Humans Retain Consciousness Bias Against AI, Despite Behavioural Equivalence

A recent study suggests that individuals are significantly less likely to attribute consciousness to artificial intelligence systems compared to humans, even when both exhibit identical behavioural outputs. This research indicates a robust bias in how humans perceive the internal states of AI versus biological entities.

The study involved presenting participants with various scenarios depicting either a human or an AI engaging in specific behaviours. These behaviours included both low-awareness actions, such as solving arithmetic problems, and high-awareness actions, like expressing pain or gratitude. Participants consistently rated humans as more conscious than AI across all scenarios, irrespective of the complexity or emotional valence of the behaviour displayed.

Interestingly, the perception of awareness showed a different pattern. When AI systems performed high-awareness behaviours, such as articulating feelings of pain, participants were as likely to attribute awareness to the AI as they were to a human performing the same action. This suggests a disconnect: while specific behaviours might imply awareness, they do not necessarily lead to an attribution of full consciousness to AI.

The researchers employed a seven-point Likert scale for participants to rate human-like consciousness. Participants consistently rated AI as less conscious, even when the AI's responses mirrored those of a human counterpart. The findings highlight a fundamental difference in how human minds categorize and understand consciousness when applied to non-biological systems.

This perceptual gap between attributing awareness and full consciousness to AI could inform future human-AI interaction design. It raises questions about how societal norms regarding consciousness might evolve with increasingly sophisticated AI capabilities. What implications does this hold for the ethical development and deployment of advanced AI systems?