When AI Listens First: Crisis Triage and Emotional Support in Mental Health Helplines
Publication Date
2-20-2026
Abstract
Problem definition: Mental health helplines play a critical role in providing timely support to individuals experiencing acute distress, yet they face severe capacity constraints and high call abandonment. Generative artificial intelligence (AI), particularly large language models (LLMs), has emerged as a potential first point of contact to support triage and engagement. In this setting, AI serves not only as a screening tool but also as a source of emotional support. We study when such AI systems improve helpline operations and how they should be designed and tuned to enhance outcomes. Methodology/results: We develop a stylized queueing model of a mental health helpline in which calls first interact with a generative AI system and may be escalated to human counselors. The model captures the bifunctional role of AI in risk identification and emotional support, while accounting for waiting and abandonment. Our analysis shows that AI deployment can have nonmonotonic effects on outcomes: improvements in triage accuracy or conversational support do not necessarily improve system performance once congestion and abandonment are considered. The optimal LLM effort allocation exhibits a corner solution structure under mild conditions: effort should be concentrated entirely on triage or emotional support, depending on their relative effectiveness and the degree of functional synergy between them. Managerial implication: Our findings offer insights into operational conditions under which it is beneficial to introduce AI in mental health helplines and identify concrete tuning targets for LLM design that improve system performance while balancing safety, access, and speed. They also highlight the key operational and behavioral parameters that helplines must calibrate in advance, enabling informed LLM development and deployment and reducing the risk of unintended access losses.
Document Type
Article
Keywords
Mental Health Helplines, Large Language Models, Triage, Call Support
Disciplines
Business Administration, Management, and Operations
Source
SMU Cox: IT & Operations Management (Topic)
Language
English
