Replication Data for: Generative Multimodal Models for Social Science: An Application with Satellite and Streetscape Imagery

This repository provides sample code and resources for implementing Law and Roberto's (2025) Social Science Framework for Image Analysis with Generative Multimodal Models. This framework, which extends Grimmer, Roberts, and Stewart's (2022) "agnostic approach" to computational text analysis, consists of three core tasks: (1) curation, (2) discovery, and (3) measurement and inference. The sample code and resources provided here demonstrate how to implement this framework with an empirical application that uses OpenAI's GPT-4o multimodal model to analyze satellite and streetscape images to identify built environment features that contribute to contemporary residential segregation in U.S. cities. Article: Law, Tina and Elizabeth Roberto. Forthcoming. "Generative Multimodal Models for Social Science: An Application with Satellite and Streetscape Imagery." Sociological Methods & Research. Preprint: https://osf.io/preprints/socarxiv/6jq32 Repository: https://github.com/tinalaw1/genmm-for-social-sci