DebateLab@KIT computational philosophy projects

Studying Rational Agency and Epistemic Communities with Large Language Models

[🖥️ Slides] [📓 Jupyter Notebook]

In a keynote at the Workshop on Computational Models in Social Epistemology 2023, organised by the RUB Research Group on Reasoning, Rationality and Science as part of the DFG network “Simulations of Scientific Inquiry”, we’ve reviewed and critically reflected on the emergent field of LLM-based multi-agent simulations of (individual and collective) epistemic processes.

To get philosophers started in that field, we publish a Jupyter notebook that walks you through an implementation of the classic Bounded Confidence model with LLM-based agents. In that model, agents can process natural language and their opinions are represented by English sentences. The Jupyter notebook may serve as boilerplate for your own projects.