Alumni Speaker - Prospective Students' Day

Reflections on Using Large Language Models in Social Science Research

Large language models (LLMs), or generative AI, are rapidly transforming the landscape of social science research. In this talk, I reflect on my recent experiences applying LLMs to the coding of complex sociological concepts. While these models have significantly reduced barriers to computational analysis—primarily through the practice of prompt engineering—they also introduce important epistemic risks concerning validity, reliability, replicability, and transparency. Using social movement studies as an illustrative example, I discuss these challenges in detail and propose a structured workflow that social scientists can adopt when integrating LLMs into their research processes. Ultimately, I argue that although the future of social science coding is likely to be increasingly automated, it will concurrently require greater human oversight to ensure robust and credible research outcomes.