Select Language:
A recent study has revealed that OpenAI’s ChatGPT has successfully passed a Turing test in the realm of therapy, raising new questions about the role of artificial intelligence in providing emotional support. According to a blog post by technology media outlet The Decoder, participants found it challenging to distinguish between the therapeutic advice given by ChatGPT and that offered by human therapists. In fact, many perceived the AI’s responses as exhibiting greater empathy.
The study, which involved 830 participants, utilized the Turing test framework, where individuals were asked to identify whether responses came from a human therapist or ChatGPT. The results indicated that participants had a slightly better than random chance of accurately identifying the responses, with a success rate of 56.1% for human therapist replies and 51.2% for those generated by ChatGPT.
Researchers noted that ChatGPT scored higher than human experts in various aspects, including therapeutic alliance, empathy, and cultural competence. The AI’s responses were generally longer, more positive in tone, and contained a richer vocabulary, making them seem more detailed and empathetic.
Interestingly, the study highlighted a bias among participants: when they believed they were reading AI-generated responses, they tended to rate them lower, regardless of the real source. Conversely, when AI responses were misidentified as coming from human therapists, they received higher ratings.
This is not the first study to suggest that AI may have potential in counseling roles. Research conducted by the University of Melbourne and the University of Western Australia found that ChatGPT provided more balanced, comprehensive, and empathetic advice on social dilemmas compared to human columnists, with preference rates between 70% and 85%. However, despite the higher ratings for AI responses, most participants still expressed a preference for advice from human counselors.
ChatGPT
Perplexity
Gemini AI
Grok AI
