AI-Driven Research in Econ-CS
Workshop at EC 2026 ᐧ Rome, Italy ᐧ July 6, 2026
Organized by Google Research
Workshop at EC 2026 ᐧ Rome, Italy ᐧ July 6, 2026
Organized by Google Research
The rapid evolution of modern AI systems—such as Gemini, Claude, and ChatGPT—has fundamentally altered the landscape of academic research. Capabilities that recently belonged to the realm of science fiction, including automated literature reviews, assisted theorem proving, and complex code generation, are now seamlessly integrated into the daily routines of researchers.
The goal of this workshop is to provide a dedicated space for the Economics and Computation (EconCS) community to exchange emerging best practices, share early experiences with AI-assisted workflows, and highlight recent breakthroughs.
The program will feature a small number of invited presentations. In addition, we solicit submissions across two distinct tracks:
AI-driven EconCS research: results of interest to the EconCS community that were solved using AI, defined in very broad terms.
EconCS Problems to be solved with AI: unsolved problems that the submitter believes are ripe for AI, but that are still open.
Both tracks focus on EconCS defined in broad terms—encompassing any topic you might find at EC, WINE, SAGT, WWW, and the EconCS sessions of STOC, FOCS, SODA, AAAI, NeurIPS, and similar venues.
The workshop will be preceded by a tutorial on AI-Driven Research in EconCS, given by Paul Duetting and Renato Paes Leme.
The workshop will be preceded by a morning tutorial on AI-Driven Research in EconCS.
[2:00-3:30pm] First Session
2:00: welcome and introduction
Renato Paes Leme (Google Research)
Contributed talks (chair: Paul Duetting)
2:10: Solving Open Problems in Operations Research Using AI
Eric Fithian, Rad Niazadeh, Pranav Nuti (University of Chicago)
2:20: EconCSLib: A Lean Library for Computational Economics and AI-Assisted Research
Xiaohui Bei (Nanyang Technological University) , Jiajun Ma (Xiamen University),
Zhan Jing (Shanghai Jiao Tong University), Hongfei Fu (Shanghai Jiao Tong University),
Zhihao Gavin Tang (Shanghai University of Finance and Economics)
2:30: EconCSLib: an AI-generated Lean library for Economics & Computation Research
Nikhil Garg (Cornell Tech)
5 min break
2:45 Stable Menus of Public Goods: AI-Enabled Progress
Sara Fish (Harvard University)
2:55 A New Lower Bound for the Random Offerer Mechanism in Bilateral Trade using AI-Guided Evolutionary Search
Yang Cai (Yale University), Vineet Gupta (Google Research), Zun Li (Google Deepmind), Aranyak Mehta (Google Research)
3:05 Adrasteia: A Human-in-the-Loop Agentic Workflow for Theory-Oriented Research in Economics
Heikichi Hayashi (Adrasteia Labs), Ruoran Lai (Sun Yat-sen University),
Shawn Yu (Adrasteia Labs; Boston College), Huanxi Zhang (University of Wisconsin-Madison),
Jiazhuo Li (University of Michigan
3:15 AI For Science
Song Zuo (Google Research)
[3:30-4:00] Coffee break
[4:00-5:30] Second Session
4:00 Keynote Talk. Title: "On the Disproof of the Unit Distance Conjecture"
Mark Sellke (Harvard and OpenAI)
4:30 Open Problem Session
Sebastien Lahaie (Google Research)
4:45 Panel How will AI tools change EconCS research and education?
Panelists: Matt Weinberg (Princeton), Yannai Gonczarowski (Harvard), Yiling Chen (Harvard)
Moderator: Simina Branzei (Purdue and Google)
As part of the workshop, we collected a set of open problems from leaders of the field. Those problems will serve as a testbeds to track future advancements in AI for tackling EconCS problems. The problems are available at https://github.com/aieconcs/econcs-bench.
Submission Deadline: May 31, 2026
Notification of Acceptance: June 5, 2026
Workshop Date: July 6, 2026
All submissions should be made through the workshop’s HotCRP submission site. Further details regarding the two tracks and specific submission requirements are provided below.
Submissions to this track consist of EconCS results derived either with AI assistance or in a fully automated way by AI. This encompasses everything from human-AI collaboration to entirely machine-generated results. We are specifically looking for contributions that explore how AI or agentic workflows are utilized directly for scientific discovery (e.g., AI systems generating novel scientific hypotheses, automating the research process, or acting as autonomous researchers).
Submissions will be selected based both on the significance of the result itself as well as the novelty of the AI workflow used to derive it. Submissions are expected to include a section describing the problem, an outline of the results and techniques, as well as a description of the workflow used to interact with AI. We recommend (but not require) that the authors share prompts and prompt best practices together with the submission. Submissions of any length are allowed, but the complete statement of the problem and results, and the description of the AI workflow must be contained in the first 5 pages. While results may not be derived by a human, it is expected that the human authors have checked and vouch for the results.
Authors of accepted submissions will give a 10 minute talk. The presenters are encouraged to focus the presentation on their AI workflow used to obtain the solution and lessons learned.
The workshop is non-archival. Authors of accepted papers are expected to post their papers online (arXiv, personal website, or equivalent). All accepted papers will be linked via a public index as a community resource. Papers that were accepted or rejected from EC are eligible. Dual submissions to other conferences or journals are permitted.
The goal of this track is to encourage the discussion of what are good problems to test AI systems on, as well as initiating a community-led effort for building a repository that collects and tracks progress on problems of interest to the EconCS community.
The submission must contain a text file (.txt or .md) with the prompt, along with a short (up to) 1-page accompanying PDF or TXT file explaining why this is a good open problem. Example prompts include:
We especially welcome problems of two types:
(1) Moonshots – problems which are recognized by the community, and where progress would imply a significant breakthrough.
(2) Technical logjams – problems where progress has been blocked by the very nature of the problem and human capabilities and taste, such as closing the gaps between upper and lower bounds, searching for stronger impossibility or hardness constructions, etc.
Accepted open problems will be included (with attribution) in a GitHub repository that will be available for the community as a testbed for AI models progress in EconCS tasks. We plan to use the workshop as a forum to discuss how this repository should be managed, and how we plan to keep track of the progress that is being made. A subset of open problems will be selected for a 5-min presentation at the workshop.
Given the experimental nature of the workshop, we will accept submissions that deviate from those guidelines (e.g., longer files specifying families of problems) as long as they fit the spirit of the workshop.
Organizers
Simina Branzei (Purdue and Google)
Yuan Deng (Google)
Paul Duetting (Google)
Sebastien Lahaie (Google)
Vahab Mirrokni (Google)
Renato Paes Leme (Google)
Song Zuo (Google)
Contact: aieconcs26@googlegroups.com