cross-posted from: https://programming.dev/post/37560470
Depression is often linked to changes in facial expressions. However, the link between mild depression, known as subthreshold depression, and changes in facial expressions remains unclear. Now, researchers have investigated whether subthreshold depression shows changes in facial expressions in Japanese young adults using artificial intelligence. The findings reveal distinct muscle movement patterns related to depressive symptoms which may help detect depression early, paving the way for timely and preventative mental health care.
As suspected, this paper is based upon Paul Ekman’s Facial Action Coding System, better known as micro-expressions.
This controversial literature borderlines on pseudo-science, with Ekman’s work having significant conflicts of interest in the form of undisclosed funding from US defence, police, and border control, who wanted Ekman to create systems that can detect lies based on supposed micro expressions. Subsequent independent meta analyses have found that micro expressions cannot be consistently read, even by Ekman’s own researchers.
Unfortunately, despite this, world governments have poured millions of dollars into private contractors to develop AI systems based on this flawed research for border control.
Munecat’s video essay on debunking body language experts goes into much greater detail:
i wouldnt use AI to diagnose any illnesses, its worst than using webmd.
If you feel you are not properly sedated, call 348-844 immediately. Failure to do so may result in prosecution for criminal drug evasion.
— THX 1138
Plot twist - it was the AI that is making the students depressed in the first place.
Assumptions like these tend to go wrong for neurodivergent groups.
Do one that detects psychopathy and try it on politicians, heh.
I don’t think we need AI to know they’re all psychopaths
Isn’t it easier to just assume everyone is depressed in 2025?
A system like that can claim 95% accuracy by just saying “yes” always
Yes but if you don’t look miserable you are hiding it.
Perfect, we can now detect at what level to hold the populous so they’ll keep working, and not revolt.
I recall reading that one application of sentiment analysis in voice recognition — like, determining what a speaker’s mood is — is that if someone gets upset on a call talking to a computer, the system will route them to a human.
Brother, I think you are depressed.
Correct, please send this one to the Emotional Recalibration Camp as soon as possible.
If you’re a good little citizen, they might even teach you how to concentrate while you’re at camp!
Wait. No stop I DIDN’T MEA-
