This advanced course explores state-of-the-art AI methods for social science research, focusing on LLM-based
approaches including synthetic participants, AI-augmented surveys, and computational simulations of human
behavior. Students will critically evaluate recent work through
paper discussions and an original research project. Projects will involve empirical work with publication
potential, such as replication studies comparing synthetic to human data, meta-analyses of field practices, or
development of validation frameworks for the emerging field of AI-augmented social science.
Prerequisites: Having taken a graduate-level computer science course in Artificial Intelligence or
Machine Learning (including CSS, NLP, Computer Vision, etc.) is expected. Students must be comfortable with
reading recent research papers and discussing key concepts and ideas.
Course Goals: Specific outcomes for this course are that students will be able to:
- Determine appropriate use cases for AI-augmented methods by comparing their strengths and weaknesses against
traditional social science approaches (surveys, experiments, qualitative interviews) for specific research
questions.
- Design and execute rigorous studies that assess whether AI-generated data (synthetic survey responses,
simulated behaviors, agent interactions) accurately represent human populations and social phenomena.
- Critically evaluate methodological claims in published research, identifying validity threats, confounds,
and limitations in studies using LLMs for social science applications.
- Implement reproducible research practices including pre-registration, comprehensive reporting of prompts and
model parameters, sensitivity analyses, and proper documentation of AI system configurations.
- Conduct original empirical research that contributes to methodological knowledge in AI-augmented social
science, whether through replication studies, validation frameworks, meta-analyses, or novel applications.
Thank you to Daniel Khashabi for sharing the course website template!
