Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[silverspawn]

So I think I should probably state what I expect MBTI to do:

- I expect MBTI to be predictive of people's romantic interests. But that does not mean that profile A will want to date similar profiles, or that profile A will always date profile B. It means that if you are interested in someone with profile B, then anyone else you're interested in is much more likely to have a profile similar to B. So a correlation among different interests of the same person. I expect this to be stronger than Big5
- Same for friendships
- I expect MBTI to be super predictive of various professions, like becoming a processional chess or starcraft player (or any other esport that doesn't have team play), and probably also for lots of other professions (although not all). I expect some of these to be much stronger than Big5, although certainly not all
- I don't expect it to strongly correlate with income or happiness or longevity or health or IQ or whatever because there's just no obvious reason why it would correlate with these particular things. There are probably weak correlations because weak correlations exist between lots of things, but this isn't what it's primarily measuring.
- Obviously you should use percentages and not categories for all the above

Basically, I expect MBTI to be better at all the interesting stuff like actual personality differences and worse at all or at least most of the boring stuff like future income of job success

[faust]

So I think I should probably state what I expect MBTI to do:

- I expect MBTI to be predictive of people's romantic interests. But that does not mean that profile A will want to date similar profiles, or that profile A will always date profile B. It means that if you are interested in someone with profile B, then anyone else you're interested in is much more likely to have a profile similar to B. So a correlation among different interests of the same person. I expect this to be stronger than Big5
- Same for friendships
- I expect MBTI to be super predictive of various professions, like becoming a processional chess or starcraft player (or any other esport that doesn't have team play), and probably also for lots of other professions (although not all). I expect some of these to be much stronger than Big5, although certainly not all
- I don't expect it to strongly correlate with income or happiness or longevity or health or IQ or whatever because there's just no obvious reason why it would correlate with these particular things. There are probably weak correlations because weak correlations exist between lots of things, but this isn't what it's primarily measuring.
- Obviously you should use percentages and not categories for all the above

Basically, I expect MBTI to be better at all the interesting stuff like actual personality differences and worse at all or at least most of the boring stuff like future income of job success

Why?

[silverspawn]

Because the four axes seem to capture really important psychological traits (moreso than Big5), and despite everyone shitting on it, I haven't seen any disconfirming evidence. In fact the evidence about communities I keep bringing up, e.g. 4 of 5 prismata players in our former team being INTJ despite INTJ being like 2.1% of the population, seem to strongly suggest that it captures something important.

And ofc I'm cautiously generalizing from myself

[silverspawn]

But I'm open to change my mind if someone can demonstrate that it doesn't do the things I think it does

[silverspawn]

KIMBA YOU'RE BAD

[Awaclus]

The source just links to the personality test. (I was gonna book mark it to use it as evidence against the next person who claims MBTI is meaningless.)

The source is the personality test.

[Awaclus]

Because the four axes seem to capture really important psychological traits (moreso than Big5), and despite everyone shitting on it, I haven't seen any disconfirming evidence. In fact the evidence about communities I keep bringing up, e.g. 4 of 5 prismata players in our former team being INTJ despite INTJ being like 2.1% of the population, seem to strongly suggest that it captures something important.

And ofc I'm cautiously generalizing from myself

I bet there exists a combination of Big5 results excluding neuroticism, compressed to binary values for each axis, that a majority (or close enough to majority that 4 out of 5 happens by chance reasonably often) of dedicated hypercompetitive strategy game players have in common.

The problem with that sample is that it is at a very extreme end of the general population. The vast majority of INTJs are not hypercompetitive strategy game players or have any particularly similar interests to that, and I'm pretty sure whoever the fifth one in the team was (or, say, me who is also a non-INTJ Prismata player) had a much more similar personality to everyone else there than the average INTJ does.

[silverspawn]

Because the four axes seem to capture really important psychological traits (moreso than Big5), and despite everyone shitting on it, I haven't seen any disconfirming evidence. In fact the evidence about communities I keep bringing up, e.g. 4 of 5 prismata players in our former team being INTJ despite INTJ being like 2.1% of the population, seem to strongly suggest that it captures something important.

And ofc I'm cautiously generalizing from myself

I bet there exists a combination of Big5 results excluding neuroticism, compressed to binary values for each axis, that a majority (or close enough to majority that 4 out of 5 happens by chance reasonably often) of dedicated hypercompetitive strategy game players have in common.

maybe but I'm not convinced that's true

[silverspawn]

ok so I'm mostly done with the chess course. Time to pick up online chess again and figure out a way not to spend too much time.

The goal is to reach 2000. But first I have to get back to 111111111111188888888888888888888888880000000000000000000000000000000

[faust]

But I'm open to change my mind if someone can demonstrate that it doesn't do the things I think it does

The way I see it, this seems hardly possible.

- friendships and romantic interests: This just seems hard to study, given that you can't just use a random sample of the population, you'd have to get all ex-partners of some person or a whole friend group to participate. So I don't expect to find any data on this.
- profession prediction: You have put a qualifier on this so that even if I found evidence that indicated some profession has about the same distribution as the general population, you could always turn around and say "well, that just means this is not one of the profesion that MBTI is a predictor for!"

So in that way, you claims are all but non-falsifiable.

[silverspawn]

But I'm open to change my mind if someone can demonstrate that it doesn't do the things I think it does

The way I see it, this seems hardly possible.

- friendships and romantic interests: This just seems hard to study, given that you can't just use a random sample of the population, you'd have to get all ex-partners of some person or a whole friend group to participate. So I don't expect to find any data on this.
- profession prediction: You have put a qualifier on this so that even if I found evidence that indicated some profession has about the same distribution as the general population, you could always turn around and say "well, that just means this is not one of the profesion that MBTI is a predictor for!"

So in that way, you claims are all but non-falsifiable.

If you found no correlation for a bunch of reasonable professions then I wouldn't do that, and also I included friendships which are easier to do -- but nonetheless, yes, I agree that the position is unlikely to be falsified, especially since MBTI is so out of academic favor. But I didn't adjust my views to be hard to falsify, this is just what I actually expect the test to do.

[Awaclus]

If you found no correlation for a bunch of reasonable professions then I wouldn't do that, and also I included friendships which are easier to do -- but nonetheless, yes, I agree that the position is unlikely to be falsified, especially since MBTI is so out of academic favor. But I didn't adjust my views to be hard to falsify, this is just what I actually expect the test to do.

For what strengths of correlation would you consider there to be "no correlation"? It's almost certainly going to be an extremely weak correlation just because most people are close to the middle with their MBTI non-binary values and the most common professions are common only because we need a ton of people to do those jobs, so there's some combination of high pay, low skill requirements and ease of getting employed that causes a lot of people to do the job even if it's not what they would ideally prefer to do.

If you limit the professions to something extreme, like professional e-sports, you are still not going to find much of a correlation because the vast majority of people, including the vast majority of people who do have the relevant MBTI type, won't have that profession (because they're too busy having those aforementioned common professions).

[silverspawn]

If you limit the professions to something extreme, like professional e-sports, you are still not going to find much of a correlation because the vast majority of people, including the vast majority of people who do have the relevant MBTI type, won't have that profession (because they're too busy having those aforementioned common professions).

The number of people in esport actually doesn't matter for the strength of the correlation.



It's a bit more complicated since you have 4 continuous variables and not 1, but that won't change the principle

[silverspawn]

Wait I'm stupid it totally does since then p0*p1 is smaller. I'm now feeling pretty confused about this

[silverspawn]

So yeah I had no idea that correlation between a binary and a continuous variable is sensitive to how skewed the binary variable is. Very interesting. In this case what I actually intended with my prediction is not the correlation but just the difference between the two means. I.e., mean of, e.g., the J-P axis for professional esport people vs. the general public.

[silverspawn]

I guess it makes sense since correlation tells you how much knowing one thing reveals about the other thing, and if one group is very small, then knowing whether someone is in that group will on average not tell you much because the answer is almost always no. But I've never really grasped what that means, of if I have then I've forgotten about it.

Like the correlation between smoking and lung cancer will go down if fewer people smoke, but that's not exactly helpful for someone considering whether or not to stop smoking.

[Awaclus]

It doesn't even have to be a binary value, it works the same way with two continuous variables (or two binary values). If the relationship between x and y is not linear but e.g. exponential, that isn't a very strong correlation between x and y. This is also not just "well technically correlation only means this specific thing" bullShiT, but it is actually important that correlation only means the specific thing that it does — e.g. there is obviously a connection between how many coal power plants we have running and how bad climate change is, but the relationship is far more complicated than a correlation, so we can't just be like "oh well, climate change is tolerable so far so we'll keep building more coal power, and as soon as it becomes too bad, we'll turn one of the power plants off and we get the maximum benefit from coal power while climate change won't be too much of a problem" (which is a reasonable approach to things that are actually correlations where the first thing is good, some amount of the second thing is acceptable, and too much of the second thing is way more bad than the first thing is good).

Like the correlation between smoking and lung cancer will go down if fewer people smoke, but that's not exactly helpful for someone considering whether or not to stop smoking.

We can still take prevalence of smoking over time, prevalence of lung cancer over time, and find the same correlation there always was.

[silverspawn]

I grant all that (good thing at least one of us knows math~) but I think for cases like smoking, the most reasonable measure is what you get by just dropping the sqrt(p0*p1) term from the correlation, which apparently is then called Cohen's effect size or Cohen's d. Like there are some cases where it makes sense to adjust for size of the binary group and some where it doesn't. And in our original case of looking at MBTI scores for esport people vs general population, it definitely doesn't.

[Awaclus]

I grant all that (good thing at least one of us knows math~) but I think for cases like smoking, the most reasonable measure is what you get by just dropping the sqrt(p0*p1) term from the correlation, which apparently is then called Cohen's effect size or Cohen's d. Like there are some cases where it makes sense to adjust for size of the binary group and some where it doesn't. And in our original case of looking at MBTI scores for esport people vs general population, it definitely doesn't.

It definitely does! You're expecting MBTI to be "super predictive of" e-sports, but if you're just going to go around asking random people's MBTI scores and trying to predict whether or not they're professional e-athletes based on just the MBTI score, you're going to make a godawful prediction almost every time you run into someone with the score you think predicts being an e-athlete.

[Awaclus]

In other news, I just found out that there's a 3rd party desktop program for ChatGPT which is based on the OpenAI API keys, which is nice for two reasons:

1) It's way more convenient than using a browser for it
2) You pay the API request rates instead of having to pay a monthly subscription fee to access GPT4, which could actually cost more money if you use it a lot, but I (and probably most people) don't use it anywhere near that much.

[Awaclus]

It's called NoFWL.

[silverspawn]

I grant all that (good thing at least one of us knows math~) but I think for cases like smoking, the most reasonable measure is what you get by just dropping the sqrt(p0*p1) term from the correlation, which apparently is then called Cohen's effect size or Cohen's d. Like there are some cases where it makes sense to adjust for size of the binary group and some where it doesn't. And in our original case of looking at MBTI scores for esport people vs general population, it definitely doesn't.

It definitely does! You're expecting MBTI to be "super predictive of" e-sports, but if you're just going to go around asking random people's MBTI scores and trying to predict whether or not they're professional e-athletes based on just the MBTI score, you're going to make a godawful prediction almost every time you run into someone with the score you think predicts being an e-athlete.

Right. So I said it definitely doesn't because it's not the fault of the test. If the two groups have different means, it seems like the test is working, and if the one group is small, well that's not really something the test can help. So Cohen's d seems more fair

[silverspawn]

In other news, I just found out that there's a 3rd party desktop program for ChatGPT which is based on the OpenAI API keys, which is nice for two reasons:

1) It's way more convenient than using a browser for it
2) You pay the API request rates instead of having to pay a monthly subscription fee to access GPT4, which could actually cost more money if you use it a lot, but I (and probably most people) don't use it anywhere near that much.

Does it know about current events? (The monthly version has a knowledge cutoff on January 2022.)

[silverspawn]

Imagine how nice it must be to have zero social "intelligence". You just don't notice when you're annoying people don't hence don't even have to worry about it.

[silverspawn]

This does not refer to events related to this thread