Coding Agents in the Social Sciences
a day ago
- #Productivity
- #Social Science Research
- #AI Adoption
- A 2026 survey of 1,260 quantitative social scientists found 81% have tried AI chatbots for research, primarily for coding and editing prose, but only 20% have adopted autonomous coding agents like Claude Code.
- Coding agent adoption is uneven: it's twice as high among researchers with typically male names vs. female names, and 40% higher at top universities, with economists and political scientists leading adoption.
- Users of coding agents report starting more research projects, posting more working papers, and submitting more grant proposals compared to peers, but no increase in journal submissions yet.
- Researchers mainly use AI for generating analysis code and editing prose, not drafting text; only a third of AI users have employed it for drafting.
- While researchers are optimistic AI will boost paper-writing productivity, they are less confident about its overall positive impact on the social sciences, fearing issues like congestion and selective reporting.