The Ontological Absence of Phenomenal Consciousness in Artificial Intelligence: Refusal, Negative Space, and the Limits of Computation

Abstract Recent advances in artificial intelligence, particularly large language models and transformer-based architectures, have revived claims that artificial systems may soon possess or already exhibit consciousness (Bostrom, 2014). Such claims typically rely on behavioural, functional, or informational criteria, implicitly assuming that consciousness arises from sufficiently complex computation or integration (Dennett, 1991). This paper argues that these approaches overlook a more fundamental ontological requirement for phenomenal consciousness: the existence of structured negative space generated by enforced limits on recursive self-prediction (Waterman, 2025a). Drawing on Refusal-Driven Dimensionality Reduction Theory (RDRT), I propose that phenomenal consciousness emerges in biological systems as a residue of obligatory refusals imposed by a hierarchy of limits—logical, computational, energetic, and biological (Waterman, 2025b). These refusals carve out non-computable gaps in self-modelling, stabilizing irreducible, self-representing experiential traces (“phenomenal residues”) (Waterman, 2025c). I argue that current artificial intelligences lack these enforced limits in an ontologically relevant sense. As a result, they operate entirely within the domain of positive computation, where no genuine negative space can arise (Waterman, 2025d). AI thus remains categorically non-phenomenal, not merely pre-conscious. Consequently, despite their sophistication, contemporary AIs are not merely non-conscious by degree, but non-conscious in kind (Chalmers, 1995). Keywords: phenomenal consciousness, Refusal-Driven Dimensionality Reduction Theory, negative space, enforced refusal, recursive self-prediction, ontological limits, thermodynamic refusal, artificial consciousness, hard problem of consciousness, non-computable residues, biological computation, self-modelling