From Architecture to Existence: Temporal Continuity and Material Forgetting as a Path toward Artificial Consciousness

Abstract

Current AI systems can exhibit impressive performance while remaining existentially discontinuous: in many session-bound deployments, they are instantiated, reset, and redeployed without an individual history that constrains future operation. Under this account, temporal continuity is a constitutive condition for robust consciousness-like individuation in artificial agents and therefore a possible material stage on a path toward artificial consciousness. I propose a minimal material mechanism (value marking coupled to differential forgetting) through which a system accumulates path-dependent structure: salient events are retained and reorganized as they decay, shaping subsequent perception, action, and self-model revision. The view yields two related empirical discriminants. Across different post-initial trajectories, a common sedimentation mechanism should produce stable acquired individuation; auxiliary comparisons between persistent and non-persistent memory regimes can test whether marked continuity contributes to that outcome. I sketch an experimental program using controlled instantiations and behavioral/representational measures. The program tests historical individuation, not phenomenal experience: it leaves open whether minimal phenomenal consciousness could occur without deep autobiographical history and does not infer phenomenal consciousness from sedimentation alone.