In the spirit of Mad Studies’ mission to redefine science and empowering users, I’m sharing a vision for a non-profit smartphone app that could transform psychiatric care. Research shows AI can detect depression from facial expressions (using just 4 minutes of video) with accuracy effectively surpassing clinicians, even when only trained off data sets as small as ~200 participants. A free app could make this objective, scalable tool ubiquitous, addressing concerns about stigma, treatment efficacy, and even the definition of mental states themselves. Key features:
-
Initial capability: Detects depression via facial expressions, leveraging existing academic models.
-
Self-Improving: Optional user-provided depression scores (e.g., PHQ-9) expand the AI platform training database, boosting accuracy with each use.
-
Future expansion: Incorporate other smart-phone accessible biomarkers (audio, motion, pupillometry, saccades, etc.) with user consent. Expand to anxiety, PTSD, and more.
-
Ethical model: Run as a non-profit, user data stays protected (including contractual protection with every user) and user access remains free. Monetization through licensing algorithms to healthcare systems funds growth without compromising privacy.
The mental health crisis is profound – costs exceed all agriculture and education expenditures combined – yet diagnostic remain subjective, placing the field atop a house of cards. This app could democratize objective insight, aligning with Mad Studies’ push for user-driven, transformative approaches.
As the concept initiator, I’m confident in its potential, but lack the expertise to lead this one. I’m seeking a Principal Investigator or other project lead with associated research or fundraising interests.
For-profit apps exist, but a non-profit’s trust and scale could outpace them, especially with community-driven data contributions. However, to best incubate the idea and protect from added for-profit pressure, I’m not posting this for arbitrarily large audiences. Please DM or email me eric[at]diogenix dot org.