Large Language Models For Social Science

Indira Sen

May 29, 2025 – 9am-1pm, Online via Zoom

The Institute for Social Research Population Dynamics and Health Program at the University of Michigan presents Response Quality Assessment in Surveys, a PDHP Workshop presented by

Please join as the PDHP workshop series resumes with “Large Language Models For Social Science” presented by Indira Sen of the University of Mannheim. This workshop is geared toward researchers and data analysts of all experience levels, particularly those interested in incorporating LLMs as part of their workflow for open-ended text data from surveys. This workshop will introduce LLMs as a social science tool, including an introduction to how LLMs work, and implementation and prompting of open-source LLMs such as Llama & Mistral. Attendees will also receive hands-on practice analyzing open-text survey data and simulating survey responses, primarily using Python. Finally, we will cover biases in LLMs, and discuss ethical considerations for using them in social research.

Topics covered:

  • Overview of LLMs and how they work
  • Introduction to running and prompting open-source LLMs (eg Llama, Mistral)
  • Applying LLMs to social research with Python, including analyzing open-text data and simulating survey responses
  • Biases in LLMs and ethical considerations for using them in social research

Video:

Workshop video is available here.

Slides & Lab Materials:

Slides are available for download here.
Python notebooks are available for download here.

Recommended Software:

— Huggingface login and API for serverless access
— OpenAI API Key to use commercial models
Please see the slides linked here for further details.

 

View Past PDHP Workshops