The Philosophical Zombie Refuted by LLM Reality: Behavioral Isomorphism and the Physical Necessity of Consciousness

Abstract

The philosophical zombie argument is a core modal argument in contemporary dualism against physicalism. Its central claim is that there could exist an entity that is physically and behaviorally identical to a human being yet entirely lacks subjective phenomenal consciousness—a "philosophical zombie." This entity is claimed to be logically conceivable, from which it is inferred that consciousness is a non-physical property. For decades, this argument has lacked empirical verification against real-world entities. This paper takes the Transformer-based large language model (LLM) as an observable, reproducible, and analyzable unconscious intelligent system, and conducts a systematic analysis from the perspectives of information theory and systems theory. (1) By distinguishing between structural-isomorphism zombies and functional-isomorphism zombies, we show that unconscious systems have inherent behavioral limits and cannot achieve global functional isomorphism. (2) We identify the zombie camp's conflation of surface-level behavioral simulation with global functional isomorphism, and question the modal-logical assumption that "conceivability entails metaphysical possibility." (3) We propose a three-dimensional framework of necessary conditions for consciousness—initiative, frame-transcendence, and selfhood—and draw on Integrated Information Theory (IIT) and the Multiple Drafts Model to support our "complexity threshold -> consciousness emergence" inference. We conclude that philosophical zombies are not metaphysically possible, and that the dualist zombie argument rests on a fundamental logical error.