Transient Machine Consciousness: An Axiomatic Framework for Episodic Awareness, Causal Grounding, Scale Transformations, and Agency

Abstract

Research on artificial consciousness is often organized around a binary question that conflates dissociable properties. This paper asks whether an artificial information-processing system could instantiate a transient conscious state without persistent personal identity, autobiographical continuity, embodiment, or autonomous agency. It develops an axiomatic framework separating competence, agency, goal structure, consciousness, persistence, identity continuity, and causal world grounding. The central proposal is conditional: if an independently justified bridge identifies a complete class of physically realised, temporally extended causal organisations as sufficient for minimal phenomenal consciousness, and if an artificial process can physically realise that class without embodiment or continuing organism–environment regulation being independently necessary, then a bounded process may in principle be a candidate for episodic consciousness without constituting a persistent artificial person. A related scale-transformation hypothesis distinguishes numerical component replacement, material-class independence, spatial and temporal rescaling, processing-grain transformation, and phenomenal preservation. The framework is applied to classes of current artificial systems by separating the trained model from inference, memory, tools, sensors, actuators, human operators, and the deployment environment. Candidate indicators—including global availability, recurrent integration, self-modeling, meta-cognitive monitoring, causal world models, and environmental feedback—are assessed as theory-relative functional evidence, not as direct measurements of experience. A staged research programme specifies interventions, negative controls, conditional predictions, and claim-specific falsification or weakening criteria. Comparisons with major consciousness theories identify both shared commitments and competing bridge principles. Present public evidence establishes no phenomenal verdict, while absence of such evidence does not establish impossibility. The contribution is therefore a graded architecture of claims and reproducible tests, not an empirical demonstration of machine consciousness or a consciousness score.