Reading Anthropic's Survey of 1,260 Social Scientists on Coding Agents
Anthropic surveyed 1,260 quantitative social scientists (Feb–Mar 2026) on coding agents: tools that, unlike chatbots, run analyses themselves, interpret the results, and iterate. The report looks at how those tools change the way research gets done. Here are five numbers from it that stood out.
Insight 1: 81% have tried AI, 20% use coding agents regularly
Only 20% of respondents use coding agents regularly (more than weekly), even though 81% have tried AI chatbots for research. Among coding agent users, 86% use Claude Code.
The 81% counts anyone who has tried AI chatbots for research; the 20% counts weekly-or-more use of coding agents. The two numbers measure different tools.
Insight 2: The adoption gap for coding agents is wider than for AI in general
Researchers with typically male names use coding agents at more than twice the rate of those with female names, and top-25 university researchers are 40% more likely to adopt. Doctoral students and postdocs use them weekly at just over 25%; tenured professors at less than half that rate.
Use splits sharply by gender, institution and career stage. On career stage the direction is reversed: doctoral students and postdocs use coding agents weekly at more than twice the rate of tenured professors.
Insight 3: Coding agents are mostly used for analysis code
97% of coding agent users (versus 77% of other AI users) use them to generate analysis code. Editing prose comes second; drafting prose is a minority practice outside economics and management.
Compared with other AI users (77%), coding agent users lean further toward code.
Insight 4: More projects and working papers, no more journal submissions
Coding agent users start about 25% more projects and post about 50% more working papers. But there is no evidence that coding agent users submit more new papers to journals. (The comparison is descriptive; selection effects can't be ruled out.)
The report itself notes the comparison is descriptive: researchers who were already more productive may simply have adopted coding agents first (a selection effect).
Insight 5: More optimistic about their own productivity than about the field
88% are optimistic about AI's contribution to paper productivity (above the midpoint on a 1-10 scale), but 70% are more optimistic about their own paper productivity than about the field's overall impact.
The 88% are optimistic that AI will help paper productivity; the 70% rate their own paper productivity more optimistically than AI's impact on the field as a whole.
