Hacker News Reader

> A problem with modern psychology is the diagnostic criteria aren't very much based on underlying causes as on being able to reliably diagnose patients based on symptoms regardless of physician.

That’s generally true of diagnosis when we don’t understand the mechanism, and since where we do understand the mechanism we treat that as a reason to move the thing-with-behavioral-symptoms out of psychology and into some other medical domain, so its essentially an irreducible problem with psychology.

A diagnostician which crosses all specialties would be really nice. I imagine this is an area where an easily replicated computing tool might be the best advancement in the field. Initial symptoms and the low hanging easy tests (as well as any others 'that may as well also be run' for any labs or test types indicated given a moderate extra effort) would drive a root cause diagnosis and recommendation for a specialist who can best treat any of the possibly multiple issues.

The reason doctors become specialized is that they can't learn multiple, and definitely not all, specialities to a good enough degree to be able to diagnose and treat patients across them. There's probably a better chance to have a diagnosis team that meets and handles this type of thing, but it would likely significantly increase the cost of medical care if every patient had to go through a team of specialists.

And as for computer aid in this area, it would be awesome, but I fear we are still far away from it, especially since diagnosis is an interactive, continuous process, and is not separate from treatment. The patient presents with some symptoms a history and a current course of treatment, you order some tests and may even prescribe some medication for a preliminary diagnosis, then you wait and adjust as new information comes along. I'm not sure current approaches in machine learning deal very well with this type of process.

I find it remarkable that conventional expert systems didn't succeed at that. It seems like a straightforward application. If you have symptoms X, Y, and Z, then it could be P, Q, or R, so run tests T1 and T2.

Obviously I'm not a doctor and can only assume that this is like somebody telling me "Debugging is easy, you just set a breakpoint..." But I'd be curious to know more about why this approach failed.

I think the problem is that the expert system's fact base would have to be essentially the world's most complex medical encyclopedia, except that encyclopedias are addressed to thinking humans, whereas this would have to have enough detail so that it can be "read" by an extremely simplistic algorithm. And not only that, but it would have to be in constant update as new medical knowledge is discovered, new medications (with new interactions) come out etc.

I also wonder how well the inference algorithms would actually perform with the scale of facts you would need to add.

For well understood you don't need complicated machine learning you just need a searchable dictionary of test outcomes and flow diagnosis flow charts, these things already exist.

What you describe is essentially a general practitioner. If you look at what they do (and how their training is organised), they essentially follow decision trees to either a clear diagnosis or "refer to specialist". The difficulty for using computers for this tasks is that several of the things going into the tree are based on observations of the patients behaviour, look, speech etc., which is difficult to categorise for a machine.

> A problem with modern psychology is the diagnostic criteria aren't very much based on underlying causes as on being able to reliably diagnose patients based on symptoms regardless of physician. This leads to people agreeing what your diagnosis is

That is true. The groupings persist because they are useful in facilitating communication regarding potential complications, prognosis, and patterns of response to treatments.

People seem intent on throwing away the above utility with a backhanded “but that’s not the Real Diagnosis.” It’s not, but within the limitations of our current understanding of neurology, it’s the best we have come up with so far (allowing for some limitations due to the pace of spread of innovations, politicking, etc.)

Clinician and researcher in mental health here. In my view, you're being too charitable.

> The groupings persist because they are useful in facilitating communication regarding potential complications, prognosis, and patterns of response to treatments.

These are reasons, but they're a distant second. Mainly, they persist because those groupings - diagnoses, in other words - have become necessary for billing. The DSM, for instance, is an economic document, not a scientific one. For most people, it sits on a shelf, gathering dust.

As for "the best we've come up with so far", you'd perhaps be interested to learn how often the decisions about these groupings are made because of politics and economics, and not by science [1].

[1] https://www.amazon.com/They-Youre-Crazy-Paula-Caplan/dp/0201...

Thank you for coming out with us! I tell my doctors all the time, I’m tired of being diagnosed, I want you to listen to how I feel. That has got to be so much further in helping me than any diagnosis has.

> The point is "depression" or "PTSD" probably don't exist among many others. There are perhaps dozens or hundreds of unique conditions currently lumped together because they present in vaguely similar ways - the good news is that we are making progress towards much more arguably "real" understandings, the bad news is we're still quite far away from diagnosing mental conditions with the same specificity we can diagnose cancers or viruses.

One sees this in a bizarre number of fields, that specialists act as though the classifications actually be as rigid and objective as in, say, physics, and that something objectively is or is not something, and then debate it.

I've actually read some debates between linguistics that are quite passionately debating whether something is an allophone of the same phoneme, or a different phoneme altogether, and I do not find such distinctions to be all that objective as the language they used suggests.

Name magic.

To know a thing's name is to have power over it.

From ancient mythos to modern specialists and especially philosophy, naming things gets a whole lot of focus. It does because it is terrifically useful! Naming, separating, making distinctions, this is an essential part of understanding.

However a whole lot of these kinds of debates don't acknowledge what kind of a debate they are actually happening. Is the subject of debate the limits of where a definition matches and were it doesn't? Is it whether the subject fits within already agreed limits? Is it really about an unspoken assumption that differs between the parties arguing? Many also don't address the quality of the definition. Is it a useful definition that represents defensible differences between "has this name" and "doesn't have this name" or is it an entirely arbitrary classification (say, look at the color spectrum and define areas which are red and which are orange; now zoom in on a boundary between the two, is there any defensible reason why the border should not be moved slightly to add one color to red? No, probably not)

> I've actually read some debates between linguistics that are quite passionately debating whether something is an allophone of the same phoneme

In this example are they really arguing about the something in question or are they arguing about the definition of "allophone" without realizing it. The followup question is "what is the reason why this edge case is relevant?" and "if this edge case isn't relevant, is the term 'allophone' even useful?"

> In this example are they really arguing about the something in question or are they arguing about the definition of "allophone" without realizing it. The followup question is "what is the reason why this edge case is relevant?" and "if this edge case isn't relevant, is the term 'allophone' even useful?"

I do not believe they are arguing about the definition either, nor do I believe that the classification has any use as whichever way it falls changes nothing.

Linguistics is a field plagued by classification for it's own sake, as many others are. Objects are classified but no conclusions can be drawn from the classification. — perhaps it is but simply a form of mental file organization.

In the majority of US States, psychologists cannot prescribe medications.

https://www.verywellmind.com/can-psychologists-prescribe-med...

These conversations tend to conflate psychology and psychiatry. Which isn't altogether unreasonable.

I would have used a more generic term to include both if I knew one. Psyche-related healthcare providers and the science behind them? Meh.