Jake M. Hofman

Sayash Kapoor, Emily Cantrell, Kenny Peng et al. · princeton

Machine learning (ML) methods are proliferating in scientific research. However, the adoption of these methods has been accompanied by failures of validity, reproducibility, and generalizability. These failures can hinder scientific progress, lead to false consensus around invalid claims, and undermine the credibility of ML-based science. ML methods are often applied and fail in similar ways across disciplines. Motivated by this observation, our goal is to provide clear reporting standards for ML-based science. Drawing from an extensive review of past literature, we present the REFORMS checklist ($\textbf{Re}$porting Standards $\textbf{For}$ $\textbf{M}$achine Learning Based $\textbf{S}$cience). It consists of 32 questions and a paired set of guidelines. REFORMS was developed based on a consensus of 19 researchers across computer science, data science, mathematics, social sciences, and biomedical sciences. REFORMS can serve as a resource for researchers when designing and implementing a study, for referees when reviewing papers, and for journals when enforcing standards for transparency and reproducibility.

Jamelle Watson-Daniels, Solon Barocas, Jake M. Hofman et al.

Prediction models have been widely adopted as the basis for decision-making in domains as diverse as employment, education, lending, and health. Yet, few real world problems readily present themselves as precisely formulated prediction tasks. In particular, there are often many reasonable target variable options. Prior work has argued that this is an important and sometimes underappreciated choice, and has also shown that target choice can have a significant impact on the fairness of the resulting model. However, the existing literature does not offer a formal framework for characterizing the extent to which target choice matters in a particular task. Our work fills this gap by drawing connections between the problem of target choice and recent work on predictive multiplicity. Specifically, we introduce a conceptual and computational framework for assessing how the choice of target affects individuals' outcomes and selection rate disparities across groups. We call this multi-target multiplicity. Along the way, we refine the study of single-target multiplicity by introducing notions of multiplicity that respect resource constraints -- a feature of many real-world tasks that is not captured by existing notions of predictive multiplicity. We apply our methods on a healthcare dataset, and show that the level of multiplicity that stems from target variable choice can be greater than that stemming from nearly-optimal models of a single target.

Jake M. Hofman, Angelos Chatzimparmpas, Amit Sharma et al.

Amid rising concerns of reproducibility and generalizability in predictive modeling, we explore the possibility and potential benefits of introducing pre-registration to the field. Despite notable advancements in predictive modeling, spanning core machine learning tasks to various scientific applications, challenges such as overlooked contextual factors, data-dependent decision-making, and unintentional re-use of test data have raised questions about the integrity of results. To address these issues, we propose adapting pre-registration practices from explanatory modeling to predictive modeling. We discuss current best practices in predictive modeling and their limitations, introduce a lightweight pre-registration template, and present a qualitative study with machine learning researchers to gain insight into the effectiveness of pre-registration in preventing biased estimates and promoting more reliable research outcomes. We conclude by exploring the scope of problems that pre-registration can address in predictive modeling and acknowledging its limitations within this context.

Brendan Lucier, Nicole Immorlica, Markus Mobius et al.

Motivated by agentic markets -- two-sided markets in which consumers and businesses are assisted by AI tools that facilitate consumers' search -- we study the impact of improved search technology on learning and welfare in markets. We put forth a model where consumers engage in costly search to acquire signals of product fit prior to purchase. The market tracks indications of fit for searched products and indications of quality for chosen products, thereby guiding searches. We characterize the long-run steady-state of the resulting dynamics as well as the impact of improving search technology. We find cheaper search improves learning and consumer surplus, whereas more informative search can degrade both unless the market learns as much as consumers about the products by, for example, ``reading the transcripts'' of agentic conversations. Finally, we consider the impact of search improvements on how businesses set prices. At equilibrium prices in symmetric markets, consumer surplus is improved by cheaper search but may be decreased by more informative search, due to weakened inter-business competition.

Hancheng Cao, Sofia Eleni Spatharioti, Daniel G. Goldstein et al.

Numerical perspectives help people understand extreme and unfamiliar numbers (e.g., \$330 billion is about \$1,000 per person in the United States). While research shows perspectives to be helpful, generating them at scale is challenging both because it is difficult to identify what makes some analogies more helpful than others, and because what is most helpful can vary based on the context in which a given number appears. Here we present and compare three policies for large-scale perspective generation: a rule-based approach, a crowdsourced system, and a model that uses Wikipedia data and semantic similarity (via BERT embeddings) to generate context-specific perspectives. We find that the combination of these three approaches dominates any single method, with different approaches excelling in different settings and users displaying heterogeneous preferences across approaches. We conclude by discussing our deployment of perspectives in a widely-used online word processor.

Jacy Reese Anthis, Hannah Cha, Solon Barocas et al.

The capabilities of artificial intelligence (AI) lie along a jagged frontier, where AI systems surprisingly fail on tasks that humans find easy and succeed on tasks that humans find hard. To investigate user reactions to this phenomenon, we developed an incentive-compatible experimental methodology based on diagram generation tasks, in which we induce errors in generative AI output and test effects on user reliance. We demonstrate the interface in a preregistered 3x2 experiment (N = 577) with error rates of 10%, 30%, or 50% on easier or harder diagram generation tasks. We confirmed that observing more errors reduces use, but we unexpectedly found that easy-task errors did not significantly reduce use more than hard-task errors, suggesting that people are not averse to jaggedness in this experimental setting. We encourage future work that varies task difficulty at the same time as other features of AI errors, such as whether the jagged error patterns are easily learned.

Kristina Gligoric, Tiziano Piccardi, Jake Hofman et al.

Science is facing a reproducibility crisis. Previous work has proposed incorporating data analysis replications into classrooms as a potential solution. However, despite the potential benefits, it is unclear whether this approach is feasible, and if so, what the involved stakeholders-students, educators, and scientists-should expect from it. Can students perform a data analysis replication over the course of a class? What are the costs and benefits for educators? And how can this solution help benchmark and improve the state of science? In the present study, we incorporated data analysis replications in the project component of the Applied Data Analysis course (CS-401) taught at EPFL (N=354 students). Here we report pre-registered findings based on surveys administered throughout the course. First, we demonstrate that students can replicate previously published scientific papers, most of them qualitatively and some exactly. We find discrepancies between what students expect of data analysis replications and what they experience by doing them along with changes in expectations about reproducibility, which together serve as evidence of attitude shifts to foster students' critical thinking. Second, we provide information for educators about how much overhead is needed to incorporate replications into the classroom and identify concerns that replications bring as compared to more traditional assignments. Third, we identify tangible benefits of the in-class data analysis replications for scientific communities, such as a collection of replication reports and insights about replication barriers in scientific work that should be avoided going forward. Overall, we demonstrate that incorporating replication tasks into a large data science class can increase the reproducibility of scientific work as a by-product of data science instruction, thus benefiting both science and students.

Serina Chang, Ashton Anderson, Jake M. Hofman

With the rapid adoption of LLM-based chatbots, there is a pressing need to evaluate what humans and LLMs can achieve together. However, standard benchmarks, such as MMLU, measure LLM capabilities in isolation (i.e., "AI-alone"). Here, we design and conduct a user study to convert MMLU questions into user-AI conversations, by seeding the user with the question and having them carry out a conversation with the LLM to answer their question. We release ChatBench, a new dataset with AI-alone, user-alone, and user-AI data for 396 questions and two LLMs, including 144K answers and 7,336 user-AI conversations. We find that AI-alone accuracy fails to predict user-AI accuracy, with significant differences across multiple subjects (math, physics, and moral reasoning), and we analyze the user-AI conversations to provide insight into how they diverge from AI-alone benchmarks. Finally, we show that fine-tuning a user simulator on a subset of ChatBench improves its ability to estimate user-AI accuracies, increasing correlation on held-out questions by more than 20 points, creating possibilities for scaling interactive evaluation.

Calvin Isley, Joshua Gilbert, Evangelos Kassos et al.

While large language models (LLMs) challenge conventional methods of teaching and learning, they present an exciting opportunity to improve efficiency and scale high-quality instruction. One promising application is the generation of customized exams, tailored to specific course content. There has been significant recent excitement on automatically generating questions using artificial intelligence, but also comparatively little work evaluating the psychometric quality of these items in real-world educational settings. Filling this gap is an important step toward understanding generative AI's role in effective test design. In this study, we introduce and evaluate an iterative refinement strategy for question generation, repeatedly producing, assessing, and improving questions through cycles of LLM-generated critique and revision. We evaluate the quality of these AI-generated questions in a large-scale field study involving 91 classes -- covering computer science, mathematics, chemistry, and more -- in dozens of colleges across the United States, comprising nearly 1700 students. Our analysis, based on item response theory (IRT), suggests that for students in our sample the AI-generated questions performed comparably to expert-created questions designed for standardized exams. Our results illustrate the power of AI to make high-quality assessments more readily available, benefiting both teachers and students.

Gabriel Kaptchuk, Daniel G. Goldstein, Eszter Hargittai et al.

A growing number of contact tracing apps are being developed to complement manual contact tracing. A key question is whether users will be willing to adopt these contact tracing apps. In this work, we survey over 4,500 Americans to evaluate (1) the effect of both accuracy and privacy concerns on reported willingness to install COVID19 contact tracing apps and (2) how different groups of users weight accuracy vs. privacy. Drawing on our findings from these first two research questions, we (3) quantitatively model how the amount of public health benefit (reduction in infection rate), amount of individual benefit (true-positive detection of exposures to COVID), and degree of privacy risk in a hypothetical contact tracing app may influence American's willingness to install. Our work takes a descriptive ethics approach toward offering implications for the development of policy and app designs related to COVID19.

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.

Amit Sharma, Jake M. Hofman, Duncan J. Watts

We present a method for estimating causal effects in time series data when fine-grained information about the outcome of interest is available. Specifically, we examine what we call the split-door setting, where the outcome variable can be split into two parts: one that is potentially affected by the cause being studied and another that is independent of it, with both parts sharing the same (unobserved) confounders. We show that under these conditions, the problem of identification reduces to that of testing for independence among observed variables, and present a method that uses this approach to automatically find subsets of the data that are causally identified. We demonstrate the method by estimating the causal impact of Amazon's recommender system on traffic to product pages, finding thousands of examples within the dataset that satisfy the split-door criterion. Unlike past studies based on natural experiments that were limited to a single product category, our method applies to a large and representative sample of products viewed on the site. In line with previous work, we find that the widely-used click-through rate (CTR) metric overestimates the causal impact of recommender systems; depending on the product category, we estimate that 50-80\% of the traffic attributed to recommender systems would have happened even without any recommendations. We conclude with guidelines for using the split-door criterion as well as a discussion of other contexts where the method can be applied.

Prem Gopalan, Jake M. Hofman, David M. Blei

We develop a Bayesian Poisson matrix factorization model for forming recommendations from sparse user behavior data. These data are large user/item matrices where each user has provided feedback on only a small subset of items, either explicitly (e.g., through star ratings) or implicitly (e.g., through views or purchases). In contrast to traditional matrix factorization approaches, Poisson factorization implicitly models each user's limited attention to consume items. Moreover, because of the mathematical form of the Poisson likelihood, the model needs only to explicitly consider the observed entries in the matrix, leading to both scalable computation and good predictive performance. We develop a variational inference algorithm for approximate posterior inference that scales up to massive data sets. This is an efficient algorithm that iterates over the observed entries and adjusts an approximate posterior over the user/item representations. We apply our method to large real-world user data containing users rating movies, users listening to songs, and users reading scientific papers. In all these settings, Bayesian Poisson factorization outperforms state-of-the-art matrix factorization methods.