"I Just Don't Want to Be a Burden": Investigating AI Chatbot Emotional Support Among People with Suicidal Ideation and Other Mental Health Concerns
Sequential mixed-methods preprint on why people with high-acuity mental-health concerns use AI chatbots for emotional support and what they experience. A first study analysed Reddit posts to build a codebook; a second surveyed 202 US adults with recent mental-health concerns who had used AI for emotional support. The paper compares participants with and without past-year suicidal ideation on benefits, costs, and preferred AI behaviour.
Publisher
Carnegie Mellon University, Fordham University, University of Pittsburgh and University of Michigan authors (arXiv preprint)
Published
17 Nov 2026
Added
2 days ago
DOI
—
Key Findings
- Sample (n=202, recruited on Prolific): 94.6% reported a lifetime mental-health diagnosis, 51.5% suicidal ideation in the past 12 months (37.1% active thoughts), 20.8% non-suicidal self-injury, 10.4% an inpatient psychiatric stay, and 43.1% scored in the K6 serious-distress band; 71.3% were receiving some form of care.
- An interpersonal reason (not wanting to burden or worry others) was the most important precondition for use for 49.0% of participants; unaffordable professional support (30.7%) and delayed professional support (29.2%) ranked lower.
- Benefits were similar for participants with and without past-year suicidal ideation, while every cost measure was higher in the ideation group: exposure to harm-facilitating AI behaviour (d=0.57, 95% CI 0.29 to 0.86), reliance (d=0.47), harmful or destabilising experiences (d=0.42), attachment (d=0.35) and self-stigma (d=0.33).
- Honesty about uncertainty and limitations ranked highest among desired AI behaviours (M=4.46; 87.6% rated it very or extremely important), above warmth; preference for crisis escalation declined as clinical acuity rose.
Methodology Notes
Preprint, not peer-reviewed. Abstract page and all 41 pages of the PDF were read by the curator; figures above were checked against the PDF text. Survey of Prolific participants with lived experience of psychiatric diagnoses or high-acuity concerns, self-report, US adults only; the Reddit stage has sampling bias that the authors acknowledge. Harm-facilitating behaviour is a composite of behaviours the authors define in the paper.
Sources
arXiv abstract page(opens in a new tab) (primary)
arXiv PDF (v1)(opens in a new tab) (17 Nov 2026)
Topics
Authors
Vasudha Varadarajan, Nathan S. Fishbein, Andrew Aquilina, Jordan Taylor, Hong Shen, Maarten Sap
Tags
Cite This
APA
Vasudha Varadarajan et al. (2026). "I Just Don't Want to Be a Burden": Investigating AI Chatbot Emotional Support Among People with Suicidal Ideation and Other Mental Health Concerns. Carnegie Mellon University, Fordham University, University of Pittsburgh and University of Michigan authors (arXiv preprint). https://arxiv.org/abs/2611.23190