The role of large language models in mental health : a scoping review

• Main Author: Gomes, Tiago
• Publication Date: 2024
• Format: Master thesis
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10400.14/47745
• topic: Large language models (LLMs); Mental health; Applications; Clinical data analysis; Generative pre-training (GPT); Screening; Risk detection; Treatment; Recommendations; Ethical challenges; Data privacy; Communication; Therapeutic interventions; Natural language processing (NLP); Artificial intelligence; Grandes modelos de linguagem (GMLs); Saúde Mental; Aplicações; Análise de dados clínicos; Pré-treino generativo (GPT); Rastreio; Deteção de risco; Tratamento; Recomendações; Desafios éticos; Privacidade dos dados; Comunicação; Intervenções terapêuticas; Processamento de linguagem natural (PLN); Inteligência artificial

Mental health disorders affect nearly one billion individuals worldwide, with a growing prevalence over year, caused in part due to stigma and lack of treatment causing a high burden for healthcare systems. In this context, Large Language Models (LLMs), such as GPT-4, have emerged as transformative tools with the potential to improve mental health care. This master thesis conducts a scoping review of research published from 2023 onwards to explore the current applications of LLMs within the realm of mental health, with the objective of offering a thorough overview of their existing and prospective applications in clinical practices and data analysis. While LLMs hold promise in improving mental healthcare through early diagnosis, treatment planning, and the communication between patients and clinicians, this review has also pointed out the limitations the current models have, such as the high-risk mental health crisis, an inability to understand emotional subtleties which are crucial in the treatment of mental health, and concerns about ethics and data privacy in relation to the inherent biases of the training data. For future research, key areas include enhancing LLMs' skills in recognizing crises, creating tailored models for mental health for higher sensibility, and addressing significant ethical issues like bias and data privacy, which are essential for the gradual integration into the mental health field. LLMs integration in the mental health sector require a careful integration in order ensure patient safety and maintaining trust. It is imperative to have human oversight while using these tools, especially in high-risk clinical environments.