Petter Törnberg - Simulating Social Media with Large Language Models

When: 2024-05-06 14:30-16:30 (Tallinn time)
Where: online

The event is public via zoom: https://zoom.us/j/94927755198?pwd=VWpWU25DY1JlVFJsbDF4aUtyL3Zvdz09 Websites: https://www.pettertornberg.com/

Abstract – Large Language Models have shown capable of convincingly mimicking human conversation. This presentation discusses the potential of using this capacity to pursue a new form of social simulation. We will focus on a project that seeks to create synthetic social media platforms populated by thousands of interacting LLMs, calibrated and validated to behave as real-world social media users. This allows us to experimentally examine how designs and algorithms of social media platforms shape behavior and public conversations, and to identify how social media could be redesigned to have better societal outcomes.

Additional information – Petter Törnberg is Assistant Professor in Computational Social Science at the University of Amsterdam. He studies the intersection of AI, social media, and politics, and draws on computational methods and digital data for critical inquiry. His recent books include “Intimate Communities of Hate: Why Social Media Fuels Far-Right Extremism” (with Anton Törnberg), and “Seeing Like a Platform: An inquiry into the condition of digital modernity” (with Justus Uitermark, in print.)