Perspectival Control Identity Theory

Abstract

Perspectival Control Identity Theory (PCIT) proposes a falsifiable, intervention-based identity program for phenomenal consciousness. The central claim is an a posteriori identity: phenomenal consciousness is identical to a specific kind of internal control variable, a Perspectival Control State (PCS)—a temporally extended, viability-weighted stream that makes an agent's competing needs comparable and coordinates a coalition of constitutive consumers whose control and learning depend on the stream and whose outputs feed back to shape it. PCIT makes two specific, testable bridge commitments. Degree of consciousness tracks how much the PCS stream causally matters for closed-loop viability regulation under intervention. Content tracks the decoder-indexed equivalence classes over PCS states and the induced similarity geometry determined by constitutive consumers. Advances in machine learning make this kind of "synthetic phenomenology" experimentally tractable: we can build agents-in-worlds with known internal organization, intervene on proposed PCS implementations and their decoders, and measure downstream effects on behavior and long-horizon viability. The payoff is a research program with concrete invariance and dissociation tests—and a principled scientific foundation for questions about AI moral status, animal sentience, and disorders of consciousness that currently lack one.