Contemporary artificial intelligence, largely shaped by large-scale architectures optimized for formal reasoning, has achieved impressive success in cognitive tasks. Yet it remains fundamentally disembodied. This work argues that this limitation is not a matter of scale, but of architecture. We introduce CAINE, a cognitive framework that reorients artificial intelligence away from abstract cognition toward the affective grounding of social behavior. This shift rests on the view that emotion is not opposed to reason, but is one of its necessary conditions.The framework is built on a dual foundation: an agent architecture grounded in a dynamic emotional matrix functioning as a regulatory mechanism for entropy minimization, and a formal ontology of social interaction "the Room Framework" which models the emergence of intersubjectivity as a form of entropy reduction through relational coherence and synchronization. The core principle of this approach is cumulative social dissonance. Psychological and social evolution are driven not by convergence of perspectives, but by the historical accumulation of divergences between the lived experiences of interacting agents. Each relationship is constituted by irreducible asymmetry: no relational state is ever fully shared, and this asymmetry is preserved by design as a response to the problem of intersubjectivity. Each agent possesses a complete inner life through which social actions acquire meaning from lived experience. Memory is not a neutral repository, but a context- and affect-dependent process in which experiences are transformed, condensed, and reactivated according to their emotional significance. Personality itself is not fixed, it gradually reorganizes in response to psychosocial feedback, allowing lasting imprints to form, particularly in situations of conflict or trauma. Through simulations, we show that these principles are sufficient to generate rich social dynamics from first principles. Relationships exhibit emergent life cycles, psychological wounds persist over time, and affective meaning arises organically from interaction rather than being predefined. These phenomena are not explicitly programmed but emerge from the enduring divergence of subjective perspectives. Our results suggest that subjective asymmetry is not a flaw to be eliminated, but a necessary condition for the emergence of meaningful social dynamics. This work points toward a new paradigm for socially aware artificial intelligence and offers a falsifiable computational perspective on enactivist and phenomenological accounts of social cognition.
BRICHE, Georges, Patrick, David (Wed,) studied this question.