The article presents a new opportunity in creating computational models of social systems using generative artificial intelligence. This is done through Generative Agent-Based Models (GABMs) that use large language models like ChatGPT to represent human decision-making in social scenarios. The article provides a GABM case that incorporates human behavior in simulation models by combining a mechanistic model of human interactions with a pre-trained large language model. The authors examine various scenarios and the sensitivity of the results to changes in the prompt. The aim is to guide the building of diffusion models that include realistic human reasoning and decision-making.
Publication date: Not provided in the text
Project Page: ?
Paper: https://arxiv.org/pdf/2309.11456
