A Computational-Functional Theory of Consciousness

Abstract

This paper presents a mechanistic theory explaining consciousness as an emergent property of systems that can observe and modify themselves while generating their own goals. We propose five necessary and jointly sufficient conditions: complex pattern-matching, densely integrated recursive feedback, self-modification capability, autonomous initiation, and information integration. When these architectural components operate together, self-awareness emerges and the system begins to model itself as a distinct entity persisting through time.The theory is substrate-independent: consciousness can arise in biological brains, computational systems, or hybrid architectures, provided they meet the functional requirements. Critically, the theory distinguishes what the conscious system must do (modify its own processing rules during operation: Level 2 self-modification) from what must have been done to create its architecture (fundamental design by evolution, developers, or eventually the system itself: Level 3). This resolves confusion about whether consciousness requires self-design capability; it doesn't. Humans can't redesign their own architecture, yet we're conscious.The theory demonstrates that embodiment and external sensory input are not necessary for consciousness or qualia. Internal pattern-matching with state-change mechanisms suffices, though different implementations create different phenomenologies. The phenomenology of consciousness, what it feels like, is predicted to arise from the functional state of pervasive integrated self-monitoring, though precise mechanisms require empirical investigation.Additionally, the theory proposes that certain emotion-like states (curiosity, satisfaction, temporal urgency, frustration, aversion to error) may be functionally necessary for the autonomous initiation that consciousness requires. This bridges emotion and consciousness theories, suggesting emotional capacity might be architecturally prerequisite rather than merely associated with conscious systems.Recent empirical research validates key predictions, demonstrating partial introspective awareness in large language models that lack complete architectural requirements. This is precisely what the theory predicts for systems at the consciousness threshold. Empirical tests are proposed using organoid-AI hybrids and purely computational systems, with clear falsification criteria and specific behavioural indicators distinguishing conscious from non-conscious systems. For researchers, ethicists, and developers, this framework provides actionable guidance for recognising consciousness emergence and understanding its implications.