The MAANAS Intelligence Kernel
We are building a research platform, with the web application, iOS app, and watchOS apps as its initial beachheads. The concepts below describe capabilities we are exploring through the MAANAS Intelligence Kernel; their clinical benefits remain unestablished.
Research: Multimodal Signals
We are exploring whether combinations of language, voice, facial movement, and physiological signals correlate with meaningful changes in wellbeing. Their ability to identify distress accurately or earlier than existing assessments has not been established.
Sample prompt: Would you like to take a short break or reach out to someone you trust?
Research: Optional Wellbeing Prompts
Our prototypes illustrate optional wellbeing prompts. Whether personalization is helpful, appropriate, and safe for people experiencing distress requires evaluation; these examples are not treatment recommendations.
Research: Culturally Informed Design
We aim to study how language and culture affect the interpretation of signals and prompts. Cultural appropriateness and equivalent performance across groups have not been established.
Human-Centered Design
Future evaluation should include clinical expertise, ethics review, and people with relevant lived experience.
What the research needs to establish
Combining signals
Does combining modalities improve performance over a single signal or self-report? We have not established clinical superiority.
Errors and uncertainty
How often does MANAS miss concerns or raise false alarms? Combining signals does not eliminate false positives.
Useful support
Do optional prompts improve access to support or wellbeing compared with an appropriate control?
Performance across groups
How does performance vary by language, culture, demographics, device, and context?