Jingyi Qiu
Ph.D. Candidate, School of Information, University of Michigan
On the 2026–27 academic and industry job markets.
jaqiu [at] umich [dot] edu
I am a Ph.D. candidate in Information Science at the University of Michigan School of Information, advised by Yan Chen, Qiaozhu Mei, and Alain Cohn. I study how to turn swarms of AI agents into effective organizations, and how to keep human and AI judges from rewarding hype over substance. I am currently a Frontier Agents research intern at Together AI, where I study how different ways of organizing agent swarms affect how effectively LLM agents work together, including how agents evaluate one another’s work and how they perform in strategic settings. Previously, I was a research intern at Microsoft Research (Economics and Computation).
My other work examines how social media promotion influences job market outcomes through a field experiment (PNAS 2026), and develops LLM-based measurement methods, such as a counterfactual framework for measuring rhetorical style (ICLR 2026). My research combines field and lab experiments, causal inference, and LLM-based measurement, so that claims about what AI systems do can be tested rather than assumed. I received a B.A. from Duke University and a B.Econ. from Duke Kunshan University.
Research Highlights
A preregistered field experiment in the economics job market shows that promoting candidates' job market papers on social media to the right audience improves their real labor market outcomes.
Covered by the Financial Times, Michigan News, and New Things Under the Sun.
Turns an LLM from an unreliable rater into a controlled measurement instrument. Across 8,485 ICLR submissions and 250,000+ counterfactual writings, promotional framing predicts citations and media attention beyond peer-review scores, and rises sharply after 2023 with LLM writing assistance.
In a hiring experiment for a real position, applicants suspected of using AI receive up to 40.6% lower evaluations whether or not they used it; recruiters detect AI use barely better than chance and infer lower ability, not lower effort.
How the organization of LLM agent collectives, model capability, reasoning, the agent harness, and the allocation of compute across agents shape multi-agent performance, including agents evaluating one another and agents in dynamic zero-sum games.
News
| Nov 2026 | Presenting at the 2026 INFORMS Annual Meeting (San Francisco, Nov 1–4), where I am also organizing a session, and at CIST 2026, the Conference on Information Systems and Technology (San Francisco, Oct 31–Nov 1). |
| Sep 2026 | Started as a Frontier Agents Research Intern at Together AI, studying how the organization of agent swarms shapes multi-agent performance. |
| May 2026 | Received the Best Poster Award at the AI for Social Science Research Methods Workshop at Yale University. |
| May 2026 | Our paper Social Media Promotion Improves Job Market Outcomes (with Yan Chen, Alain Cohn, and Alvin E. Roth) is published in PNAS. |
| Summer 2026 | Research intern at Microsoft Research, Economics and Computation group. |
| Jan 2026 | Our paper Counterfactual LLM-based Framework for Measuring Rhetorical Style (with Hong Chen and Zongyi Li) is accepted at ICLR 2026. |
| 2025 | Received support from the OpenAI Researcher Access Program and an Innovations Growth Lab Seed Grant (co-PI with Qingyi Wang). |
Honors & Awards
| 2026 | Best Poster Award, AI for Social Science Research Methods Workshop, Yale University |
| 2026 | Empowering Research with AI Award, University of Michigan |
Grants & Funding
| 2025 | OpenAI Researcher Access Program ($10,000) |
| 2025 | Innovation Growth Lab Seed Grant (co-PI with Qingyi Wang, $5,000) |
| 2022, 2024 | Rackham Graduate Student Research Grants, University of Michigan ($2,000; $3,000) |
In the Media
- Financial TimesWhen is it OK for economists to experiment on people?
- Michigan NewsEconomist retweets boost visibility, outcomes for doctoral job seekers
- New Things Under the SunTwitter and the Spread of Academic Knowledge
Selected Talks
| 2026 | Upcoming: INFORMS Annual Meeting (presenter and session organizer); Conference on Information Systems and Technology (CIST) |
| 2026 | Business Implications of Generative AI Conference (MIT) |
| 2025 | ML-Driven Discovery Workshop (MIT); Carnegie Mellon University (Nihar Shah group); Caltech Behavioral & Experimental Workshop |
| 2024 | Stanford Institute for Theoretical Economics (SITE); Econometric Society Interdisciplinary Frontiers Conference on Economics and AI+ML (Cornell); Caltech (Colin Camerer group) |
| 2022–2023 | ACM Conference on Economics and Computation (EC); Advances with Field Experiments (UChicago) |