PI: Claudio Altafini, Linköping University
co-PI: Emma Tegling, Lund University
This project investigates how opinions and decisions evolve in large networks that include both human participants and AI agents. A central motivation is that social interaction increasingly occurs on online platforms whose content flows and visibility are shaped by algorithms and automated agents. As AI-driven accounts become more capable of generating persuasive or human-like content, it becomes essential to understand how they alter information propagation, how to distinguish them from human users based on behavior and sentiment, and how their presence affects large-scale collective dynamics.
In this project, we will develop methods for identifying the structure and dynamics in opinion-dynamics models from empirical data. This will enable us to model interactions in human-AI networks and the corresponding influence asymmetries along with, for example, risks for spreading of mis- and disinformation. Methods to discern human from AI behaviors will be central to the question of model inference.
Finally, we explore how principles from social dynamics, such as principles for consensus-seeking and “wisdom of crowds”, can improve distributed optimization, estimation, and federated learning. By applying social learning principles to engineered settings, we hope to reap some of the benefits of the emerging complex human-AI landscape.
Project number: F1
