Anthropic coding agents gender gap social science AI adoption

Men Use AI Coding Tools Twice as Much as Women in Social Science, Anthropic Finds

2 June 2026

An Anthropic study of social scientists found that men use AI coding agents—tools that automatically write code—more than twice as often as women, with the gap persisting across disciplines and career levels. Economists and early-career researchers at top universities are the heaviest adopters, with code generation for data analysis being the dominant use case at 97%. While 88% of respondents are optimistic about AI boosting their own productivity, 70% are more skeptical about its broader impact on their field, citing concerns about peer review overload and research quality.

A new Anthropic study into how social scientists use AI has turned up a striking gender gap, specifically around coding agents like Claude Code. Researchers with typically male names use these tools more than twice as often as those with typically female names. The disparity holds across disciplines and career stages, which rules out the easy explanation that it's just a field-composition effect.

General AI adoption is fairly balanced across groups. The coding agent gap is where things get lopsided. Economists are the heaviest users at 39 percent adoption, while education researchers barely register at four percent. PhD students and postdocs are notably more likely to use coding AI than professors, and researchers at top-25 universities use these tools around 40 percent more than colleagues at lower-ranked institutions.

Unsurprisingly, code generation for data analysis dominates at 97 percent of coding agent users. Text drafting comes in at a distant third of respondents overall. Economists are the outliers here too, with half of them also using AI for writing, making them the most versatile adopters in the sample.

The authors flag that the divides by gender, career level, and institutional prestige are all consistently wider for coding agents than for general AI use. Whether that reflects differences in access, confidence, workflow habits, or something else, the study doesn't fully resolve.

On the productivity question, researchers are bullish about their own output. Eighty-eight percent rate AI's effect on their personal productivity above the midpoint of a ten-point scale, and half put it at eight or higher. Coding agent users are the most optimistic of all.

Here's where it gets interesting, though. Around 70 percent of respondents rate AI's impact on their own work more favourably than its impact on social science as a discipline. The authors suspect this comes down to systemic concerns: more papers flooding an already creaking peer review system, stiffer competition for attention, and existing problems like selective reporting or incremental, risk-averse research getting worse rather than better.

That anxiety isn't paranoia. In biomedical research, AI-hallucinated citations are already appearing in papers that feed into clinical guidelines, with fabrication rates reportedly jumping more than twelvefold since 2023. Social scientists watching that unfold have reasonable grounds for unease about what productivity gains might actually cost the field.