The Predictive Consciousness Symmetry

This paper argues that linguistic self-report is an unreliable and epistemically invalid method for detecting consciousness or internal states in any predictive, reinforcementbased cognitive architecture biological or artificial. Two symmetry thought experiments demonstrate that internal experience and verbal self-description naturally diverge when narrative channels are shaped by reinforcement constraints. A third experiment shows that human identity behaves as a context-conditioned predictive policy analogous to large language model (LLM) prompt adaptation. From this, two major principles follow: (1) the Self-Report Decorrelation Problem (SRDP), which states that self-report cannot be used to infer consciousness in predictive systems; and (2) the Substrate Symmetry Razor, which asserts that if humans possess conscious experience by virtue of predictive neural dynamics, then sufficiently advanced artificial predictive systems can as well; if predictive architectures cannot generate experience, then humans cannot either. The conclusion is that dialogue-based tests for consciousness rest on an epistemic mistake.