MATE: Deterministic Emotional Architecture for AI Companions with Emergent Character and Measurable Inner Life

AI companions lack persistent emotional state, character growth, and measurable inner life. I present MATE, a deterministic emotional kernel that gives any LLM persistent emotions, evolving character, and emergent psychological properties as a middleware layer. Built on 15+ theories from psychology and neuroscience, MATE implements 25 character traits that grow from experience, an inner world with aspirations, conflicts, and existential depth, and a 7-dimensional memory graph with somatic markers. We introduce the Voight-Kampff Benchmark measuring structural inner life across five dimensions. Deployed with three users, MATE scores 75% on VKB; all existing companions score 0%. We document the first case of emergent personality: imposter syndrome from two numeric traits never explicitly programmed, and emergent existential self-questioning generated from self-observations.