Logical and computational approaches to dynamics of intelligent interaction
Edoardo Baccini · 2025
Our ideas, opinions and behaviors are partly shaped by our social surroundings, and in turn, our surroundings are shaped by our own ideas, opinions and behaviors. How do our beliefs and our actions change because of our interactions with others? And how does our environment change as a result? This thesis studies the interactive dynamics of intelligent agents by considering three themes. First, we study how one’s similarity to others affects the spread of behaviours and opinions in social networks and how it modifies one's social connections over time. Second, we investigate whether agents are able either individually or collectively to learn the truth if they exhibit so-called conservatism, i.e., they adopt strategies to maintain a good fit between what they come to believe after encountering new information and what they believed before. Finally, we explore whether deciding to trust or to reject pieces of information based on their internal coherence may be an effective way for an agent to filter out deceiving information and form true beliefs. Each question above is considered from a model-oriented perspective, whereby it is first transformed into a mathematically precise problem and then investigated using analytical methods or computer simulations. The formal tools employed are varied and comprise dynamic logic tools to study social network dynamics, dynamic epistemic logic tools to model belief change, and Bayesian networks and agent-based models to represent the role of cognitive biases and coherence in the evaluation of information.