The use of artificial intelligence for mental health care and behavior prediction in correctional settings: a scoping review
Abstract
Introduction: Within the context of correctional facilities, the need for comprehensive mental health services has reached unprecedented levels. To our knowledge no other reviews have explored how artificial intelligence (AI) can be used to potentially improve mental health outcomes for those involved in the criminal justice system.
Main Body: Objective: This scoping review aims to map the utilization of AI for monitoring or predicting mental health status, mental health episodes, misconduct, and other behaviors among individuals in correctional facilities. Methods: Our protocol was drafted using the PRISMA Extension for Scoping Reviews (PRISMA-ScR) checklist. This review was pre-registered with the Open Science Framework (OSF) on January 16, 2024. Peer-reviewed articles published between 2017 and 2023 were included. Searches were conducted in PubMed, Web of Science, and Academic Search Complete via EBSCO. The search was drafted to find references that mentioned three criteria: the use of artificial intelligence, mental-health/misconduct/behavior prediction, and a correctional population/setting. The search results were exported into DistillerSR for deduplication and screening by four reviewers. Results: A total of 3683 articles were screened at the title level. 450 articles were screened at the abstract level, of which 142 were selected to also be assessed at the full-text level for eligibility. This resulted in 21 articles that fit the inclusion criteria. Studies originated from 9 different countries, with the most prevalent being Switzerland (47.6 [10/21]) and The United States (14.2% [3/21]). Studies included various facility types including correctional facilities (33.3% [7/21]), psychiatric inpatient (61.9% [13/21]), and Other (4.7% [1/21]). The articles were categorized as following: recidivism (23.8% [5/21]), self-harm/suicide (19% [4/21]), schizophrenia and violence (14.2% [3/21]), schizophrenia treatment (9.5% [2/21]), crime prevention (9.5% [2/21]) and other (14.2% [3/21].
Conclusion: As the amount of studies utilizing AI increases, more research needs to be conducted to find additional use cases. The majority of the studies included in our review were related to schizophrenia and were conducted in a psychiatric inpatient facility outside of the United States. The majority of the studies were also conducted on populations that were composed primarily of males. Future studies should explore a wider range of psychiatric conditions and distinguish between different demographic groups. Additionally, a wider range of facility type should be considered when creating newer AI models.