[논문 리뷰] Benefits and Harms of Large Language Models in Digital Mental Health

Munmun De Choudhury, Sachin R. Pendse외 1명|arXiv (Cornell University)|2023. 11. 07.

Digital Mental Health Interventions인용 수 8

한 줄 요약

본 논문은 디지털 정신 건강에서의 대형 언어 모델의 기회와 위험을 생태계 프레임워크를 통해 분석하며, care-seeking, community care, institutional care, 및 societal care ecologies를 가로지르는 생태계 맥락에서 다룬다.

ABSTRACT

The past decade has been transformative for mental health research and practice. The ability to harness large repositories of data, whether from electronic health records (EHR), mobile devices, or social media, has revealed a potential for valuable insights into patient experiences, promising early, proactive interventions, as well as personalized treatment plans. Recent developments in generative artificial intelligence, particularly large language models (LLMs), show promise in leading digital mental health to uncharted territory. Patients are arriving at doctors' appointments with information sourced from chatbots, state-of-the-art LLMs are being incorporated in medical software and EHR systems, and chatbots from an ever-increasing number of startups promise to serve as AI companions, friends, and partners. This article presents contemporary perspectives on the opportunities and risks posed by LLMs in the design, development, and implementation of digital mental health tools. We adopt an ecological framework and draw on the affordances offered by LLMs to discuss four application areas -- care-seeking behaviors from individuals in need of care, community care provision, institutional and medical care provision, and larger care ecologies at the societal level. We engage in a thoughtful consideration of whether and how LLM-based technologies could or should be employed for enhancing mental health. The benefits and harms our article surfaces could serve to help shape future research, advocacy, and regulatory efforts focused on creating more responsible, user-friendly, equitable, and secure LLM-based tools for mental health treatment and intervention.

연구 동기 및 목표

• Introduce an ecological framework to assess LLMs in digital mental health.
• Identify affordances of LLMs relevant to care at multiple ecological levels.
• Discuss potential benefits and harms in four application areas: careseekers, crisis response, clinical decision support, and telehealth futures.
• Highlight ethical, safety, and equity considerations to shape responsible deployment of LLMs.

제안 방법

• Adopts an ecological, social-ecological framework based on Insel’s four application contexts and the four-tier public health model.
• Maps LLM affordances to the four ecological levels to organize benefits and harms.
• Synthesizes evidence and examples from the literature on LLMs in psychotherapy, crisis response, and clinical decision support.
• Discusses real-world incidents and regulatory considerations to inform responsible design and use.

Figure 1: An ecological conceptualization of the use of large language models in digital mental health, based on the Social Ecological Model [ 34 ] .

실험 결과

연구 질문

• RQ1 What are the potential benefits and harms of LLMs across individual, caregiver, institutional, and societal levels in digital mental health?
• RQ2 How can affordances of LLMs be leveraged or constrained to improve care while mitigating risks in tele-mental health, crisis response, clinical decision support, and psychotherapy?
• RQ3 What ethical, safety, and equity concerns arise with LLM-enabled mental health tools, and how might regulation or guidelines address them?

주요 결과

• LLMs offer improved reach and accessibility for psychotherapy and self-guided interventions.
• LLMs enable personalization and data-driven insights for care, while raising concerns about therapeutic alliance and data privacy.
• Crisis-response applications can benefit from rapid assessment and culturally tailored guidance, but risk inappropriate outputs and data misuse.
• Clinical decision support can synthesize vast medical knowledge and support differential diagnosis, but may propagate misinformation or overlook nuanced mental-health contexts.
• There are significant ethical, legal, and equity considerations in automated decision-making and data governance that require careful governance and safeguards.