A Unified Computational Substrate for Consciousness-Relevant Cognition: Operational Predictions from Neural Manifold Geometry

Abstract

This paper reports a unified computational framework grounded in neural manifold geometry, fuzzy attractor dynamics, cross-manifold interference matching, and value-weighted attention allocation (VIRT) that spans perception, concept formation, planning, information fusion, and attention allocation. The framework was developed to address practical problems in speech understanding, concept learning, and planning under uncertainty, not to explain consciousness. Its relevance to consciousness follows from what the architecture requires: recurrent cross domain connectivity, goal-driven resource allocation, self-referential monitoring, and graded novelty detection. These are properties that major theories of consciousness predict and that the empirical literature on consciousness-relevant behavior demands. The paper's method is operational. For each of the eight most rigorously developed philosophical objections to machine consciousness, from Chalmers' hard problem to Searle's Chinese Room, Block's access/phenomenal distinction, and Tononi's Integrated Information Theory, we translate the objection into a behavioral or architectural prediction and examine whether the framework satisfies it. For every objection that yields to operationalization, it does. For the objections that do not yield, principally the explanatory gap and the zombie argument, we state precisely what they require and why current methods cannot address them. The paper's contribution is methodological as much as substantive. It demonstrates that the standard philosophical objections to machine consciousness, when translated into testable predictions, are addressable by architectural analysis grounded in independent empirical evidence. The framework does not resolve the hard problem of consciousness. It provides a substrate from which all operationalizable consciousness-relevant properties can be derived, and from which genuinely unresolved questions can be more precisely identified.