Then let me try to help clarify. MBTI has no predictive power ("I'm an INTP, so... ?") and is not falsifiable. Those are the delineators between science and pseudoscience, so that's why I bucket it with astrology.

Since mbti essentially neuters it's own usefulness in the "round those percentages down to binary" stage, you could end up with a situation where an INTP and an ESFJ have only an 8% total difference in mbti-measured personality between them.

The question isn't whether people who are in the middle of each criteria exist and could easily swap classifications to their polar opposite. The question is what percentage of people would do that and whether it's high enough that there's too little signal to noise to extract any meaningful information out of the classification system.

Don't invert the burden of proof. It's up to the proponents of MBTI to demonstrate its effectiveness.

As for signal-to-noise, mbti is a useful tool for measuring the traits it measures, it just, for some reason, cuts the signal in half by removing the percentage from the label.

Without the percentages, you can never assume a persons traits don't hover at that middle ground. And thus, can never assume any part of their label is accurate.

If the occurrence of these hover cases were more like <4%, I'd be more inclined to agree with its usefulness, but as it is right now, the 4bit label, on its own, is just noise.

i'm pretty sure someone who has gotten INTP every single test in his life will behave differently on average than someone who has always gotten ESFJ

I think its a mildly useful categorization which allows a simpler and more coarse view of personality than the full spectrum of data would normally allow.

If I ask you what your weight and height is today, it's indicative of what your weight and height will be a year from now even if the number doesn't fit into exactly the same rounded range. Doesn't mean weight and height are meaningless if the answers change classifications a bit.

A much better analogy would be blood sugar level. For many healthy people its sorta constant-ish. The fact that its not nearly constant for a subset of people doesn't prove the test is therefore worthless. It just proves some people have untreated diabetes. The test is not worthless because it can be used in diabetes screenings, unless the point of the test was to tap people like maple syrup trees or something really weird like that.

A major problem with the test is we know people are not terribly good at self introspection, we're all above average drivers and lovers and thinkers, all of us. I am good enough at self introspection to tell I'm pretty well introverted, yet I can imagine someone not so observant or self aware or perhaps neurotic about the whole topic, getting a fairly random score. Or I'm bad enough at self introspection to not realize I'm no good at it at all, and I'm actually a closet extrovert, although I find that rather unlikely (although I would, wouldn't I, if I failed at introspection).

In that way as a secondary measurement I'd propose theoretically that the wider the variation in test results, the worse the person would score in some kind of self introspection ability test. It seems a reasonable hypothesis. You'd need correction factors to deal with people having diagnosed or undiagnosed swings in brain neurotransmitters or use of addictive substances (including sugar and caffeine). But it could be run as a study.

That said there are other criticisms, for instance: Are thinking and feeling really an axis or are they two independent traits? Are there other useful axes not included?

While I sort of cringe at the idea of using this for assigning people jobs, there area few people in my life that fit some of the type narratives fairly well and it helped me understand their motivations a bit better, which did legitimately help my relationships with them.

That doesn't seem like the proper question. Specifically, it works only to predict future MBTI results, not anything further. That's fine for height and weight because those values are robust to small/moderate changes in value.

MBTI already has low predictive power about real-world content, and high sensitivity to change (on the margins; it also suffers from applying a clustered model to non-clustered data). So if retest reliability is anything less than excellent, I worry that our predictive power has been almost totally destroyed.

To provide a new analogy, this is like variance in number of fingers. The significance of 11 vs 10 vs 9 isn't proportional to magnitude, so we wouldn't accept "small" counting errors.

None of that says that "I'm an INTP" is meaningless, but I suspect that MBTI is no better, or even worse, than non-formalized personality summaries.

(Of course, MTBI is neither accurate nor precise. It is not accurate because it does not provide us any useful information about personality. It is not precise because test results vary wildly.)

Of course not. All the companies are simply wrong. They're falling for management/hiring fads, or relying on old superstitions. This happens all the time. "Companies are using it" says nothing about whether it's actually reliable.

Specifically, many studies and meta-studies have failed to show correlation between MBTI results and job performance. Here is one such summary: http://www.opd.net/abstracts5.html

Furthermore, your willingness to buy into the narrative (and others' willingness to also believe it) that by categorizing certain people into stereotypes and assuming that most of these people tend to excel at a particular set of tasks is most likely what limits them from excelling at tasks that lie outside of their respective sets just as much as their true predisposition to not excelling at them.

Saying that any profiling technique has zero predictive value is a very bold claim.

Even one's astrological sign has predictive value: [1]. And yes, this has been scientifically proven.

[1] http://www.medicaldaily.com/how-birth-month-influences-perso...

For example, companies fruitfully extract predictive power from whether a user clicked a link or not. Facebook extracts predictive power from whether users like videos, et cetera. So why shouldn't an elaborate test that was developed to have meaning, not have any predictive power?

There are results that show MBTI has no predictive power against job performance. It has not been tested against any other measure.

That said, MBTI shows bad orthogonality in the factors in its response. This suggests the factors are ill defined.

Could you provide a reference?

Does this have predictive power? Yes. If I needed somebody to take care of small children, would this test be useful? Yes of course. If I had to pick people blindly for this task, except for the results of this test, then this test would be very useful. Now, MBTI is more sophisticated than this simple test. So to say that it has no predictive power is just silly.

I'm saying that the fact that some people are using it for some purpose doesn't mean that it's actually useful for that purpose.

As for your response, the burden of proof really lies with its proponents - is there any proof of its alleged predictive power?

Without this, I don't think it's reasonable to assume that it is in fact predictive.

For example, the simple test of whether a user clicks a certain link or not already has predicitive power (companies are using this, fruitfully). So a more sophisticated test like MBTI certainly has some predictive power.

[1] https://en.wikipedia.org/wiki/Phrenology

Additionally, the added (relative) complexity doesn't imply anything about its qualities.

This is a null argument.

Unless you show that those predictions actually assess whether someone is fit for those tasks this is absolutely content free.

MBTI is a subjective categorization. Astrology is more like your DNA. You're born with a specific set pattern. Just like we don't understand what most of our genetic patterns mean, we don't understand what most of our astrological patterns mean.

http://www.nytimes.com/2015/03/08/magazine/is-most-of-our-dn...

The problem is not variation over time - that's fine. The problem is that when someone gets a result, that result doesn't say what it claims to about them. It's similar to astrology because while it has a series of well-defined rules for making a decision, that decision does not reflect reality. It is therefore effectively arbitrary.

Both MBTI and astrology are internally valid - that is, the same person will get the same or similar results the majority of the time. That doesn't mean they mean anything.

If say some person knows two people and that one tested as INTP and the other as ESFJ and this person doesn't know for a fact who had which test results. Are you really sure they won't have better than even odds at figuring out which is which based on their interactions with the people that don't involve asking for their MBTI test results? If so, based on what study?

It's clear I wouldn't be able to tell them apart based on their birthday (unless they were super young or something), but it's not at all clear that it's the case for MBTI.

If you say "the sky is green," it's not my job to prove to you it's not. It's your job to prove it is. You're the one making the claim.

But here is a (more politely worded than I) paper on the MBTI's ineffectiveness for your perusal: http://epublications.bond.edu.au/cgi/viewcontent.cgi?article... - simply googling "mbti external validity" will also give you plenty of explanation. I also recommend looking at the many studies that review cites, since its whole purpose is amassing information on whether the MBTI is effective.

There's a difference between saying the extent of something's effectiveness hasn't been established by existing research and saying that it is ineffective. I've read the publication you are linking to before and it is saying the latter without the evidence necessary to back up the claim.

I don't want to be "credentials guy," but I have a degree in cognitive science and I've conducted human subjects research. What you are saying is simply wrong.

That publication cites dozens of studies to back up its claims. Are you suggesting they are all invalid? If so, why?

Combined with the fact that this is basically a listicle ("quoticle?") clickbait from a mid/low-brow popsci website, with no analysis or synthesis, and you can pretty much throw TFA away, regardless of your side of what's shaping up to be a religious war here on HN.

I can't prove a negative, all that a study can do is say "no results observed", but there's plenty of that. From the researcher cited in the article: "Many very specific predictions about the MBTI have not been confirmed or have been proved wrong." "Finally, there is no evidence that the MBTI measures anything of value."

This aligns with all of the other at-all-convincing research I've seen on MBTI: it assigned binary distributions to normally distributed data, and has very low predictive power. The majority of people will be near the middle of those distributions, where MBTI is at its least useful.

Further, I think your example is unfair: MBTI is a one-way function. I can meet people and anticipate their scores to with 1-2 errors (a huge margin, but we talk like it's 'close'!) I can't go from an MBTI score to useful predictions about the person though, because it's so massively information-destroying. Even on the things it purports to measure, the normal-to-binary problem means that it has awful predictive power. MBTI defines a somewhat-consistent function from people to letters, but since it's an arbitrary clustering system it doesn't actually allow new predictions or efficiently compress information about people.

So yes, I could probably tell some people apart by MBTI, but you could also assemble 10 different ESFJ's who are more dissimilar than the ESFJ and the INTP. It's like a compression algorithm that can't be reversed, and about as useful.

I don't disagree with the fact that it's not necessarily super effective because the distributions of personalities of people are not bimodal, as would be ideal for this simplistic sort of classification. It's also unlikely to be supremely effective at clustering in that way because the design of the metric wasn't really data driven in any systemic way. It's not even like someone used principal component analysis to really establish that the four classifiers are reasonably orthogonal. It's probably not super obvious/trivial to find human language-friendly orthogonal aspects of personality that are each bimodal, but I imagine the surveys could be improved informed by better data science.

It's certainly a compression algorithm, as are all classification based metrics based on answers to multiple questions, and I'm not arguing it's the most effective one around. However I think before making claims like it has no "external validity" the degree to which it might have it should actually be tested with reasonable methodology and I'm offering something that can absolutely be tested scientifically. There's nothing impossible about having some people take the test and having others who know them try to classify them without the results and see how well the two coincide. It's also possible to evaluate the degree to which people can decompress the data. You can set it up where people have equal time spent with two people of each MBTI and see how well they can guess their answers to the original survey questions with additional knowledge of the MBTI as opposed to without it. It won't establish an upperbound for how useful an MBTI is but it will help gauge a lower bound and I'm pretty confident the lower bound will be higher than that of a zodiac.

I imagine we can design a better classification, probably with even the same survey questions, but just because a metric wasn't well-designed doesn't mean the information it provides is meaningless.

"INFJs are strong leaders" is as arbitrary as saying "aquarius's are strong leaders"

The problem i, and many other people find with the mbti is that most of its users, and even scientific apologists, can't help but fall into the trap of categorizing people into perceived behavioral trends based on their types, which is at best, scientifically invalid, and at worse, harmful.

I'm really a very nice person. I'm magnanimous, full of goodwill and compassion and everyone I meet immediately likes me. I am intelligent and brave and very gregarious. I am the best person to be around when you need a shoulder to cry on because I have a deep understanding of human nature, thanks to my very high emotional intelligence.

That is the problem with taking into account people's stated behaviour and inclinations.

Of course other parts of the article talk about the predictive power of the personality test. On the other hand, those excerpts don't talk about horoscopes.