Coding agents in the social sciences

Worauf sich diese Seite beruft, wörtlich, abgerufen am 12.10.2026:

• can take a research idea and a dataset, write and run an analysis, interpret the output, and iterate autonomouslybestätigt 10.11.2026
• We present results from a survey of 1,260 social scientists about AI and coding agent use, fielded in February and March 2026.bestätigt 10.11.2026
• Only 20% of respondents use coding agents.bestätigt 10.11.2026
• Claude Code is the most common coding agent tool reported, with 86% of users reporting Claude Code usebestätigt 10.11.2026
• is for coding up analysis of quantitative databestätigt 10.11.2026
• respondents were recruited for a study that offered access to Claude Max accounts, so selection into the sample could tilt toward researchers curious about AI toolsbestätigt 10.11.2026
• We expect that the respondents are both heavier users and more optimistic about LLMs than non-responders.bestätigt 10.11.2026
• regularly (more than once a week) use an AI coding assistant integrated into your command linebestätigt 10.11.2026
• We launched the survey only a month after the release of Claude Coworkbestätigt 10.11.2026
• This means we do not count researchers who use AI agents exclusively via more general purpose desktop apps.bestätigt 10.11.2026
• Outcomes are self-reported by respondents.bestätigt 10.11.2026
• In percentage terms, coding agent users look around 10% (empirical projects started) to 75% (working papers posted) more productive than others in their discipline and career stage.bestätigt 10.11.2026
• We find no evidence that coding agent users are submitting more new papers to journals or resubmitting papers more quickly.bestätigt 10.11.2026
• Differences should not be interpreted as causalbestätigt 10.11.2026
• Finally, we only look here at the number of projects researchers report, and report nothing about their quality.bestätigt 10.11.2026
• The survey is the baseline wave of a larger ongoing study of how coding agents affect research productivity, including a randomized experiment providing researchers with access to Claude Code.bestätigt 10.11.2026
• We will publish results from this experiment in the future.bestätigt 10.11.2026
• the respondents were fairly similar to an earlier sample that received a more generic invitationbestätigt 10.11.2026
• self-reported output over the six months before the surveybestätigt 10.11.2026