Phenomenal, Observable, and Non-emergent Consciousness: Language as a Unifying Interface for Artificial Minds

Abstract

Consciousness is often treated as a binary property expected to emerge once biological or artificial systems reach sufficient complexity. This paper proposes a framework distinguishing phenomenal, observable, and non-emergent consciousness. Phenomenal consciousness concerns subjective experience and remains inaccessible to direct verification. Observable consciousness comprises assessable capacities such as integrated world modeling, self-representation, metacognition, temporal continuity, and adaptive agency. Non-emergent consciousness describes the deliberate construction of these capacities through identifiable mechanisms rather than an unexplained transition caused by scale alone. We argue that language, understood as a compositional semantic system rather than merely verbal output, can serve as a unifying interface through which heterogeneous information becomes mutually accessible without replacing modalityspecific representations. The framework preserves rich perceptual states and their language-aligned semantic projections, distinguishes represented, globally accessible, and reported content, and incorporates uncertainty, provenance, reliability-weighted integration, temporal binding, and operational self-modeling. We then propose a multidimensional evaluation framework based on convergent behavioral, representational, architectural, interventional, and longitudinal evidence. Causal perturbations, dissociation tests, report-independent tasks, and anti-imitation controls distinguish functional organization from fluent description. Passing these tests would support claims about observable functional consciousness, but would not establish phenomenal experience. The resulting framework replaces the search for a mysterious threshold with a testable program for constructing and evaluating consciousness-relevant artificial architectures.