AI 'Moonshot' Taps Smartphones to Predict Mental Health Crises

EUGENE, OR – May 14, 2026 – A groundbreaking initiative backed by a federal "moonshot" agency aims to turn the smartphone in your pocket into a powerful tool for predicting and managing mental health crises. Digital health firm Ksana Health has been awarded a contract worth up to $17.9 million from the Advanced Research Projects Agency for Health (ARPA-H) to lead a multi-institutional effort to build a new class of artificial intelligence for behavioral health.

The project, titled "Building a Behavioral Health Foundation Model," will develop a Large Health Behavior Model (LHBM). This AI will be trained to analyze continuous, real-world data passively collected from personal smartphones—such as sleep patterns, physical activity, and social engagement—and link it to electronic health records. The goal is to create a system that can detect the subtle behavioral shifts that often precede a mental health or substance use crisis, enabling earlier, more personalized, and proactive interventions.

A New Paradigm for Behavioral Health

The current approach to mental healthcare is often reactive and fragmented. Behavioral factors are linked to an estimated 90 percent of the nation's staggering $4.9 trillion in annual health expenditures, yet the system largely waits for patients to report symptoms, which are then assessed subjectively. This leads to a time-consuming process of trial and error for treatments, with clinicians often struggling to predict patient outcomes.

"Behavioral factors contribute to 90 percent of the nation's $4.9 trillion in annual health expenditures, yet we still rely on fragmented, episodic approaches to detection and intervention," said Nick Allen, PhD, CEO of Ksana Health and the project's Principal Investigator. "This project will harness the sensing capabilities of people's personal smartphones and link those behavioral signals to health records at scale. The goal is to build a foundation model that can enable a new era of proactive, personalized behavioral healthcare."

This initiative is funded under ARPA-H's EVIDENT program, which is designed to foster breakthroughs in measuring and treating behavioral health conditions. ARPA-H itself was established to fund high-risk, high-reward research with the potential to transform health outcomes, and this project fits that mold perfectly. By aiming to create objective, data-driven metrics for behavioral health, the project could provide the kind of evidence-based tools that have long been missing from the field.

The Technology of Transformation

At the heart of the project is the concept of digital phenotyping—using data from personal digital devices to create a detailed, moment-by-moment picture of an individual's behavior. The LHBM will be trained on vast datasets capturing patterns of physical activity, sleep duration and quality, mobility via GPS, social connection through call and text logs (analyzing metadata, not content), and even language use patterns.

This effort builds on three converging technological trends: the ubiquity of smartphones, the digitization of medical records, and recent breakthroughs in AI foundation models that can process diverse, multi-modal data. Ksana Health's EARS (Effortless Assessment Research System) platform, a leading infrastructure for mobile sensing in research, will serve as the primary data collection engine.

The market for this technology is expanding rapidly. The digital phenotyping sector, valued at over $1 billion in 2024, is projected to surge past $8 billion by 2033. While numerous companies are entering the behavioral health AI space with solutions like chatbots or single-purpose apps, Ksana's approach is more ambitious. By building a comprehensive foundation model that integrates clinical and real-world data at an unprecedented scale, the project aims to create a foundational resource that could power a wide array of future applications, from clinical decision support to personalized wellness nudges.

Building Trust with an Ethical Blueprint

The prospect of an AI continuously monitoring personal behavior for signs of mental distress raises immediate and significant ethical questions about privacy, bias, and consent. The project's leaders are addressing this head-on by embedding an ethical framework into the project from its inception.

A central component is the Lived Experience and Ethics Panel (LEEP), a group composed of individuals with personal experience of mental health challenges, alongside experts in AI, bioethics, and healthcare. This panel is not a mere advisory board; it is an integral part of the governance structure. The LEEP will be tasked with evaluating model development, assessing training data for potential biases that could harm marginalized communities, and co-developing the guidelines for the technology's safe and effective use.

"We believe that the people most affected by behavioral health conditions must have a meaningful voice in shaping the tools designed to help them," Allen stated. "The LEEP panel is not an afterthought—it is built into the project's structure from day one."

This proactive approach aligns with emerging best practices for AI in healthcare, which demand strict adherence to privacy laws like HIPAA, transparent model performance, and robust human oversight. By placing ethics and patient advocacy at its core, the initiative hopes to build the trust necessary for such a powerful technology to be accepted and adopted responsibly.

A Coalition for National Impact

Tackling a problem as vast as the nation's behavioral health crisis requires a collaborative effort. The project brings together a powerful coalition of partners, each contributing unique expertise.

Major healthcare systems Providence Health & Services and MedStar Health will lead participant recruitment, ensuring the data collected is geographically and demographically diverse, which is crucial for building an equitable AI model. Their involvement also provides a direct pathway for translating the research into real-world clinical practice. Providence, in particular, is known for its Digital Innovation Group and a track record of developing and commercializing new health technologies.

The core computational modeling will be spearheaded by scientists at the University of Washington's prestigious Paul G. Allen School of Computer Science & Engineering. Their deep expertise in behavioral data science and foundation model architecture will provide the academic and technical rigor needed to build the sophisticated AI.

The project will unfold in phases, starting with a proof-of-concept and scaling to tens of thousands of participants. The ultimate goal is to make the resulting model, with appropriate safeguards developed in consultation with the LEEP, available to the broader scientific and clinical community.

"This initiative augments Ksana's current efforts to shift behavioral healthcare from episodic, subjective assessment toward continuous, data-driven health promotion, reducing healthcare spending, improving quality of life, and reaching populations that currently lack access to effective behavioral health support," said Tony Scripa, COO of Ksana Health.