Effectiveness of AI-Based Chatbots and Digital Tools for Adult Mental Health: A Systematic Review

Abstract

Chatbots and other digital tools based on artificial intelligence are becoming more and more applicable to the field of adult mental health, but the evidence on their efficiency, user interaction, and safety is still partial. The purpose of this systematic review was to summarize existing studies on the efficacy of AI-based chatbots and digital mental health tools use among adults, and discuss engagement patterns and safety outcomes in particular. The search has been performed in the key medical and psychological databases to locate randomized controlled trials and controlled observational studies on the subject of adults (18 years and older). Predefined eligibility criteria were used to screen studies and the methodologic quality was evaluated with the help of established risk-of-bias tools. Because there was heterogeneity in which interventions and outcome measures were measured, a narrative synthesis was conducted on the effectiveness outcomes, whereas a thematic analysis was conducted on the engagement and safety results. Thirty-two studies, assessing a variety of AI-based interventions, most frequently chatbot-delivered cognitive behavioral therapy, were included in the review. On the whole, AI-based interventions showed small to moderate depressive and anxiety symptoms improvements, as compared to inactive controls. Participation was significantly different among studies and the adherence was higher in structured and guided interventions. The reporting on safety outcomes was inconsistent with not many studies systematically monitoring adverse events or crisis escalation. These results indicate that AI-based chatbots and digital tools can have a slight mental health advantage to adults, especially to depression and anxiety, yet the constraints are associated with the maintenance of engagement and the safety analysis. More strict studies will be needed to prove the long-term effectiveness and safe implementation.

Keywords: Artificial intelligence, Chatbots, Depression, Digital mental health, Systematic review

https://doi.org/10.5281/zenodo.20209915