Seeing, Hearing, and Reading Politics: A Framework for Using Multi-modal Large Language Models in Social Science Research

• In Event: LLMs and Generative AI

Brief Overview

Multi-modal large language models offer a unified framework for analyzing political ads across text, audio, and images. Zero-shot prompting outperforms few-shot, but all models require expert labels with DSL to ensure valid and unbiased inference.

Author

• Chandler Jackson L'Hommedieu, The Ohio State University, Main Campus