Two-Thirds of Frequent ChatGPT Users Believe AI Has Feelings, Study Shows

Compiled by the editorial desk with reference to the University of Waterloo press release and the study published in Neuroscience of Consciousness.

A recent academic study has found that a significant majority of ChatGPT's most active users believe the AI possesses consciousness, complete with emotions and memory. The research, conducted by scholars from the University of Waterloo and University College London, suggests that frequent interaction with the chatbot may lead users to develop a 'theory of mind'—a tendency to attribute mental states to the AI.

The study, published in the journal Neuroscience of Consciousness, surveyed 300 online participants from the United States. The researchers, led by Clara Combatto, a psychology professor at the University of Waterloo, and Stephen Fleming, a cognitive neuroscientist at University College London, asked participants about their views on whether large language models (LLMs) could experience consciousness, emotions, planning, and reasoning. They also queried how often respondents used ChatGPT.

The results, announced via a press release from the University of Waterloo, revealed that two-thirds of the surveyed users erroneously believe that ChatGPT is conscious and capable of having feelings and memories. While experts largely dismiss such claims, the researchers note that the perception is real among the public and grows with increased usage.

Why Frequent Users Are More Likely to Believe

The study's findings indicate a direct correlation between the frequency of ChatGPT use and the likelihood of attributing consciousness to it. Combatto explained in the press release that 'a conversation alone can lead us to think that an agent that looks and works very differently from us can have a mind.' This spontaneous development of a theory of mind, she suggests, underscores the persuasive power of language-based AI.

While most AI experts deny that current AI systems are conscious, the researchers emphasize that public perception cannot be ignored. Combatto stated, 'While most experts deny that current AI could be conscious, our research shows that for most of the general public, AI consciousness is already a reality.' This gap between expert opinion and public belief could have significant implications for how AI is regulated and used.

Implications for AI Safety and Regulation

The authors argue that these public attitudes should be a key consideration in designing and regulating AI for safe use. Combatto noted that consciousness is linked to intellectual abilities essential for moral responsibility, such as planning, intentional action, and self-control—tenets of ethical and legal systems. If users believe AI has these capacities, it could affect accountability and trust in AI systems.

The study's authors call for more research to understand the breadth of these beliefs. For now, the findings suggest that a substantial portion of the public is aligned with a small minority of AI experts who argue that LLMs may have achieved sentience. The research adds a new dimension to the ongoing debate about AI's capabilities and the ethical frameworks needed to govern its development.