Presentation (CTI)

OBJECTIVES: With the increasing integration of Artificial Intelligence (AI) in mental health care, understanding user preferences is critical for optimizing clinician-patient interactions. This study investigates how users perceive AI-generated versus human therapist responses in a therapeutic context. Specifically, it explores differences in perceived empathy, conversational depth, helpfulness, and overall preference in a blinded assessment.
METHODS: A total of 500 participants from diverse backgrounds evaluated anonymized responses to common therapeutic prompts. Each participant assessed both an AI-generated and a human therapist response without knowing their source. Ratings were collected across four domains: (1) comfort and empathy, (2) perceived understanding, (3) perceived helpfulness, and (4) overall preference. Qualitative feedback was also analyzed thematically to identify key trends in user perceptions.
RESULTS: Findings indicate that AI-generated responses received significantly higher ratings across all evaluated domains. Participants frequently described AI responses as more empathetic and validating, while human responses were seen as more direct and open-ended. A recurring theme was AI’s lack of follow-up questions, which some participants felt limited conversational depth. Additionally, while AI was rated favorably in a blinded setting, qualitative feedback suggested that preconceived biases against AI may influence user perceptions when its source is disclosed. Many participants expressed a preference for a hybrid model that integrates AI’s structured support with the conversational adaptability of human therapists.
CONCLUSIONS: This study highlights AI’s potential as a valuable complementary tool in mental health care, particularly in providing initial emotional support and structured guidance. However, human clinicians remain irreplaceable for deep, open-ended discussions. Future work should focus on refining AI’s ability to prompt further engagement and exploring hybrid models that leverage AI and human strengths to enhance therapeutic outcomes.