On Types of Typologies

My MBTI type is “the type of person who did some looking into it years ago and knows that the MBTI is neither particularly scientific nor particularly consistently applied”. Or, as it’s also called, INTJ
— tropylium.tumblr.com

I’m sick of people hating on the Myers-Briggs Type Indicator.

The argument against Myers-Briggs is that it’s not scientific. The argument for Myers-Briggs is that I’m also the kind of person who did some looking into it and realizes that MBTI is neither scientific nor consistently applied, and I also test consistently as INTJ, so clearly something is going on here. And every time I read a description of INTJ I have to facepalm because I so consistently recognize myself in it.

(Yes, I’m familiar with the Forer effect and have compared it to descriptions of different types. Yes, I could totally believe there is a Forer + placebo effect where knowing that you have been assigned a certain type makes it sound more relevant to you than other types you read. Yes, I’m still impressed with how well descriptions of INTJs fit me. Also, I notice that people on Less Wrong, ie people like me, are seven times more likely to be INTJ than the general population. That seems like a nice objective result.)

I think it’s easy to reconcile “Myers-Briggs is not scientific” with “Myers-Briggs is a useful and real descriptive tool”.

When we say a personality type theory is scientific, part of what we want is certainty that it describes the world in the correct number of categories in the correct way. For example, MBTI has four different variables, each of which has two possible values. This could be interpreted as an implicit claim that there are four dimensions to personality, and only two types of personality within each dimension. And Myers and Briggs had no basis on which to make such a claim and just sort of armchaired it. Interpreted that way, the theory is pseudoscientific and dumb.

We compare it to the more scientifically accepted five-factor model, which used lots of data and a statistical technique called factor analysis to clump out what dimensions people actually varied on. They got five, only one of which (extraversion) is obviously the same as any of Myers-Briggs’ dimensions (although I’ve read other sources saying MBTI intuition is very highly correlated with FF openness). Now it’s not clear the five-factor people got it right; there’s a small cottage industry in finding correlations among the five factors and reducing them to fewer factors, such as two factor or one factor models. But clearly the five factor people are trying in a way the MBTI people aren’t.

But I don’t think MBTI needs to try to be scientific to do what it does. It’s not obvious that MBTI gives you any more than you put into it, and it’s not obvious that it should.

By this I mean that you answer a lot of personality questions on a test, like “Do you like spending time around other people?”, and you say “no”, and then later the test tells you “You’re an introvert”, and then you think “Oh my god, this is amazing, it’s like it’s known me my whole life!”

The claim that MBTI gives you new information would be a bold scientific claim and would require bold scientific evidence. I don’t know to what degree the MBTI people make this claim, but I don’t think it’s necessary for me to enjoy the test and consider it useful. All it needs to do is condense the information you put into it in a way that makes it more relevant and digestible.

To say it does so perfectly would be a bold scientific claim. To say it does so at a nonzero level would be an antiprediction, and one that I’m perfectly happy to hand to it.

So let’s talk about European countries.

European countries were created in a very unscientific way: some people conquered some other people, and then some organizations like the Congress of Vienna drew a lot of little borders. If we interpret European border-drawing as an attempt to capture “types” of Europeans, it makes some obvious errors – Basques thrown in with Spaniards, Switzerland having Germans and French and Italians, Belgium making NO SENSE, et cetera.

On the other hand, it is massively better than nothing. If I say “Andreas is Greek, but Franz is German”, you can make a lot of stereotyped but useful assumptions about them. Andreas probably has darker hair and eyes. Franz is probably more likely to be a rich banker. Andreas drinks more wine, Franz more beer. And so on.

Compare the two-factor model of European classification. We break down variance among Europeans into two dimensions. We find that Andreas is (38, 24) and Franz is (52, 13). That’s…nice. Now when I tell you the two dimensions are latitude and longitude of their location, you feel like you know what I’m saying. But it’s not very human-processable information. If I told you that Maria was at (46, 6) and asked you to describe her, probably your first impulse would be to check a map and see what country that corresponds to – in other words, abandon the perfectly logical Two Factor model and go back to the awful, unscientific Metternich-Bourbon Type Indicator.

Five Factor and MBTI are trying to do fundamentally different things. Five Factor is trying to give us a mathematized, objectively correct version of personality useful for research purposes. MBTI is trying to separate people into little bins that put continuous personality space into discrete and easy-to-think-about terms suitable for human processing, and even very poorly drawn bins will do a pretty good job, just like European countries.

If you want Five Factor to compete with MBTI for some purpose – and really, I’m not sure why you’d care, but some people seem really into this – then it has to compete in the “useful discrete bins” field. Invent a binary opposition between OCEAN – Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism – and SUIDR – Shutness, Unconscientiousness, Introversion, Disagreeableness, Relaxedness. Then let me tell people I’m OCIDN, and they can be like “I’m OCIDN too, it’s like I’ve known you my whole life!” And we can make a web form called “The OCcIDeNt” or something where only OCIDNs are allowed, and talk about how no one else understands the problems that OCIDNs have to face.

Then make a gif about what Harry Potter character each of the thirty-two possibilities corresponds to. Realistically, combine two of the five factors so there are only sixteen possibilities, or else you’re going to be scraping the bottom of the Harry Potter barrel: “My type is Charity Burbage’s pet owl!”

I would be interested in seeing if such a system catches the popular imagination the same way Myers-Briggs does. I would be interested in seeing whether it gives the same intuitive right answers, like having certain types massively overrepresented on Less Wrong. My guess is that it would. But I don’t think MBTI does too much worse in that regard or that the differences would be worth stressing out over.