Anthropic has studied how social scientists use AI and found that far more men than women use coding agents, AI tools like Claude Code that write program code automatically.

Researchers with typically male names use these tools more than twice as often as those with typically female names. The gap holds even within the same disciplines and career levels.

Economists lead in coding agent adoption at 39 percent, while education researchers sit at the bottom with just four percent. PhD students and postdocs use coding AI far more than professors, and researchers at top-25 universities use the tools 40 percent more often than their peers. The dominant use case is code generation for data analysis, at 97 percent. Only a third use AI for writing text.

The authors note that gaps by gender, career level, and university rank are all wider for coding agents than for general AI use.

Researchers are bullish on their own productivity but skeptical about their field

Respondents are optimistic about AI's effect on their own paper output: 88 percent rate it above 5 on a 10-point scale, and half rate it at 8 or higher. Coding agent users are even more optimistic than others.

But 70 percent of respondents are more upbeat about their own productivity than about AI's impact on the social sciences as a whole. The authors suspect researchers worry that more papers could overload the peer review system, intensify competition for attention, and worsen existing problems like selective reporting and risk-averse, incremental research.

That concern tracks with what's already happening in other fields. In biomedical research, AI-hallucinated citations are creeping into papers that shape clinical guidelines, with fabrication rates jumping more than twelvefold since 2023.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

Subscribe now