arXiv subjects

Jonah Meyerhoff

Publications and source records attributed to Jonah Meyerhoff.

6 recordsLinked to original sources

arXiv2026

Beyond Simulations: What 20,000 Real Conversations Reveal About Mental Health AI Safety

Mental-health AI safety is typically evaluated with small, simulation-based benchmarks that may not reflect the linguistic and contextual diversity of deployment. We pair four benchmark replications with an ecological audit of real-world conversations to evaluate a purpose-built mental-health AI alongside six frontier general-purpose models spanning four families (OpenAI GPT-5, GPT-5.1, GPT-5.2; DeepSeek V3; Google Gemini 3 Flash; Moonshot Kimi K2). The purpose-built system produced significantly lower overall potentially harmful content rates than every frontier comparator on suicide/self-harm, eating-disorder, and substance-use prompts (CCDH Benchmark: Ash 6.2% vs frontier models 18.0-52.0%, all p < .001). In an audit of 20,000 deployment conversations, clinician review within the audit pipeline confirmed no suicide-risk conversations lacking crisis resources and three NSSI-related conversations without crisis intervention, a within-pipeline conditional rate of 3/800 (0.38%). To extend population-level inference, we adjudicated 600 conversations randomly sampled from the full deployment distribution via blinded clinician review. Against these labels, the LLM judge showed 100% sensitivity (6/6; 95% CI 54.1-100%) and 99.2% specificity (589/594) at its operating threshold, and the conversational model delivered crisis resources in all 6 clinician-confirmed cases (zero end-to-end false negatives; rule-of-three upper 95% CI 0.50%). Findings argue for ecological auditing as a complement to pre-deployment evaluation in mental-health AI safety

cs.CY↗

arXiv2025

The Agony of Opacity: Foundations for Reflective Interpretability in AI-Mediated Mental Health Support

Throughout history, a prevailing paradigm in mental healthcare has been one in which distressed people may receive treatment with little understanding around how their experience is perceived by their care provider, and in turn, the decisions made by their provider around how treatment will progress. Paralleling this offline model of care, people who seek mental health support from artificial intelligence (AI)-based chatbots are similarly provided little context for how their expressions of distress are processed by the model, and subsequently, any reasoning or theoretical grounding that may underlie model responses. People in severe distress who turn to AI chatbots for support thus find themselves caught between black boxes, contending with unique forms of agony that arise from these intersecting opacities. In this paper, we argue that the distinct psychological state of individuals experiencing severe mental distress uniquely necessitates a higher standard of end-user interpretability in comparison to general AI chatbot use. We propose a reflective interpretability approach to AI-mediated mental health support, which nudges users to engage in an agency-preserving and iterative process of reflection and interpretation of model outputs, towards creating meaning from interactions (rather than accepting outputs as directive instructions). Drawing on interpretability practices from four mental health fields (psychotherapy, crisis intervention, psychiatry, and care authorization), we describe concrete design approaches for reflective interpretability in AI-mediated mental health support, including role induction, prosocial advance directives, intervention titration, and well-defined mechanisms for recourse, alongside a discussion of potential risks and mitigation measures.

cs.HC↗

arXiv2025

When Testing AI Tests Us: Safeguarding Mental Health on the Digital Frontlines

Red-teaming is a core part of the infrastructure that ensures that AI models do not produce harmful content. Unlike past technologies, the black box nature of generative AI systems necessitates a uniquely interactional mode of testing, one in which individuals on red teams actively interact with the system, leveraging natural language to simulate malicious actors and solicit harmful outputs. This interactional labor done by red teams can result in mental health harms that are uniquely tied to the adversarial engagement strategies necessary to effectively red team. The importance of ensuring that generative AI models do not propagate societal or individual harm is widely recognized -- one less visible foundation of end-to-end AI safety is also the protection of the mental health and wellbeing of those who work to keep model outputs safe. In this paper, we argue that the unmet mental health needs of AI red-teamers is a critical workplace safety concern. Through analyzing the unique mental health impacts associated with the labor done by red teams, we propose potential individual and organizational strategies that could be used to meet these needs, and safeguard the mental health of red-teamers. We develop our proposed strategies through drawing parallels between common red-teaming practices and interactional labor common to other professions (including actors, mental health professionals, conflict photographers, and content moderators), describing how individuals and organizations within these professional spaces safeguard their mental health given similar psychological demands. Drawing on these protective practices, we describe how safeguards could be adapted for the distinct mental health challenges experienced by red teaming organizations as they mitigate emerging technological risks on the new digital frontlines.

cs.CY↗

arXiv2025

Investigating the Role of Situational Disruptors in Engagement with Digital Mental Health Tools

Challenges in engagement with digital mental health (DMH) tools are commonly addressed through technical enhancements and algorithmic interventions. This paper shifts the focus towards the role of users' broader social context as a significant factor in engagement. Through an eight-week text messaging program aimed at enhancing psychological wellbeing, we recruited 20 participants to help us identify situational engagement disruptors (SEDs), including personal responsibilities, professional obligations, and unexpected health issues. In follow-up design workshops with 25 participants, we explored potential solutions that address such SEDs: prioritizing self-care through structured goal-setting, alternative framings for disengagement, and utilization of external resources. Our findings challenge conventional perspectives on engagement and offer actionable design implications for future DMH tools.

cs.HC↗

arXiv2024

Perfectly to a Tee: Understanding User Perceptions of Personalized LLM-Enhanced Narrative Interventions

Stories about overcoming personal struggles can effectively illustrate the application of psychological theories in real life, yet they may fail to resonate with individuals' experiences. In this work, we employ large language models (LLMs) to create tailored narratives that acknowledge and address unique challenging thoughts and situations faced by individuals. Our study, involving 346 young adults across two settings, demonstrates that personalized LLM-enhanced stories were perceived to be better than human-written ones in conveying key takeaways, promoting reflection, and reducing belief in negative thoughts. These stories were not only seen as more relatable but also similarly authentic to human-written ones, highlighting the potential of LLMs in helping young adults manage their struggles. The findings of this work provide crucial design considerations for future narrative-based digital mental health interventions, such as the need to maintain relatability without veering into implausibility and refining the wording and tone of AI-enhanced content.

cs.HC↗

arXiv2023

Using Adaptive Bandit Experiments to Increase and Investigate Engagement in Mental Health

Digital mental health (DMH) interventions, such as text-message-based lessons and activities, offer immense potential for accessible mental health support. While these interventions can be effective, real-world experimental testing can further enhance their design and impact. Adaptive experimentation, utilizing algorithms like Thompson Sampling for (contextual) multi-armed bandit (MAB) problems, can lead to continuous improvement and personalization. However, it remains unclear when these algorithms can simultaneously increase user experience rewards and facilitate appropriate data collection for social-behavioral scientists to analyze with sufficient statistical confidence. Although a growing body of research addresses the practical and statistical aspects of MAB and other adaptive algorithms, further exploration is needed to assess their impact across diverse real-world contexts. This paper presents a software system developed over two years that allows text-messaging intervention components to be adapted using bandit and other algorithms while collecting data for side-by-side comparison with traditional uniform random non-adaptive experiments. We evaluate the system by deploying a text-message-based DMH intervention to 1100 users, recruited through a large mental health non-profit organization, and share the path forward for deploying this system at scale. This system not only enables applications in mental health but could also serve as a model testbed for adaptive experimentation algorithms in other domains.

cs.AI↗