of disagreements amongst human experts is a ubiquitous one in both machine learning and medicine. In medicine, this often corresponds to doctor disagreements on a patient diagnosis. In this work, we show that machine learning models can be trained to give uncertainty scores to data instances that might result in high expert disagreements. In particular, they can identify patient cases that would benefit most from a medical second opinion. Our central methodological finding is that Direct Uncertainty Prediction (DUP), training a model to predict an uncertainty score directly from the raw patient features, works better than Uncertainty Via Classification, the two-step process of training a classifier and postprocessing the output distribution to give an uncertainty score. We show this both with a theoretical result, and on extensive evaluations on a large scale medical imaging application.

cs.LG↗

arXiv2019

Machine-Learning Tests for Effects on Multiple Outcomes

In this paper we present tools for applied researchers that re-purpose off-the-shelf methods from the computer-science field of machine learning to create a "discovery engine" for data from randomized controlled trials (RCTs). The applied problem we seek to solve is that economists invest vast resources into carrying out RCTs, including the collection of a rich set of candidate outcome measures. But given concerns about inference in the presence of multiple testing, economists usually wind up exploring just a small subset of the hypotheses that the available data could be used to test. This prevents us from extracting as much information as possible from each RCT, which in turn impairs our ability to develop new theories or strengthen the design of policy interventions. Our proposed solution combines the basic intuition of reverse regression, where the dependent variable of interest now becomes treatment assignment itself, with methods from machine learning that use the data themselves to flexibly identify whether there is any function of the outcomes that predicts (or has signal about) treatment group status. This leads to correctly-sized tests with appropriate $p$-values, which also have the important virtue of being easy to implement in practice. One open challenge that remains with our work is how to meaningfully interpret the signal that these methods find.

stat.ML↗