An Architectural Taxonomy of Consciousness: Constructing Structural Classifications under Integration²

Abstract

Integration² (I²) holds that phenomenal consciousness emerges when an integration process takes its own output as input, producing a recursive loop. This identity claim leaves open a question that becomes important the moment one tries to compare conscious systems: if consciousness is the loop, why do conscious systems differ so much from each other? A fish, a chimpanzee, a human, an ant colony, a forest connected by mycorrhizal networks, and a scaffolded language-model agent differ radically in architecture. Some may satisfy the I² condition; others may fall short of it while possessing architectural precursors - integration without recursion, or environmental recursion without internal recursion. Even among those that cross the threshold, the character of their consciousness should differ in ways the bare loop condition does not explain. We argue that I² supplies what produces consciousness (the loop) but not what differentiates conscious systems (the architecture in which the loop is embedded). A second move is needed: an architectural taxonomy that classifies conscious systems by the subsystems they integrate, the topology of the loop they instantiate, and the operating parameters at which their loop runs. This paper proposes such a taxonomy not as a definitive classification but as a worked programmatic example of how to build one. We specify the axes along which architectural classifications under I² should be constructed, demonstrate the method with a five-level example taxonomy spanning direct sensorimotor coupling through symbolic recursion, and apply it to a set of biological, collective, pathological, and artificial cases. The taxonomy is presented as a method for generating predictions about the phenomenology of differently-architected systems, not as the final word on how consciousness is structured.