Related Experiment Video
Updated: Jan 15, 2026
Implementation of a Real-Time Psychosis Risk Detection and Alerting System Based on Electronic Health Records using CogStack
Published on: May 15, 2020
Challenges in Implementing a Mobile AI Chatbot Intervention for Depression Among Youth on Psychiatric Waiting Lists:
Junichi Fujita1, Yuichiro Yano2, Satoru Shinoda3
1Department of Child Psychiatry, Yokohama City University Hospital, 3-9, Fukuura, Kanazawa-ku, Yokohama, 236-0004, Japan, 81 0457872800.
Background:
The mental health of children and adolescents is a growing public health concern, with increasing rates of depression and anxiety impacting their emotional, social, and academic well-being. In Japan, access to timely psychiatric care is limited, leading to extended waiting periods that can range from 3 months to a year. Artificial intelligence (AI)-driven chatbots, such as emol (Emol Inc) that integrates acceptance and commitment therapy, show potential as digital solutions to support young patients during these waiting times. The AI chatbot emol was selected based on a comprehensive review of Japanese mental health technology apps, including in-person evaluations with company representatives.
Objective:
This exploratory parallel-group randomized controlled trial examined the feasibility of using an AI chatbot emol with pediatric and adolescent individuals on psychiatric waiting lists.
Methods:
Participants aged 12-18 years were recruited from 4 hospitals in Kanagawa Prefecture and randomly assigned to either an intervention group, receiving 8 weekly chatbot sessions, or a control group, receiving standard mental health information. The primary outcome was the change in scores on the 9-item Patient Health Questionnaire from pre- to postintervention. Secondary assessments, such as voice and writing pressure analysis, provided additional engagement metrics, with data collected at baseline, during the intervention, and at week 9.
Results:
Of the 96 eligible individuals on psychiatric waiting lists, 8 expressed interest and 3 provided initial consent. However, all participants subsequently withdrew or were excluded, resulting in no evaluable data for analysis. Low engagement may have been influenced by the perceived irrelevance of digital tools, complex protocols, and privacy concerns.
Conclusions:
Significant barriers to engagement suggest that digital interventions may need simpler protocols and trusted environments to improve feasibility. Future studies could test these interventions in supportive settings, like schools or community centers, to enhance accessibility and participation among youth.
