One worry here is that AI only classifies as well as the labeled datasets that it trains on. And my concern would be the way humans apply labels to other humans. Diagnosing mental illness is notoriously messy. DSM-IV descriptions of various illnesses including borderline personality disorder are disputed, not settled. If we can catch psychosis early, great -- but what's psychosis?
Exactly. What good does it do it the AI over-diagnoses something like (to take a relatively benign, common example) ADHD at just the same rate as human doctors? Would a nearly 75% miss rate really be acceptable? It would be a bit like 'Psycho Pass', but without any underlying competence.
The problem with Psychiatry is that it treats symptoms, not causes. The hope represented by this article is that by catching symptoms earlier, or by better pinpointing linguistic markers, vulnerable patients can be treated earlier with more-appropriate medications.
I think that it is much more important to find causes behind "mental disorders" than to describe them better.
From the article:
> According to Schwoebel, it took over 10 primary-care appointments before his brother was referred to a psychiatrist and eventually received a diagnosis. After that, he was put on one medication that didn’t work for him, and then another. In the years it took to get Schwoebel’s brother diagnosed and on an effective regimen, he experienced three psychotic breaks. For cases that call for medication, this led Schwoebel to wonder how to get a person on the right prescription, and at the right dose, faster.
What are the conditions that led this person to have a psychotic break? They didn't care, and just put Mr. Schwoebel's brother on maintenance medication for life. I suspect that such patients are usually 'exhausted' [1].
"How Psychiatry Lost Its Way: Identifying new mental disorders and their "curse" had become a productive industry; it is also bad medicine" [2], by Paul McHugh M.D., is very helpful for giving context to the mental health industry's status quo.
> The problem with Psychiatry is that it treats symptoms, not causes.
No, the problem is that we know a lot more about symptoms than causes, so we diagnose by and treat symptoms, which may or may not be treating causes (though many treatments are chosen because they treat something that there is at least a plausible hypothesis is the cause of the symptoms, and the fact that they have efficacy in treating the symptoms is at least suggestive evidence that both the hypothesized cause is correct and that it is being treated.)
This is also far less different than lots of other areas of medicine that critics of psychiatry try to make it out to be.
Robert Whitaker [1] has looked into the "mental health" industry, and proposes that today's patients do worse on their medications than their unmedicated predecessors. It used to be that people had "episodes", now they have chronic diseases from which they never recover (unless they stop taking their pills).
Thank you for this link. this seems right up my alley - I'm trying to avoid being an uninformed critic. i've had bad experiences in my life with psychiatry, been in and out of the system since i was 12 and witnessed a LOT of ... let's say bad practices. do you have any other suggested reading / sites for someone in my position?
> Thank you for this link. this seems right up my alley - I'm trying to avoid being an uninformed critic.
You're welcome, thanks for responding.
> i've had bad experiences in my life with psychiatry, been in and out of the system since i was 12 and witnessed a LOT of ... let's say bad practices.
Having seen the inside of this world (as a visitor), I can sort-of imagine what you've been through. People do have problems; the system knows not that it does not always help.
> do you have any other suggested reading / sites for someone in my position?
depends on what you're interested in. I switch to this account whenever I have something to say about mental health, so there might be some other links there. (I just surveyed my comments, there aren't too many. I post the link to Robert Whitaker's work fairly regularly, but there are a bunch of others. They depend on the context...)
https://news.ycombinator.com/item?id=11974769 - this has a link about the monopolists' approach to medicine. I posted that comment from my phone, and didn't feel like fishing out the full link. But the article is quite helpful for illuminating the source of bias in the "Medical" world.
To make an example of this... "the patient is responding well to medication/treatment for $problem, therefore $problem is the cause and/or part of a criteria that could finalize a diagnosis for $problem"
I think I see what you're saying. If $problem is kidney failure, not many people would blink an eye at this method. But something about $problem being purely "mental" has a stigma. A really nasty stigma. Yet this method can still be useful to get people help - that wasn't my personal experience, but I've seen it happen for others.
Then again, when I delve into thinking about that, I can't help but shake the thoughts about placebo effect and what exactly defines quality of life in relation to symptoms.
I digress. Anyone have extra thoughts on new / uncommon / "non traditional" science in the field? e.g. neuroimaging, genetic factors (SNPs)?
I was going to write something about how ominous it is to defer to the judgment of a machine for a person's sanity, but the machine of psychiatry to which we defer is no more or less subject to the weaknesses of that kind of deference.
You've identified that a group of humans will make mistakes, possibly as much as machines, and that's not incorrect.
The real hazard, on the other hand, is the unrestrained laziness humans might adopt, once the norm of machine assessment matures. Always trusting the machine answer. Never (or rarely ever) contemplating the judged person. Optical sorting. Rubber stamps.
No longer simple stigma. But endemic callousness, as people stop trying to help anyone, because machines now adjudicate who to simply avoid, until those untouchables rectify their unit tests.
The isolation induced by a machine-endorsed judgement could be far more rigid, than times when people would have had to get to know you, feel you out. And this process could have been theraputic for at least one side of the conversation. Maybe that goes away now.
In a recent Harvard study about how Instagram could predict depression with trained algos, they shared some very high depression rate figures regarding Mechanical Turk workers. We should be doing more to investigate this issue, as the decisions made on MT fuel much current work.
14 comments
[ 0.25 ms ] story [ 42.4 ms ] threadI think that it is much more important to find causes behind "mental disorders" than to describe them better.
From the article:
> According to Schwoebel, it took over 10 primary-care appointments before his brother was referred to a psychiatrist and eventually received a diagnosis. After that, he was put on one medication that didn’t work for him, and then another. In the years it took to get Schwoebel’s brother diagnosed and on an effective regimen, he experienced three psychotic breaks. For cases that call for medication, this led Schwoebel to wonder how to get a person on the right prescription, and at the right dose, faster.
What are the conditions that led this person to have a psychotic break? They didn't care, and just put Mr. Schwoebel's brother on maintenance medication for life. I suspect that such patients are usually 'exhausted' [1].
[1] https://news.ycombinator.com/item?id=12331317 (my comment from this past weekend)
"How Psychiatry Lost Its Way: Identifying new mental disorders and their "curse" had become a productive industry; it is also bad medicine" [2], by Paul McHugh M.D., is very helpful for giving context to the mental health industry's status quo.
[2] https://www.commentarymagazine.com/articles/how-psychiatry-l...
"The Doctor Isn't In" (same author) is also quite helpful: https://news.ycombinator.com/item?id=10315705
No, the problem is that we know a lot more about symptoms than causes, so we diagnose by and treat symptoms, which may or may not be treating causes (though many treatments are chosen because they treat something that there is at least a plausible hypothesis is the cause of the symptoms, and the fact that they have efficacy in treating the symptoms is at least suggestive evidence that both the hypothesized cause is correct and that it is being treated.)
This is also far less different than lots of other areas of medicine that critics of psychiatry try to make it out to be.
http://www.MadInAmerica.com/
You're welcome, thanks for responding.
> i've had bad experiences in my life with psychiatry, been in and out of the system since i was 12 and witnessed a LOT of ... let's say bad practices.
Having seen the inside of this world (as a visitor), I can sort-of imagine what you've been through. People do have problems; the system knows not that it does not always help.
> do you have any other suggested reading / sites for someone in my position?
depends on what you're interested in. I switch to this account whenever I have something to say about mental health, so there might be some other links there. (I just surveyed my comments, there aren't too many. I post the link to Robert Whitaker's work fairly regularly, but there are a bunch of others. They depend on the context...)
https://news.ycombinator.com/item?id=11974769 - this has a link about the monopolists' approach to medicine. I posted that comment from my phone, and didn't feel like fishing out the full link. But the article is quite helpful for illuminating the source of bias in the "Medical" world.
https://news.ycombinator.com/item?id=10974230 - I responded to this person's comment to thank them for their link about "SSRI's"
What are you interested in? anti-psychiatry? self-hacking mental health? The history of Medical Monopoly?
I think I see what you're saying. If $problem is kidney failure, not many people would blink an eye at this method. But something about $problem being purely "mental" has a stigma. A really nasty stigma. Yet this method can still be useful to get people help - that wasn't my personal experience, but I've seen it happen for others. Then again, when I delve into thinking about that, I can't help but shake the thoughts about placebo effect and what exactly defines quality of life in relation to symptoms.
I digress. Anyone have extra thoughts on new / uncommon / "non traditional" science in the field? e.g. neuroimaging, genetic factors (SNPs)?
The real hazard, on the other hand, is the unrestrained laziness humans might adopt, once the norm of machine assessment matures. Always trusting the machine answer. Never (or rarely ever) contemplating the judged person. Optical sorting. Rubber stamps.
No longer simple stigma. But endemic callousness, as people stop trying to help anyone, because machines now adjudicate who to simply avoid, until those untouchables rectify their unit tests.
The isolation induced by a machine-endorsed judgement could be far more rigid, than times when people would have had to get to know you, feel you out. And this process could have been theraputic for at least one side of the conversation. Maybe that goes away now.
Expanded thoughts here: https://medium.com/@dbreunig/do-algorithms-find-depression-o...