Artificial Intelligence and Machine Learning in Mental Health Services: An Environmental Scan

Key Message

• There has been an increase in investment, funding, and interest in the use of artificial intelligence (AI) in the mental health area, especially over the past five years. The types of AI in use for mental health problems and illnesses include neural networks, support vector machines, logistic regression, linear discriminant analysis, and random forests.
• Currently, the use of AI for mental health is limited in a clinical environment. Consultations revealed that the majority of AI applications are in research and development and have not been expanded to clinical or patient use. The applications of AI in mental health include conversational agents, computerized adaptive testing, diagnosis of mental health conditions, prediction of behaviour, and prediction of prognostic outcomes in treatment.
• The professions currently using and researching AI are primarily researchers in the fields of psychology and computer science. There are limited AI applications available commercially, but some AI applications, such as chatbots, are available for the general public through mobile application stores such as Apple and Google Play.
• Four key domains of research and development in the use of AI for mental health are diagnosis, prognosis, public health, and clinical administration; most of the identified research is in the area of diagnosis.
• Research and development initiatives for mental health diagnosis using AI include the use of biomarkers, neuroimaging, genetics, metabolomic data and proteomic data to diagnose or detect mental illness, and new data collection methods such as smartphone-based data collection or wearable sensors paired with AI applications. Research initiatives for treatment include improvements in currently available chatbots, the creation of new AI chatbots, and research on whether individuals will respond to particular treatments based on their specific characteristics. Prognosis initiatives include predictions of mental health trajectories and potential future costs of illness.
• Trends in the development of AI for mental health include chatbots for mental health treatment, increased explainability of models, and an increase in wearable devices and smartphone-based sensors for data collection.
• In regard to specific life stages, there are some identified AI algorithms used in children and youth, including Kids’ Help Phone’s crisis text service, Bark’s AI parental control application, and AI algorithms for the diagnosis of internalizing disorders and bipolar disorder. In older adults, there are several AI applications to combat loneliness in seniors using companion robots, and some AI specific to seniors for diagnosing depression.
• Policy and program initiatives in mental health–based AI include the translation of lab-based research initiatives into clinical applications. This translation may require careful planning to ensure that ethical standards are met, that assessments are carried out to determine the suitability of the AI for each aspect of mental health diagnosis and treatment, and that generalizable and culturally sensitive algorithms are created

This environmental scan was commissioned by The Mental Health Commission of Canada to address the role of AI in mental health services. This report is a companion to a Rapid Response review on clinical effectiveness and guidelines for AI in mental health [Artificial Intelligence and Machine Learning in Mental Health Services: A Literature Review. Ottawa: Canadian Agency for Drugs and Technology in Health (CADTH), Mental Health Commission of Canada (MHCC); 2021 June.]