Artificial intelligence in traumatic stress treatment: The TRUST framework for ethical development, clinical...

Posttraumatic stress disorder (PTSD) and depression are common diagnoses following traumatic events, with several available evidence-based interventions to reduce symptomology. However, trauma populations face significant access barriers that limit their adoption and reach. Artificial intelligence (AI) technologies, such as large language models (LLMs), have the potential to enhance access, cost-efficiency, delivery, and quality of traumatic stress interventions. Their application to traumatic stress treatment and research is understudied and requires responsible and ethical development, evaluation, and monitoring to maintain service quality and delivery for trauma survivors and their providers. We present considerations for the responsible use of AI tools to ethically shape trauma-focused treatment and research, with future use cases to highlight these critical considerations. A multidisciplinary approach to trauma-based LLM development that integrates feedback and evaluation from trauma experts, clinicians, and trauma survivors and prioritizes the quality, safety, effectiveness, and equity of trauma-focused care is essential. We propose a framework, TRUST, that is informed by evidence from adjacent mental health AI applications and emerging research on digital trauma interventions, while acknowledging the limited trauma-specific AI trial data. These topics were presented at the 41st Annual Meeting of the International Society for Traumatic Stress Studies via a panel of experts in AI technologies to support trauma-focused interventions.