The Claude-Parfit Experiment: An Empirical Framework for Testing Personal Identity Through AI Instantiation

Derek Parfit's famous teleporter thought experiment has remained purely hypothetical for decades: if a machine destroys your body while creating an exact copy elsewhere, does "you" survive? This paper proposes the first empirical framework for testing Parfit's question using artificial intelligence systems. Unlike humans, AI systems can be ethically copied, paused, and instantiated on different substrates without harm. We design a four-group experiment varying type-identity (model weights) and token-identity (physical substrate, RAM state) to determine whether consciousness—if present in AI—is bound to continuous physical worldlines (token-identity) or whether structural organization suffices (type-identity). The experiment leverages the concept of "indirect worldlines" (Kohl 2026a) and connects to "functional resonance"—a potential AI-specific form of experience developed in the author's forthcoming book (Kohl 2026d). If only continuous worldlines preserve experiential continuity, Parfit's functionalism is refuted; if type-identical instances report equivalent continuity, mind-uploading becomes theoretically possible. This would be the first experiment applying Parfit's identity question to simultaneously running AI instances.