Hacker News Reader

Comment by geor9e | Hacker News Reader

Reminds me a bit of the MBTI personality test, where they make up 4 types of question to arbitrarily split the population, so 2x2x2x2 = 16 subtypes. It's true by it's own definition. Which is fine, but are these particular arbitrary boundaries useful? Perhaps. Could the splitting lines have been just as useful in different arbitrary place. Perhaps. A lot of people who take the MBTI find they're on the boundaries flip-flopping into a few different pigeonholes depending on different times they take the test. So it's important to let people know they can be in multiple buckets (are a bit in all of them), and take a little advice from each community.

For this one they split autism into 3 groups (core,social,developmental) then split core into (mild,severe), to make 4 total.

View on HN

> Troyanskaya and her colleagues investigated the variants associated with a person’s collection of characteristics. They applied a statistical model to data on the traits and behaviors of 5,392 autistic people from the SPARK research cohort. By adjusting the number of groups, the team found the most significant similarities among participants when the model sorted the cohort into four autism subtypes.

https://www.medrxiv.org/content/10.1101/2024.08.15.24312078v...

Not my field, but the statiscal analysis seems legit.

Have not read the article in depth yet, but I am very familiar with the Troyanskaya lab. They are good and their articles are always worth a read.

They are using a geodesic finite mixture model, which I am not familiar with, but seems to be an extension of mixture models to non Euclidean metric?

Pleiotropy can be tricky though, as traits / diagnoses / observations can have causal relationships amongst themselves. For example, a variant that impacts obesity may be statistically associated with heart attacks; but the relationship between the variant and heart attacks is not directly causal in the sense of something like a variant that alters function of an ion channel expressed in cardiac smooth muscle.

It’s probably more helpful to think of them as spectra or overlapping circles of behavior. When we talk about psychological disorders, they are clustered and only useful insofar as they allow professionals to talk to one another and mean the same things.

Undoubtedly each human displays a unique case and manifestation of whatever behavior we choose to specify, and that individuality can pull from different “trait” or “behavior” groups on a scale.

It’s only when we start to lump together many cases that we begin to discern between types and patterns, but it’s silly to make a harsh distinction between clusters and then pretend like they are emphatically different from the next. Grouping can have its uses, but we shouldn’t forget each person presents uniquely.

Well said. Psychological diagnoses are genres for the brain.

There’s a better way to frame this issue: being able to consistently label a characteristic is only useful if it can help in identifying the underlying mechanism or is helpful in choosing an effective treatment.

A whole lot of mental health knowledge fits into the category of repeatably identifiable categories of dubious usefulness.

Think phrenology: it is indeed possible to categorize the shape of people’s head in a statistically valid way… it just has nothing to do with their mental health.

Are you a libra? This sounds like something a libra might would say! /s

“Repeatable identifiable categories of dubious usefulness”

Yep, there’s a lot of this in modern science.

“Is this method of classification useful? wrong question! All you need to ask is, is it easy to write a reproducible paper (or shareable web content) on this method of classification? Yes, yes it is!”

There’s a lamppost problem at work here.

Eh, it's more like the DSM was written primarily with the concept of billing and charting in mind. If you track and bill medical conditions, you need a consistent language for things that will be consistent across medical providers.

It solves a problem, just not the problem most people think it solves.

https://en.wikipedia.org/wiki/DSM-5

If we consider science as a way to document reality (among other things) then "information -> decision" is an elegant point of view (if information doesn't help make decision, it is just noise). https://news.ycombinator.com/item?id=41881872

It reminds me of "brain (mostly) for movement " (no decisions to make, no brain is necessary). Sea squirts being a prime example: "As an adult, the sea squirt attaches itself to a stationary object and then digests most of its own brain." (Rodolfo R. Llinas I of the Vortex: From Neurons to Self) https://www.amazon.com/Vortex-Neurons-Rodolfo-R-Llinas/dp/02...

If the data is a roughly uniform distribution across the dimension space, then yes absolutely.

If, however, there is significant clustering within the dimension space, and those clusters are taken as groups, then it seems valid.

I would tend to be generous to the researchers and assume a mostly discrete clustering until I see otherwise. To apply grouping like this without discrete clusters would be unprofessionally naive on their part.

The distribution of the data is also the main problem with MBTI, AFAIK. Intra-/extraversion for example have a unimodal distribution with most people being somewhere in the middle. And exactly that middle is where MBTI makes the cut on their classification.

This is a great point. We have a generally tendency to typify in binary[black and white] terms. It overrides our ability to express nuance and un this case pushed out a larger category (if we even wanted to still categorize) of ambiverts.

MBTI speaks more on our drive for categorization than it does on the people being categorized.

Reminding is different than being the same.

For example, Big5/OCEAN/FiveFactor model uses factor analysis which has statistical reasoning behind the number of factors. In fact, the semantic meaning of the factors comes up after the analysis, not before. It's the same as clustering or GMM, the semantic meaning of the clusters is applied after the partitioning.

The concept of partitioning people is the same, however, the order by which meaning is applied is opposite which makes them completely different in practice.

descriptive vs prescriptive

Data driven vs some writers who read Jung.

They also posit correlations with epigenetic differences. If there’s distinct biological mechanisms at play, this gives some credence to splitting autism into separate conditions

It's not true by its own definition, it's just that if you soak it in, you will actually start to see the patterns behind it among people. It's still going to be pseudoscience because there are a lot of variables, but it often works and even more often people don't know what to look at or what to put together, because you'll find really A LOT of articles on the Internet that try to run mechanics on certain types without understanding what or why. I assume that if you want to put this together more precisely, there is a lot of scope here.

And lest it be said that I'm talking out of turn myself, I only became interested in this whole MBTI thing because an ENFP once told me I was an INTP after a few hours of talking about silly things. That's exactly what these tests once told me. Of course, these are still anecdotes, but we are sciency.

Is it a problem that someone has catalogued autism in this way? Is it a question of lack of precision or bad direction? Am I asking the wrong questions?

It's true by definition because MBTI splits people into two groups, four times in a row. Everyone is going to fall into one bucket or another, just as the last 4 bits of any integer is going to be one of 16. The question is whether the distinction has any practical uses.

My guess as to why MBTI is more popular than other equally valid personality tests is that the splits are defined in such a way as to capture a few traits that people find the most irksome in their everyday relationships. You want to take a rational approach, but your SO gets all emotional? Well, that's a T vs. F thing. It might not be medically meaningful, but IMO it does help us appreciate that other people's minds might be wired differently. Not wrong, just different.

Business idea: AI companions with custom personality out of the box. Choose the MBTI type you want!

You are writing about the highest level of abstraction, instead of the functions assigned to specific type and lower. Besides, it is common to combine the MBTI with the Enneagram. You literally comment on what you see at a glance.

Reply on news.ycombinator.com