[carljung]

> Saying that MB is "not scientific" is laughable. It is a measure.

One problem is that the thing being measured can decide how it wants to be measured. It's easy for me to select a type from MBTI.

The claim "you have one of 16 personality types" is not falsifiable. Why does it have to be one of those types?

The things that MBTI purports to measure - E/I, S/N, F/T, J/P are just based on Carl Jung's made up theories about how the mind works.

The questions on the MBTI are supposed to correlate precisely to Jung's scales, but what if they don't? Language is awfully malleable. Do I prefer abstract or concrete? Give me a break. (Yes, I get it, abstract is N, concrete is S.) Would I rather stay home and read a book or go to the party? Can I have a book reading party at home? Can I bring a book to the party? Can I read a book about a party? What if I don't like books or parties?

The whole concept of dualism is pure philosophy.

[raverbashing]

> One problem is that the thing being measured can decide how it wants to be measured.

This is a problem in psychology regardless of the type of test. If it's a person asking questions they might be able to detect if someone is skewing the answers

> The claim "you have one of 16 personality types" is not falsifiable. Why does it have to be one of those types?

You have 4 variables, X0 to X4, and a measure of each (let's say between -1 and 1). Then you divide this space in 2 regions per variable. That's how you get 16 (2 ^ 4) types. You might pick 3 regions per variable and have 81 types, harder to work with though)

> The questions on the MBTI are supposed to correlate precisely to Jung's scales

I agree with this, but when measuring something with lots of variables you always have a certain dimensionality reduction (and noise).

And of course it varies with age and situation, but you will be hard pressed to have a INFP as a football coach

But of course you know all that ;)

[lmm]

> The claim "you have one of 16 personality types" is not falsifiable. Why does it have to be one of those types?

We define there to be those types when we're measuring - that's like saying the claim "an object has a temperature that's a number" is not falsifiable. We can test the same object multiple times with multiple thermometers and get the same results (while getting different results for different objects). That's the part that's falsifiable - test-retest reliability. If we have that then we can be confident that our measure measures something. (Then the next step is seeing whether the measure is useful by seeing if it correlates with other things).