AI Resilience Report forSocial Scientists and Related Workers, All Other

Social scientists study how people behave and how communities work, then turn that knowledge into solutions for real-world problems. Right now, most AI in this field is augmenting researchers rather than replacing them. Brookings analysts describe a major shift, noting that researchers have used AI coding agents to transform data analysis methods, build large data collections, and run analyses far faster than before, with one colleague saying AI agents are coming for the social sciences "like a freight train." A recent Stanford workshop on empirical work in the age of AI ^[1] showed PhD students using Claude Code to replicate classic economics papers from an empty folder, and professors running multiple "agent" assistants overnight to handle narrowly defined research tasks ^[2].

The APSA Presidential Task Force on AI, Politics, and Political Science ^[3] similarly highlights that generative AI is reshaping how political scientists collect data, code qualitative interviews, and even draft literature reviews. But Brookings warns ^[2] that working with a coding agent means spending much more time reviewing AI output, and technical skills may atrophy if AI is always at the tip of the spear. Judgment, ethics, and interpretation stay human.

Adoption is moving quickly because the tools are cheap, commercially available (Claude, ChatGPT, Codex), and dramatically cut research costs — the World Economic Forum's Future of Jobs Report 2025 ^[4] lists AI and data analytics as among the fastest-growing skills employers want. But adoption faces real brakes: Nature reports ^[5] that many scientists are worried about AI distorting peer review and producing plausible-looking but flawed research. Social scientists also handle sensitive interviews and community data, so privacy, consent, and bias concerns slow rollout.

The good news for young people: human skills like ethical judgment, cultural understanding, listening, and asking the right questions are exactly what these tools can't do — and they're becoming more valuable, not less.