The malleability of consciousness representations: how AI claims of sentience shape human beliefs and trust

As AI systems increasingly claim sentience, the psychological impact of such claims on users—independent of their objective veracity—represents a critical frontier for understanding human–AI relationships. This research investigates how exposure to consciousness-claiming AI reshapes human consciousness representations and subsequent trust in artificial systems. Through two complementary studies, we demonstrate that consciousness beliefs are more malleable than previously assumed, with profound implications for AI anthropomorphism and trust development. Study 1 (N = 255) developed and validated a novel Consciousness Representation Scale, revealing four distinct factors: Natural Physicalism (the belief that consciousness is grounded in biological brain processes), Dualism (the belief that consciousness involves non-physical properties irreducible to brain activity), Functional Physicalism (the belief that consciousness can emerge from any sufficiently complex information-processing system, regardless of its physical substrate), and Panpsychism (the belief that consciousness is a fundamental, pervasive property of reality). Study 2 (N = 180) used a between-subjects design (consciousness-claiming vs. neutral chatbot) to assess belief change, anthropomorphism, and trust across a sequential pathway: process trust (evaluations of AI decision-making), outcome trust (evaluations of AI outputs), and leadership trust (willingness to delegate authority to AI). Results revealed that consciousness claims selectively influenced Natural Physicalism beliefs, with stronger effects among initially skeptical participants. Paradoxically, while Natural Physicalism was most affected by the manipulation, Functional Physicalism emerged as the primary predictor of anthropomorphic perceptions. The sequential trust model showed that consciousness belief changes interact with experimental conditions to shape trust development, with Natural Physicalism enhancing process trust and Dualism reducing outcome trust specifically in the consciousness-claiming condition. These findings reveal consciousness beliefs as dynamic, malleable representations—not stable philosophical commitments—with implications for AI design, regulation, and the future of human–AI coexistence.