Cognitive agency in socio-cultural evolution - A proposal to ground social modelling in the active inference framework

Current approaches in modelling cultural evolution have been highly successful in studying how population-level patterns in the diffusion of cultural traits emerge from individual cognitive biases. However, they do so by abstracting away the cognitive agency of individual humans, understood as the many processes by which we understand, integrate, and (re)construct culture. We argue that this abstraction - while necessary to frame many relevant research questions as computational experiments - impedes our ability to investigate the development of cultural traits. In particular, this prevents the description of the specific processes underlying the evolution of novel traits, and thus the study of open-ended cultural evolution. We propose the ecological-enactive Active Inference model of cognition as a conceptual and computational foundation for the implementation of cognitive agency and open-ended cultural evolution in silico. We further discuss the methodological implications of such an approach, and its relevance for the discipline of cultural evolution.