AI is not as conscious as humans: Essentialism in mental state attributions to artificial systems.

Abstract

Artificial systems increasingly behave in ways that resemble humans. But is that sufficient to make us see them as conscious, or as having related mental states? We contrast two broad possibilities. People may rely on a single general principle: the more appropriately an agent responds to its environment, the more conscious it is, whether human or artificial. We call this a functionalist attribution. Alternatively, people may reserve consciousness for certain kinds of beings: even an artificial system that behaves just as responsively as a human is judged less conscious. This is an essentialist attribution. A further question is whether this pattern is unique to consciousness or extends to closely related constructs such as awareness. In two pre-registered sets of studies (five experiments, N = 1092), UK participants read scenarios describing human or artificial agents whose responsiveness to spatial, social, and temporal features of the environment varied, and rated the agents' consciousness and awareness, either separately or in conjunction. Attributions of consciousness followed an essentialist logic: at similar levels of responsiveness, participants consistently attributed greater consciousness to humans. Attributions of awareness followed a functionalist pattern: for both humans and artificial agents, awareness increased steadily with responsiveness, although humans were rated more aware at the highest levels. These findings suggest that people use different inferential rules for different mental concepts. Consciousness is treated as depending on what an agent is; awareness largely on what an agent does.