zlacker

[parent] [thread] 11 comments
1. PaulHo+(OP)[view] [source] 2023-11-02 15:53:55
One of those things y'all never upvote is papers about psychodiagnostic software that uses your social media posts, cell phone location data, etc. to diagnose both chronic and acute psychiatric conditions.

I'm fairly certain that if a person is highly active on social media such a system could produce a better diagnosis than most people get when they see a professional, if only because the quality of psychodiagnosis is poor since it is often seen as a scam to satisfy insurance bureaucrats, common conditions are never diagnosed, there are fads for certain rare conditions, etc.

replies(6): >>teachr+e1 >>hutzli+15 >>spinni+ea >>phone8+Tc >>taway1+Pd >>r00fus+8e
2. teachr+e1[view] [source] 2023-11-02 15:57:25
>>PaulHo+(OP)
I don't mean to sound flip, but I don't think identifying pathological psychological conditions via web browsing habits is all that difficult. I have a friend who went through a severe depressive episode. As soon as he started watching YouTube at 2am, he started getting ads for depression meds.
replies(2): >>PaulHo+I4 >>aaroni+df
◧◩
3. PaulHo+I4[view] [source] [discussion] 2023-11-02 16:08:21
>>teachr+e1
It's not difficult at all. That's the point. A system like that collects a lot of data and very few people are going to feel that they need to dissimulate.
4. hutzli+15[view] [source] 2023-11-02 16:09:22
>>PaulHo+(OP)
"if a person is highly active on social media "

.. then the diagnosis of one of their problems sounds quite trivial.

5. spinni+ea[view] [source] 2023-11-02 16:25:57
>>PaulHo+(OP)
I think you've neatly summarised why it doesn't get upvoted:

1. "could" produce a better diagnosis. Not guaranteed. And better than what? How likely is it to really deliver a better result than appropriately trained specialists? 2. "scam to satisfy insurance bureaucrats". And you doing it digitally won't find its way to unintended recipients?

The undercurrent of this thread - and the original post - is growing awareness of the dystopian disaster that has grown out of "free" social media. So it's not surprising - to me, at any rate - that the general sentiment here is to be suspicious of any adjacent use.

6. phone8+Tc[view] [source] 2023-11-02 16:34:36
>>PaulHo+(OP)
> I'm fairly certain that if a person is highly active on social media such a system could produce a better diagnosis than most people get when they see a professional

What makes you so sure? (This is a serious question, not rhetorical.)

replies(1): >>PaulHo+Jg
7. taway1+Pd[view] [source] 2023-11-02 16:37:39
>>PaulHo+(OP)
That sounds nice! Do you know of any offline linux-compatible software that could pull this off? I'm happy to try it on myself.

What I don't understand though is why do I also need to share my browsing history with faceless american corps that sell my data for profit. This sounds unnecessary for the main point (psychodiagnostic software).

8. r00fus+8e[view] [source] 2023-11-02 16:38:17
>>PaulHo+(OP)
Well, if you've ever watched a movie like Real Genius [1], you may begin to understand some of the concerns of people building things with advanced capabilities.

Imagine how your tech could be used for evil and how profitable that would be. It could be a 2nd or even 3rd order effect, even.

[1] Film focuses on a college team building something they think is cool but really is a key part of a weapons system.

◧◩
9. aaroni+df[view] [source] [discussion] 2023-11-02 16:42:17
>>teachr+e1
Worse yet, "web browsing habits" is just one neck of the hydra. What you buy (when, as mentioned) serves as strong signal for any number of factors; all your conventional demographics of course...

I'm not in the industry but I am very curious to know if we're already in the conditional-execution phase of surveillance/ad-serving/profile-updating: is there an idea [yet] of serving a challenge, and then both recording how/if it is engaged, with automated graph traversal to "look closer"... all offered stochastically...

The simple way to put that in part is, are we now getting A/B tests run on us explicitly, rather than merely implicitly?

(Personally, I'm 100% off Meta products and TikTok—but am leaking through LinkedIn and, regrettably, Google...)

◧◩
10. PaulHo+Jg[view] [source] [discussion] 2023-11-02 16:48:02
>>phone8+Tc
Mostly because the quality of professional diagnosis is poor.

Myself I have a condition which 5-10% of people have. As a child, I had two very high quality psych evaluations for the time where people observed all the signs and symptoms (particularly the first one) but failed to draw a line between them.

Since then I saw therapists maybe 6 times in 30 years (sometimes the same one) and it was always “adjustment disorder with …” and there was some truth in that in that in each case I had some very ordinary kind of stress which was exacerbating my condition but in reality there was always a chronic aspect to that.

I’ve known numerous people who have severe mental illness (way worse than the quirk that got me kicked out of elementary school) and contact with the psychiatric system and never got a conclusive diagnosis. The first line for a lot of people is to see a primary care practitioner and get diagnosed with either “anxiety” or “depression” and prescribe the same medication in either case. A referral to an actual psychiatric nurse practitioner who is taking patients is almost impossible in 2023 in the US never mind an actual psychiatrist.

replies(1): >>diggin+DE
◧◩◪
11. diggin+DE[view] [source] [discussion] 2023-11-02 18:18:54
>>PaulHo+Jg
That's not exactly strong evidence that "quality of professional diagnosis is poor" though... it's just evidence that quality was poor in cases you're aware of. It's also not evidence at all that being tracked by facebook would have come up with better results sooner.
replies(1): >>PaulHo+221
◧◩◪◨
12. PaulHo+221[view] [source] [discussion] 2023-11-02 20:04:54
>>diggin+DE
Back when the problem was too much psychiatric care instead of not enough there was this famous experiment

https://en.wikipedia.org/wiki/Rosenhan_experiment

This one is more positive but is checking that different diagnosticians get the same answer

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5980511/

and if that was applied to the "Thud" experiment you'd have poor diagnosis with a very high kappa (interrater agreement)

[go to top]