Welcome to AI for Social Research

Social science research moves slowly for good reasons. Qualitative coding is painstaking. Interview transcription takes hours. Literature reviews can eat months. Survey analysis depends on careful, sometimes manual, classification. LLMs change a lot of that, in ways that are still being figured out. Researchers using AI well are doing studies at scales that were not possible before. Researchers using AI badly are publishing papers with fabricated quotes and misclassified themes. The methodology is still being written. This course gives you the current best practice. We focus on social science specifically: sociology, anthropology, public health, education research, organizational behavior, political science. Not data science. Not pure NLP. We'll cover: • A PhD candidate on how LLMs are changing research work (an honest intro) • Why this matters: scale and speed in qualitative research (LSE) • Real example: thematic analysis with LLMs in a public health study • A trustworthy workflow for using LLMs in qualitative analysis • How aligned is LLM coding with human coding? The evidence • AI as the interviewer: what changes when respondents talk to AI • AI-assisted interviewing: hybrid approaches that work • The limits: where AI cannot replace the ethnographer • AI in survey research: Pew's analysis • Ethics: consent, transparency, IRB approval This course is for academics, applied researchers, UX researchers, and anyone doing qualitative work. Methodological background helps. Time: ~3 hours across 10 articles.