AI & Machine Learning

Organizations delivering services through digital technology have the opportunity to use machine learning and artificial intelligence to improve their products and services.

When training and education are provided through digital technology, it is possible to customize the individual’s experience based on their characteristics as well as their prior interaction with the application. Machine learning methods can be applied to estimate what works for whom, and why. The lab specializes in adapting machine learning methods to focus on heterogeneity in the effect of interventions. Planning experiments with the goal of learning about heterogeneity is another area of interest.

Another area of focus is developing algorithms that are used to improve personalization for digital services. Innovations in machine learning in the last decade have led to improvements in algorithms for making recommendations to individuals about content, and these can be applied to, for example, recommend stories to students in an educational application. Artificial intelligence can be used to “get to know” an individual and adapt the information presented based on previous experience; the lab makes use of methods such as multi-armed bandit algorithms and reinforcement learning to accomplish this.

Many commonly used methods in artificial intelligence are designed for large systems with many users, in environments where mistakes have few consequences. The lab works to customize and improve these algorithms for the context in which we apply them, where the user base may be smaller and mistakes more costly, such as educational applications.

Project Abstracts

Read about a few of the research projects the lab is currently working on.

Machine Learning-Based Causal Inference Tutorial

The tutorial introduces key concepts in machine learning-based causal inference and is an ongoing project with new chapters uploaded as they are completed. Topics currently covered:

  • Introduction to Machine Learning
  • Average Treatment Effects
  • Heterogeneous Treatment Effects
  • Policy Evaluation
  • Policy Learning

Do Commonly Prescribed Drugs Prevent COVID-19 Symptoms?

A variety of commonly prescribed drugs have the potential to prevent or treat COVID-19 symptoms. Using historical health claims data and methods combining machine learning and causal inference, this project identifies if these commonly prescribed drugs improved the outcome of patients hospitalized during the COVID-19 crisis. The analysis compares the patient outcomes of those who were already taking these drugs to those who were not at the time of hospitalization and uses a technique that pools information from different hospital systems from around the world to enable a larger sample to be considered together.


Evaluating the Benefits of Commonly Prescribed Drugs to Prevent Poor Outcomes from Respiratory Distress

Applying recently developed machine learning and causal inference methods to historical health claims data allows the evaluation of the impact of certain medications on outcomes for patients hospitalized with respiratory conditions. The methods incorporate new techniques to estimate causal effects across distinct, proprietary data sources without merging the datasets. The results can be used to suggest candidates for clinical trials in the fight against COVID-19.


Improving In-App Recommendations and Assessments

Lab researchers are collaborating with Stones2Milestones to design a system that recommends new in-app content for users in order to improve the students’ interest and engagement in the educational content. In addition, the lab is identifying methods to improve how users are assessed both on the learning occurring within the app and across a user’s educational journey.


Boosting Efficacy, Efficiency, and Accountability in Government Training Programs

Lab researchers are building a tool to help participants in government-funded training programs to select the programs likely to be most effective, taking into account an individual’s circumstances and qualifications. With a broad set of collaborators, the lab is using administrative data, science, and technology, and the outcomes of previous program participants to create the tool. This project will also incentivize training programs to add value to the earning capacity of program attendees.

Government Services

Examining the Social Value of Targeting Interventions in Retraining Programs

This project uses administrative data, machine learning, and causal inference methods to evaluate the effectiveness of job retraining programs in Rhode Island for different types of individuals. Understanding which segments of the population do not respond well to existing programs in turn informs the design of job retraining programs as well as which programs to recommend to different individuals.

Worker Training

Academic Publications

Publication Search

Short Course

This course is a series of videos designed for any audience looking to learn more about how machine learning can be used to measure the effects of interventions, understand the heterogeneous impact of interventions, and design targeted treatment assignment policies.

Thought Leadership

Venture Beat

Helping workers find job or skill training will be an important part of addressing the instability and job loss brought on by AI, said Stanford [Professor] Susan Athey, but the United States hasn’t historically done a great job of helping displaced workers impacted by automation.


The field of computational social science has exploded in prominence over the past decade but the field has also fallen short in important ways. We suggest opportunities to address these issues, especially in improving the alignment between the organization of the 20th-century university and the intellectual requirements of the field.

Stanford Institute for Economic Policy Research

As attention on race relations increases across the nation, Stanford economists Susan Athey and Matthew Gentzkow have used a new way of measuring segregation that can provide clues to different ways for combatting its negative effects.


What might've seemed absurd just a few months ago is what we need to be trying right now: building useless factories for a good reason. It is economically beneficial to invest in capacity early, at risk.

Women in Data Science at Stanford

Susan Athey, Economics of Technology Professor at the Stanford Graduate School of Business, brings an economist’s expertise and perspective to machine learning and data science. With a prolific career spanning academia and industry, Susan’s research focuses on the economics of digitization, marketplace design, and the intersection of econometrics and machine learning.

Studieförbundet Näringsliv och Samhälle

How does AI work? How will it impact business, help government become more efficient and be beneficial for social impact?

Toulouse Network for Information Technology

Stanford’s Susan Athey discusses the extraordinary power of machine-learning and AI techniques, allied with economists’ know-how, to answer real-world business and policy problems. With a host of new policy areas to study and an exciting new toolkit, social science research is on the cusp of a golden age. Economics, in particular, will never be the same again.

Voices of Stanford GSB

Economics is an area that allows you to approach important issues that have a lot of impact on people. Susan Athey’s pioneering work as a “tech economist” has helped industry and academia alike better understand the constantly shifting digital era.


Machine-learning prediction methods have been extremely productive in applications ranging from medicine to allocating fire and health inspectors in cities. However, there are a number of gaps between making a prediction and making a decision, and underlying assumptions need to be understood in order to optimize data-driven decision-making.


Menbere Shiferaw, PhD Student at the NYU Wager Graduate School of Public Service, and Dr. Susan Athey, Professor of Economics and Technology at the Stanford Graduate School of Business, investigate machine learning applications in the context of policy analysis.


"The short answer is that I think it will have an enormous impact; in the early days, as used “off the shelf,” but in the longer run econometricians will modify the methods and tailor them so that they meet the needs of social scientists primarily interested in conducting inference about causal effects and estimating the impact of counterfactual policies (that is, things that haven’t been tried yet, or what would have happened if a different policy had been used)."