Research & Impact Orientation
The MAANAS Intelligence Kernel is the research platform we are building. Its web, iOS, and watchOS applications are initial beachheads for exploring hypotheses and developing an evaluation approach. Clinical effectiveness has not been established.
Current development status
- Implemented demonstrations: Interactive desktop, mobile, and watch prototypes with sample values and scripted interactions. They demonstrate interface behavior, not health measurement.
- Experimental research: Multimodal signal interpretation, WLOP fusion, and optional wellbeing prompts. Clinical accuracy and benefit remain unestablished.
- Proposed evaluation: Clinical studies, comparison with appropriate reference assessments, subgroup analysis, and safety monitoring. No completed clinical validation is claimed.
- Privacy research: Local processing and federated learning are approaches to evaluate. Neither establishes anonymity or product-wide compliance.
The Research Motivation
Our motivation is to explore additional ways for people to recognize changes in wellbeing and seek professional support. We have not demonstrated that MANAS can identify a clinical condition, reduce hospitalizations, or prevent crises.
Our Research Question
Can patterns in language, voice, facial movement, and physiological signals help people recognize meaningful changes in wellbeing and access support sooner? Studies must test this hypothesis, including the possibility that the approach is inaccurate or provides no benefit.
Proposed Clinical Evaluation
The following is a proposed evaluation plan, not a report of completed or enrolling clinical trials. Study design, partners, appropriate ethics and regulatory review, and participant safeguards need to be established before enrollment:
- Establishing baselines: Define the intended population, reference assessments, and thresholds in advance. Validated questionnaires may inform evaluation; their validity does not transfer to MANAS.
- Outcome definition: Evaluate accuracy, false alarms, missed concerns, time to professional assessment, and potential harms. Claims about improved outcomes require suitable comparison groups and follow-up.
- Algorithmic fairness audits: Measure performance and uncertainty across demographic, geographic, and linguistic groups rather than assuming equivalent accuracy.
MAANAS Intelligence Kernel — Technical Architecture
A detailed engineering account of the multimodal fusion pipeline — facial action units, remote photoplethysmography, voice acoustics, linguistic biomarkers, and the WLOP probabilistic fusion engine. Describes technical hypotheses and proposed evaluation. Architecture and mathematical explanations do not establish clinical efficacy or legal compliance.
Join Our Research Coalition
We actively seek partnerships with academic medical centers, psychiatry departments, and public health researchers.
Request Research Briefing