We're inviting researchers to collaborate on what actually works in AI for mental health. Express interest here.

Caitlin A. Stamatis, PhD & Thomas D. Hull, PhD

The gap we see

AI tools for wellbeing are scaling faster than the evidence base can keep up. Millions of people are interacting with conversational AI in the context of their mental health, and the field urgently needs rigorous, independent research to understand what's working, for whom, and why.

Academic researchers often have the right methods and the right questions, but limited access to the systems generating the data. We have a production platform and a front-row seat to how people actually engage with AI for their mental health — but we can't pursue every important research direction ourselves.

This program is designed to close that gap.

How we support research

The scope of each collaboration is determined on a case-by-case basis, tailored to the needs of the study and the nature of the research question. Depending on the project, support may include:

• Collaboration on study design — Working together to define research questions, cohorts, and methodologies that our environment can meaningfully support

• Aggregate or de-identified data — Where appropriate, we can provide structured data with IRB-compatible documentation and privacy safeguards

• Compute and tooling — API access, analytics infrastructure, or support for NLP and language modeling workflows as needed

• Publication support — We support open-access publication and collaboration on conference papers and presentations

A pathway, not just a single study. Promising collaborations can serve as the foundation for deeper research partnerships over time — building from early-stage exploration through to prospective study designs.

Research areas of interest

We're drawn to work that connects interaction-level data to meaningful clinical and behavioral questions. Here are some areas we find compelling — but this list is far from exhaustive. If your research interests intersect with conversational AI in a behavioral health context, we want to hear from you regardless of whether it fits neatly into a category below.

Clinical populations and outcomes

• Chronic health conditions — The intersection of behavioral health and chronic illness, particularly in populations managing diabetes, cardiovascular disease, and other conditions where psychological support impacts adherence and outcomes

• Sleep — How conversational interventions may be helpful for sleep disturbance, and whether usage patterns or language markers track with sleep-related outcomes

• Social functioning — Changes in social engagement, isolation, and interpersonal functioning as reflected in or influenced by platform interactions

• Perinatal mental health — Engagement patterns, symptom trajectories, and intervention responsiveness in pregnant and postpartum populations — a group that is both high-need and historically underserved in digital health research

• Anxiety disorders — How conversational AI supports individuals managing generalized anxiety, panic, social anxiety, and related conditions, including real-world utilization patterns and treatment response

Language, behavior, and mechanism

• Leading indicators of psychological change — Can language data reveal early signals of clinical improvement or deterioration before they show up on standard measures? What linguistic, semantic, or behavioral markers are most predictive?

• Model behaviors that matter — Which AI conversational behaviors promote both sustained engagement and positive clinical outcomes — and where do those two objectives diverge?

• Usage and engagement patterns — What real-world engagement trajectories look like, how they relate to outcomes, and what distinguishes productive use from disengagement (or overengagement)

Health economics and evidence generation

• HEOR studies — Cost-effectiveness, resource utilization, treatment patterns, and value demonstrations leveraging real-world data to inform payer and policy decisions

• Real-world evidence methodology — Novel approaches to generating rigorous evidence from observational digital health data

This is a starting point. We welcome interest from any discipline — clinical psychology, computer science, HCI, psychiatry, computational linguistics, health economics, behavioral medicine, implementation science, and beyond.

Interested? It starts with a 5-minute form.

Tell us what you're curious about and what you'd need. We review submissions on a rolling basis and will follow up within three weeks to schedule a call and explore fit. No formal proposal required upfront.

[Express Interest →]

Why we're doing this

We believe the most important questions about AI in mental healthcare won't be answered by the companies building the tools. They'll be answered by independent researchers with the training and incentives to ask hard questions and follow the data wherever it goes.

We'd rather build a product informed by that kind of evidence than one informed only by our own internal metrics. If you're working on questions we can help answer, we want to make it easy for you to do that work.

Questions before submitting? Reach out to research@slingshotai.com — happy to discuss feasibility or scope before you fill out the form.