About

John J. Horton is an Associate Professor at the MIT Sloan School of Management and a member of the MIT Initiative on the Digital Economy research team. His research focuses on the intersection of digital technology, economics, and management, contributing to understanding how AI agents for commerce could shape future markets. His work involves exploring the implications of artificial intelligence, quantum computing, and digital culture on economic behavior and market dynamics, aiming to provide insights into the future of work, productivity, and market structures in the digital age.

Research topics

• Business
• Economics
• Psychology
• Social psychology
• Demographic economics
• Political Science
• Mathematics
• Microeconomics
• Sociology
• Geography
• Demography
• Industrial organization
• Labour economics
• Engineering
• Medicine
• Statistics
• Marketing

Selected publications

• General Social Agents

  SSRN Electronic Journal · 2026-01-01

  preprintOpen accessSenior author
  PublisherDOI

• Consumer Demand with Social Influences: Evidence from an E-Commerce Platform

  Management Science · 2025-08-25

  articleOpen access

  For some types of goods, rarity itself is valued: items considered “fashionable” are demanded in part because they are unique. In this paper, we explore the economics of rare goods using auctions of limited-edition shoes held by an e-commerce platform. We model endogenous entry and bidding in multiunit auctions and construct demand curves from realized bids. We find that doubling inventory reduces willingness to pay by 8%–15%. From the monopolist’s perspective, ignoring the value of rarity leads to substantial overproduction: Auctioned quantities are 88% above the profit-maximizing level. From the consumers’ perspective, however, the negative spillovers from quantity restriction more than offset the benefits of rarer goods. This paper was accepted by Omar Besbes, revenue management and market analytics. Supplemental Material: The online appendices and data files are available at https://doi.org/10.1287/mnsc.2024.04995 .

  PublisherOA PDFDOI

• Competition Avoidance vs. Herding in Job Search: Evidence from Large-Scale Field Experiments on an Online Job Board

  Management Science · 2025-06-04

  articleSenior author

  Information about the number of applicants to a job vacancy might simultaneously signal the degree of competition and vacancy quality. We study how this information affects job search. To do so, we conduct three experiments on a large online job platform in which the treatment varies what information is shown to job seekers. Information about the number of prior applicants to a vacancy increases the number of applications and redirects them to vacancies with few prior applications. Information about vacancy age increases application rates, especially to new vacancies. To further investigate the causal mechanisms, we conduct and analyze a survey choice experiment. We conclude that job seekers prefer to avoid competition rather than use the popularity of a vacancy as a signal of quality. This paper was accepted by Karan Girotra, operations management. Funding: This research was facilitated through a research consulting agreement between A. Fradkin, J. J. Horton, and Meta. A. Fradkin and J. J Horton were employed as contractors as part of the agreement. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2023.02483 .

  PublisherDOI

• Simulating the Survey of Professional Forecasters

  SSRN Electronic Journal · 2025-01-01 · 11 citations

  preprintOpen access
  PublisherDOI

• The Death of a Technical Skill

  Information Systems Research · 2025-03-19 · 3 citations

  article1st authorCorresponding

  For managers, we show that opportunities for skill development strongly influence matching in online information technology (IT) markets. Employers cannot easily circumvent labor scarcity by adopting older technologies, as workers avoid projects with declining future skill value absent substantial wage premiums. However, older workers, having shorter career horizons, are less sensitive to declining skill value, suggesting potential benefits in matching them with legacy technologies. For policy makers, our research demonstrates that labor market tightness persists across both new and old technologies in online IT markets. This challenges the notion that employers can engage in labor arbitrage by avoiding cutting-edge technologies. Policy frameworks therefore need flexibility to address skill shortages wherever they emerge. Additionally, our findings highlight the need for more granular data collection on technical skill evolution beyond broad occupational categories. Our online context provides unique insights into how corporate decisions about technology standards cascade into labor markets. The findings underscore the importance of policies promoting continuous learning and adaptability, while suggesting that age-diverse hiring practices could help address both skill shortages and age discrimination concerns in technical fields.

  PublisherDOI

• General Social Agents

  SSRN Electronic Journal · 2025-01-01 · 1 citations

  preprintOpen accessSenior author
  PublisherDOI

• Generative AI and Labor Market Matching Efficiency

  SSRN Electronic Journal · 2025-01-01 · 2 citations

  preprintOpen accessSenior author
  PublisherDOI

• Algorithmic Writing Assistance on Jobseekers’ Resumes Increases Hires

  Management Science · 2025-04-18 · 9 citations

  articleSenior author

  There is a strong association between writing quality in resumes for new labor market entrants and whether they are ultimately hired. We show this relationship is, at least partially, causal: In a field experiment in an online labor market with nearly half a million jobseekers, treated jobseekers received nongenerative algorithmic writing assistance on their resumes. Treated jobseekers were hired 8% more often at 10% higher wages. Contrary to concerns that the assistance takes away a valuable signal, we find no evidence that employers were less satisfied. We find that the writing on treated jobseekers resumes had fewer errors and was easier to read. Our analysis suggests that writing is an imperfect signal of ability but better writing helps employers ascertain ability through clearer writing, suggesting digital platforms could benefit from incorporating nongenerative algorithmic writing assistance into text-based descriptions of labor services or products. This paper was accepted by Anindya Ghose, information systems. Funding: J. Horton and E. Wiles received funding from the online labor market on which this experiment was run. No authors received funding from the Algorithmic Writing Service. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2024.04528 .

  PublisherDOI

• Automated Social Science: Language Models as Scientist and Subjects

  arXiv (Cornell University) · 2024-04-17 · 15 citations

  preprintOpen accessSenior author

  We present an approach for automatically generating and testing, in silico, social scientific hypotheses. This automation is made possible by recent advances in large language models (LLM), but the key feature of the approach is the use of structural causal models. Structural causal models provide a language to state hypotheses, a blueprint for constructing LLM-based agents, an experimental design, and a plan for data analysis. The fitted structural causal model becomes an object available for prediction or the planning of follow-on experiments. We demonstrate the approach with several scenarios: a negotiation, a bail hearing, a job interview, and an auction. In each case, causal relationships are both proposed and tested by the system, finding evidence for some and not others. We provide evidence that the insights from these simulations of social interactions are not available to the LLM purely through direct elicitation. When given its proposed structural causal model for each scenario, the LLM is good at predicting the signs of estimated effects, but it cannot reliably predict the magnitudes of those estimates. In the auction experiment, the in silico simulation results closely match the predictions of auction theory, but elicited predictions of the clearing prices from the LLM are inaccurate. However, the LLM's predictions are dramatically improved if the model can condition on the fitted structural causal model. In short, the LLM knows more than it can (immediately) tell.

  PublisherOA PDFDOI

• The Impact of Generative AI on Labor Market Matching

  2024-03-27 · 3 citations

  articleOpen accessSenior author

  Imagine applying for jobs by simply asking your artificial intelligence (AI) assistant to “please put together a resume and cover letter based on my experiences and submit the application to senior management positions at clean-energy start-ups with fewer . . .

  PublisherOA PDFDOI

Frequent coauthors

• Ramesh Johari

  31 shared

• Philipp Kircher

  29 shared

• Joseph M. Golden

  17 shared

• Richard Zeckhauser

  Harvard University Press

  16 shared

• Moshe Barach

  University of Minnesota

  16 shared

• Apostolos Filippas

  Fordham University

  16 shared

• Daniel L. Chen

  9 shared

• Aaron Shaw

  Northwestern University

  7 shared

Labs

• John J. Horton LabPI

Awards & honors

• INFORMS ISSA Sandra A. Slaughter Early Career Award (2021)
• AI/ML Rising Star Award at the 2021 Conference on Artificial…