The hackathon will take place at the University of Cambridge, on May 11th 2026 between 10:00 and 17:00. All participants will be co-authors on the resulting Meta Paper (details here: https://osf.io/dkfzt/files/eb7nh). The tentative title and abstract are reproduced below.
In this hackathon, participants will work individually. The event is aimed at undergraduates as well as graduate students, faculty and staff. The main disciplines of focus are political science, psychology and economics. Participants’ contribution is done the day of the event only. Please share this with any colleagues and students (including 2nd and 3rd year undergraduates) who you think may be interested.
If you are interested in participating, please register below
Registration
While the organisers have a preference for in-person participation, there is also the option of hybrid participation.
You can read the outputs of some of their previous hackathons here: Reproducibility and robustness of economics and political science research and Comparing Human-Only, AI-Assisted, and AI-Led Teams on Assessing Research Reproducibility in Quantitative Social Science.
Bridging Research Gaps With AI: Expertise and Disciplinary Mobility
We test whether providing large-language-model (LLM) assistance compresses performance gaps in research skills along two prespecified dimensions. Vertically, we study expertise tiers using a three-category grouping (undergraduate, graduate student, professor/researcher) that governs both randomization and analysis while retaining self-reported titles for descriptive context. Horizontally, we study cross-discipline performance in three quantitative social sciences (Economics, Political Science, Psychology) to assess whether AI reduces the penalty from working outside one’s primary discipline (out-of-discipline, OOD). Participants are randomized 1:1 to AI access (e.g. ChatGPT) versus human-only controls within the applicable expertise strata; tasks are assigned a single discipline tag, with undergraduates always receiving inside-discipline assignments and approximately the remaining participants randomly assigned to an outside-discipline paper. Outcome measures are organized into two families: coding skills (computational reproducibility and the detection of major/minor coding errors) and non-coding skills (referee appropriateness/overall scores and robustness-check execution). The study is designed to reveal whether AI acts as an equalizer within expertise tiers and whether it enables researchers to operate effectively across disciplines.
Posted on 23/04/2026
