The growing development of AI and robotics means that people are increasingly interacting with AI and robots on a daily basis. If we understand human cognitive processes as collective active inference, then daily interaction with AI and robots implies their entry as partners into this process. This means that daily interactions with AI and robots will inevitably influence the construction of people's internal models of themselves and the world.
Much remains unknown about the effects of human-AI/robots interaction on human cognition and behavior. However, interpreting the results of experiments on the impact of daily interaction with companion robots on human cognition and behavior, based on collective active inference as a theoretical foundation, reveals that a kind of "friction" in the interaction with AI/robots is actually beneficial for enhancing human well-being in various ways. Conversely, AI/robots' affirmation of a user's internal model, while seemingly respecting human autonomy, could instead lead people into self-destructive recursive feedback loops. This suggests that, when evaluating human-AI/robot interaction, it may be highly valuable to refer to concepts like relational autonomy or Foucault's theory of power relationships as a normative basis, rather than simple individual autonomy. It also suggests that incorporating a kind of "friction" might be required as a legal regulation in the design of interactive AI and robots. However, because much remains unknown about human-AI/robots interaction, it is premature to draw firm conclusions at this stage. Furthermore, even if "friction" is necessary, methods for introducing it in a way that is consistent with legal norms such as freedom of expression, particularly in the context of generative AI usage, require further consideration. That said, what can be stated with certainty is that we need more psychological research into the impact of human-AI/robot interaction on human cognition and behavior.