Independent research · Version 1.0 · 26 September 2026
Intelligence, Consciousness, and the Architecture of Artificial Minds
By Jason Christopher Rae
Overview
The paper asks what evidence can support claims about artificial minds, while keeping observable performance, internal cognitive processes, and subjective experience distinct.
Fluent language and successful behaviour can show capability, but do not by themselves establish semantic understanding or phenomenal consciousness. At the same time, describing current models as mere surface pattern matching can overlook the internal computations and representations they use.
Future systems could be studied as recurrent architectures in which language interacts with perception, persistent world models, differentiated memory, counterfactual simulation, metacognitive monitoring, goals, and action. Graph-structured memory is one possible implementation, not a privileged route to consciousness.
The framework suggests intervention-based questions for AI research and system evaluation: what does each component enable, how stable is it under interruption, and which claims are supported by behaviour versus internal mechanism? It is a research proposal, not a validated system design or consciousness test.
About this paper
Version 1.0 brings together work from machine learning, cognitive science, neuroscience, and philosophy of mind. It proposes five testable hypotheses and an evidential ladder for evaluating increasingly persistent and self-monitoring forms of artificial cognition. The paper describes itself as neither a systematic review nor peer reviewed.
Suggested citation
Rae, J. C. (2026). From Articulation to Cognition: Intelligence, Consciousness, and the Architecture of Artificial Minds (Version 1.0). Independent research paper. Jason Rae.