Digital mental health tools need human touch

Scientists tested 75 participants across three types of short online interviews about their wellbeing: one with a semi-scripted live interviewer, one with a live but fully scripted interviewer, and one using fully automated pre-recorded videos. Participants rated both live interviewers as significantly more empathetic than the automated version, and facial recognition software detected higher expressions of joy in the live interview conditions.

As mental health services evolve, this research suggests that hybrid models combining automation with human interaction may be the most effective path forward. Looking ahead, the findings also offer a roadmap for future AI development.

Dr Thomas Nyman, lead author from the University of Reading, said: "Even when interviewers followed a script, it appears that the simple fact that a real person was present made a meaningful difference.

"With advances in AI, we may soon be able to design digital agents that successfully simulate the human qualities participants responded to in this study -- such as empathy, flexibility, and presence. Understanding what people need emotionally is the first step toward building technology that can meet those needs in human-like ways."