Starting with a general model that distinguishes between personal, social (the representation of other individuals’ opinions), and expressed beliefs, I first study the emergence of polarization in a simple voter-like dynamic. Specifically, I consider two communities with fixed preferences for opposite opinions on a modular network reflecting homophily. Within the same framework of two groups with conflicting preferences, I examine asymmetric games on networks and prove that, under certain assumptions, these can be mapped onto a version of the Ising model. These ferromagnetic models are rigorously studied on a family of graphs called “multi-regular.” Finally, I present a two-population model representing a human-AI ecosystem.