Applying open-source LLMs in Social and Behavioral Sciences

About

Open-access large language models (LLMs) like Llama-3 or Phi-3 are powerful alternatives to proprietary LLMs like ChatGPT that offer state-of-the-art performance and crucial advantages in terms of reproducibility and scale. This course introduces the use of open-access LLMs from the Hugging Face ecosystem for research in the behavioral and social sciences. In short lectures, participants will learn about key concepts (e.g., embeddings, causal attention, feature extraction, classification, prediction, fine-tuning, and token generation) and practical examples from social and behavioral science. In hands-on exercises, participants will apply language models to answer research questions from psychology, political science, decision-making research, and other fields. During and after the course, participants will engage in a personal research project applying large language models to a personal research question. Two lecturers will hold this course, implying a high level of support during the exercises and research project design

Course leader

Dirk Wulff from Max-Planck-Institut Berlin, Germany; Zakir Hussain from University of Basel, Switzerland

Target group

The GSERM Global School in Empirical Research Methods is a high-calibre programme dedicated to teaching cutting-edge research methodology. We not only welcome PhD students, post-docs, and graduate students from all kinds of fields, but also professionals working outside of academia.

Prerequisites
Basic knowledge of statistics, data analysis, and programming

Course aim

Large language models like ChatGPT or Llama-3 are incredibly useful tools for research in the social and behavioral sciences. In this course, you will (1) learn about the fundamental principles of large language models, (2) learn how to employ state-of-the-art open-access large language models using the Hugging Face ecosystem, (3) learn about the rich opportunities that large language models offer for behavioral and social science research, and (4) gain experience in applying large language models to answer personal research questions

Fee info