Artificial intelligence (AI) approaches, such as machine learning and deep learning, and their application in natural language processing (NLP) tasks, have introduced novel use cases in health and medical research. However, their use in research on the health needs of people who have contact with the criminal justice system is incipient. We conducted a rapid review using a systematic search of PubMed, Medline, EMBASE, and Scopus databases for original research into AI applications in justice health research. We reviewed abstracts and full-texts, excluding manuscripts where AI was used solely in justice or health areas, and manuscripts with unclear inclusion of justice-involved people in the study. We performed a narrative synthesis of the extracted data on characteristics and AI applications of the included studies. The 11 included manuscripts predominantly focused on machine-learning-based predictive applications, such as in the prediction of suicidality, self-harm, opioid overdoses, and length of stay in hospitals. There were limited examples of deep learning and natural language processing applications. Most studies used population-based sampling. Our rapid review revealed limited methodological and geographical diversity in AI-powered justice health research. Future research should focus on the transparency and performance of AI-powered research to improve the health of justice-involved people.
Calais-Ferreira et al. (Sun,) studied this question.