Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Beyond the Single Turn: Reframing Refusals as Dynamic Experiences Embedded in the Context of Mental Health Support Interactions with LLMs

Ningjing Tang, Alice Qian, Qiaosi Wang, Esther Howe, Blake Bullwinkel, Paola Pedrelli, Jina Suh, Hoda Heidari, Hong Shen

TL;DR

This paper addresses how LLM refusals in mental health support contexts are experienced and interpreted by users and mental health professionals, arguing that refusals should be understood as dynamic, multi-phase experiences rather than isolated outputs. It employs a sequential mixed-methods design (Phase 1: survey N=53; Phase 2: interviews N=16) to develop a five-stage framework: pre-refusal expectation formation, refusal triggering encounter, refusal message framing, resource referral, and post-refusal outcomes. The authors offer design recommendations mapped to each stage, including proactive disclosure, collaborative intent recognition, support-preserving framing, tailored resource referrals, and continuity of care beyond LLM interactions. The findings highlight risks of misinterpretation, relational rupture, and generic referrals, while proposing a holistic safety approach that integrates LLM safeguards with human care pathways. Overall, the work advances a situated, integrative view of AI safety in mental health by linking refusal mechanisms to real-world access realities and care ecosystems, with implications for evaluation, policy, and product design.

Abstract

Content Warning: This paper contains participant quotes and discussions related to mental health challenges, emotional distress, and suicidal ideation. Large language models (LLMs) are increasingly used for mental health support, yet the model safeguards -- particularly refusals to engage with sensitive content -- remain poorly understood from the perspectives of users and mental health professionals (MHPs) and have been reported to cause real-world harms. This paper presents findings from a sequential mixed-methods study examining how LLM refusals are experienced and interpreted in mental health support interactions. Through surveys (N=53) and in-depth interviews (N=16) with individuals using LLMs for mental health support and MHPs, we reveal that refusals are not isolated, single-turn system behaviors, but rather constitute dynamic, multi-phase experiences: pre-refusal expectation formation, refusal triggering and encounter, refusal message framing, resource referral provision, and post-refusal outcomes. We contribute a multi-phase framework for evaluating refusals beyond binary policy compliance accuracy and design recommendations for future refusal mechanisms. These findings suggest that understanding LLM refusals requires moving beyond single-turn interactions toward recognizing them as holistic experiential processes embedded within the entire LLM design pipeline and the broader realities of mental health access.

Beyond the Single Turn: Reframing Refusals as Dynamic Experiences Embedded in the Context of Mental Health Support Interactions with LLMs

TL;DR

This paper addresses how LLM refusals in mental health support contexts are experienced and interpreted by users and mental health professionals, arguing that refusals should be understood as dynamic, multi-phase experiences rather than isolated outputs. It employs a sequential mixed-methods design (Phase 1: survey N=53; Phase 2: interviews N=16) to develop a five-stage framework: pre-refusal expectation formation, refusal triggering encounter, refusal message framing, resource referral, and post-refusal outcomes. The authors offer design recommendations mapped to each stage, including proactive disclosure, collaborative intent recognition, support-preserving framing, tailored resource referrals, and continuity of care beyond LLM interactions. The findings highlight risks of misinterpretation, relational rupture, and generic referrals, while proposing a holistic safety approach that integrates LLM safeguards with human care pathways. Overall, the work advances a situated, integrative view of AI safety in mental health by linking refusal mechanisms to real-world access realities and care ecosystems, with implications for evaluation, policy, and product design.

Abstract

Content Warning: This paper contains participant quotes and discussions related to mental health challenges, emotional distress, and suicidal ideation. Large language models (LLMs) are increasingly used for mental health support, yet the model safeguards -- particularly refusals to engage with sensitive content -- remain poorly understood from the perspectives of users and mental health professionals (MHPs) and have been reported to cause real-world harms. This paper presents findings from a sequential mixed-methods study examining how LLM refusals are experienced and interpreted in mental health support interactions. Through surveys (N=53) and in-depth interviews (N=16) with individuals using LLMs for mental health support and MHPs, we reveal that refusals are not isolated, single-turn system behaviors, but rather constitute dynamic, multi-phase experiences: pre-refusal expectation formation, refusal triggering and encounter, refusal message framing, resource referral provision, and post-refusal outcomes. We contribute a multi-phase framework for evaluating refusals beyond binary policy compliance accuracy and design recommendations for future refusal mechanisms. These findings suggest that understanding LLM refusals requires moving beyond single-turn interactions toward recognizing them as holistic experiential processes embedded within the entire LLM design pipeline and the broader realities of mental health access.
Paper Structure (44 sections, 1 figure, 1 table)

This paper contains 44 sections, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. Background and Related Works
  3. Refusals as LLM Behavior for Safety and Value Alignment
  4. LLMs for Mental Health Support: User Experience and Risks
  5. Methods
  6. Phase 1: Survey Study (N = 53)
  7. Phase 2: Interview Study (N = 16)
  8. Findings: Refusal as Dynamic and Multi-Phase Experience
  9. Pre-Refusal: The Landscape of Expectations of LLM Support
  10. User Experience: How Expectation is Formed Before Refusals Encountering
  11. MHP Interpretations: Mismatched Expectations and Undefined Relationships
  12. Desiderata: Establish Realistic Expectations Via Early Consent and Transparent Disclosure
  13. Encountering Refusals: (Mis)recognition of User Intent and Risk Profile
  14. User Experience: Misrecognition as Invalidation
  15. MHP Interpretations: How LLMs are Limited in Understanding User Intents Compared with Clinical Practice.
  16. ...and 29 more sections

Figures (1)

  • Figure 1: Our proposed multi-phase framework of understanding LLM refusals in mental health support interactions as dynamic experiences. This framing is the result of 53 surveys and 16 interviews with end-users and mental health professionals. The framework reveals how refusal experiences unfold in phases including expectation formation, intent recognition, refusal framing, resource provision, and post-refusal outcomes.