Show HN: An API/SDK to passively detect user mental health status

[aleksdahlberg]

It's early days yet, but right now it can detect similarities in certain behaviors (steps, sleep, phone usage) associated with depression (trained on 37.5K PHQ9 samples from 3,500+ people) with up to 84% accuracy.

We've built some SDKs that make integration and data collection simple on mobile, currently supporting: Native Apple & Android, Flutter, React-Native and Capacitor. Everything runs in the background, even when an integrated app is closed.

The SDKs also provide the necessary native user-permission handling so users understand what data is being shared with integrated software, this also gives the user control over sharing such data. The data is pulled from Apple HealthKit and Google APIs (for Android).

There is also the REST API for those who want to customize things completely.

Soon to be released: passive anxiety, stress and emotion predictions using such data.

I'd love some feedback, please share!

[anigbrowl]

It's good tht you've proactively addressed privacy concerns, and your efforts here seem substantive rather than cosmetic. But I can't say I feel enthused about the video of invisible mental health diagnoses and automated interventions; this is why I have strict limits for myself on device interaction, and prefer a goofy old analog wristwatch to a digital fitness tracker. Likewise I block out some times/spaces/activities as being device free.

It seems to me that the costs of automatic this sort of cognitive/emotional effort may outweigh the potential benefits - in the aggregate, it may of course be ideal for some individuals. These are of course concerns about the undertaking rather than your particular product.

[aleksdahlberg]

Thanks for the feedback and for taking a read of things. I wouldn't treat this as a diagnosis (it's not) more of an indication as it's trained on PHQ9 data - which is not a full diagnostic tool. However, clinicians do use this tool to help assess and decide on appropriate interventions or actions to take.

[anigbrowl]

I'll certainly watch the product with interest to see where it goes. The potential benefits could be huge. It's primarily the larger economic incentives that I worry about, which are factors beyond any one researcher or firm's control.

[Nextgrid]

This is creepy as fuck - I don't see why I would ever want any local app, let alone a third-party, taking guesses at my mental state. What sort of use-cases did you build this for?

[aleksdahlberg]

I agree (though I wouldn't use the words "creepy as fuck" but I get your point), I think it's super important that this technology is available only upon permission by the end user - this is what I've done with Sahha to the best of our ability - relying on native permission handling has been super beneficial.

As for use-cases, there are plenty - and some not as direct as others but particularly within healthcare and life sciences there are plenty of uses around efficacy tracking and making sure interventions and treatments can be delivered appropriately based on objective assessment. The latter use leads into the problem many companies delivering health care interventions have with survey fatigue and subjective data bias.