Problems with MBTI Tests/New type of test

[mgb]

I've been thinking a lot lately about about MBTI testing, especially compared to the cognitive processes test, and I don't think a lot of the tests do a very good job.

I think the tests try and measure your T vs F, I vs E, N vs S and P vs J. The most simplistic ones, like say humanetrics will only measure one against the other. Some like similarminds do the same thing but on a bit of a scale.

I'm not convinced this is the best way to test. I think some of the individual components of type are different enough that they shouldn't be tested against each other, lest you get a score that is somewhat out of whack from how you really are.

I think the test should be more like an IQ test. It should test each function individually and attempt to measure your capacity in that function and determine your type from that. This might be somewhat more difficult to test I/E and J/P but I think it's possible.

I think under the current system, someone might test as say an S, but also has a very strong N, but because of the way they were tested, or specifically the questions were asked they tested as an S, when in fact they may test signicantly higher with intuition than a lot of N's do.

I'm not saying I am going to make this test up, but I think it would be interesting to see the results of a test like that and then determine your type from there. You would also get a clearer look at you complete results, rather than just half a picture of your results.

Thoughts?

[coffeezombie]

I think the test should be more like an IQ test. It should test each function individually and attempt to measure your capacity in that function and determine your type from that. This might be somewhat more difficult to test I/E and J/P but I think it's possible.


A lot of people got unorthodox orders for their functions when taking that test that determined function order. I'm not sure how getting an unorthodox order for one's functions would translate into what particular MBTI type you are, then.

[Trolsk]

The weakest link to any online test will always be the test taker. I've seen two methods worth mentioning.

The Socionics Type Assistant takes an approach where you're basically filling out a spectra of qualities that are descriptive to your behaviour. There are follow-up questions when the score is inconclusive.

There is a test at tap3x.net which attempts to verify a dominant function and gives as a score a (sorted) list of likely functional orders, with MBTI-type matching (this is because order doesn't matter much for the MBTI.)

[mgb]

A lot of people got unorthodox orders for their functions when taking that test that determined function order. I'm not sure how getting an unorthodox order for one's functions would translate into what particular MBTI type you are, then.

It wouldn't be so much about order, I think you could still have someone listed as INTP or ESTJ or whatever, but you could also have a measure of what T is, rather than a measure of T as opposed to F.

I think when people list their percentages of their functions, it's a relative term, relative to how F or S they aren't. I think that's wrong. I think someone could be both a high T and a high F. Their type would be based on which type description suits them most combined with the test scores.

As for online testing, I agree. I think this other type of test could be used in person as well.

[Hypnos]

floyd has made the case for the Global 5-SLOAN (http://similarminds.com/sloan.html) system.

[garak]

Big 5/SLOAN is definitely interesting.

[mgb]

floyd has made the case for the Global 5-SLOAN (http://similarminds.com/sloan.html) system.

I think that is just a different way of doing the same thing that MBTI does.

I'm not writing off MBTI, just the testing methods, because I don't think they adequately test for type.

I do agree with the haziness 5-SLOAN applies to typing. I don't think any person is any type to a "TEE". But it does suffer from the same one or the other in typing, undiscivered grey areas aside.

[Hypnos]

Jung typology is based on two premises:

* There are 4 faculties -- T, F, N, S and general introversion/extroversion and perceiving/judging preferences

* In a person they are architected with the following rules:
- For an introverted person, the primary function is introverted
- For a perceiving person, the strongest perceiving function (N/S) is extroverted
- The weakest function is the opposite of the dominant function (in INTPs, Ti vs. Fe)

SLOAN has a much simpler premise: what traits of a personality can be reliably measured, are independent, and predict behavioral preferences most clearly?

So it's a question of full-blown personality/psychological theory v. empirically sound measurement of personality without making a statement on architecture.

[coffeezombie]

floyd has made the case for the Global 5-SLOAN (http://similarminds.com/sloan.html) system.

Did he hire you for PR?

[Hypnos]

Did he hire you for PR?
Yes, he's paying me in Immortal Future Bucks

[mgb]

Jung typology is based on two premises:

* There are 4 faculties -- T, F, N, S and general introversion/extroversion and perceiving/judging preferences

* In a person they are architected with the following rules:
- For an introverted person, the primary function is introverted
- For a perceiving person, the strongest perceiving function (N/S) is extroverted
- The weakest function is the opposite of the dominant function (in INTPs, Ti vs. Fe)

SLOAN has a much simpler premise: what traits of a personality can be reliably measured, are independent, and predict behavioral preferences most clearly?

So it's a question of full-blown personality/psychological theory v. empirically sound measurement of personality without making a statement on architecture.


I agree with the premise, but I didn't get that from the test. I think there are still pitting the functions against each other rather than testing them individually. It might just me similar minds test, but that's how it seems to me.

Imagine instead if you could test someone's capacity to think. Then the next question could test their capacity to feel, either through test questions or scenarios. But for all of the functions. Through a series of questions like those, you could discover the strengths of the preferences on their own merits, rather than having one or the other.

Look at this question from the Global-5 test:

40) I am not sympathetic of the feelings of everyone.

Very Inaccurate . . . . . Very Accurate

If you answer Inaccurate, you are more of a feeler, or in this case maybe Accomadating (just a guess). If you answer Very Accurate you are more Egocentric.

In that way Global-5 isn't really doing anything different than a lot of online tests.

[eyebyte_atWork]

Where did you people get tested??

btw - I tested from Keirsey's book.

[Hypnos]

If you answer Inaccurate, you are more of a feeler, or in this case maybe Accomadating (just a guess). If you answer Very Accurate you are more Egocentric.

In that way Global-5 isn't really doing anything different than a lot of online tests.
Except, that accommodation and egocentricity really are mutually exclusive, unlike thinking and feeling.

And what do you mean by "online tests?" SLOAN (also called Big 5 or OCEAN) have been developed by the academic community.

[mgb]

Except, that accommodation and egocentricity really are mutually exclusive, unlike thinking and feeling.

And what do you mean by "online tests?" SLOAN (also called Big 5 or OCEAN) have been developed by the academic community.

I don't know, there was a test on similar minds. It said "click here" and I did, then a test popped up.

It might not be authorized but it was from the link you gave me.

[Hypnos]

Right, so in your example, it does examine the accommodation-egocentricity scale, which is part of the model. This, introversion-extroversion, organization, etc. are all valid oppositional scales unlike the axes in Jung typology.

[mgb]

Right, so in your example, it does examine the accommodation-egocentricity scale, which is part of the model. This, introversion-extroversion, organization, etc. are all valid oppositional scales unlike the axes in Jung typology.

I'm mostly just talking about the testing, not so much the exact preferences themselves. I think you could substitute in accommodation/egocentricity for thinking/feeling, in the test I envision (obviously changing the questions and answers to match the corresponding type theory).

I think that introversion/extroversion and perceiving/judging might work well in opposition, but I'm not sure that sensing/intuition and feeling/thinking should be considered oppositional scales. I'm not convinced that a gain in one of those preferences comes at the expense of it's opposite.

[garak]

... but I'm not sure that sensing/intuition and feeling/thinking should be considered oppositional scales. I'm not convinced that a gain in one of those preferences comes at the expense of it's opposite.
I don't think they're supposed to be seen that way anyway. It's about which you prefer. I can't imagine a person who thinks intuitively just as much as [s]he thinks sensor..uh..ly. After I figure someone out, it tends to be pretty obvious to me that they prefer one over the other. Maybe I'm delusional though.

[mgb]

I don't think they're supposed to be seen that way anyway. It's about which you prefer. I can't imagine a person who thinks intuitively just as much as [s]he thinks sensor..uh..ly. After I figure someone out, it tends to be pretty obvious to me that they prefer one over the other. Maybe I'm delusional though.

I agree, I think most tests put them in opposition though. I just think it makes the tests less accurate and doesn't give someone a true sense of their type.

[Hypnos]

I'm mostly just talking about the testing, not so much the exact preferences themselves. I think you could substitute in accommodation/egocentricity for thinking/feeling, in the test I envision (obviously changing the questions and answers to match the corresponding type theory).
I disagree. You can ask about thinking and feeling separately, but not about how accommodating/demanding you are of others.


I think that introversion/extroversion and perceiving/judging might work well in opposition, but I'm not sure that sensing/intuition and feeling/thinking should be considered oppositional scales. I'm not convinced that a gain in one of those preferences comes at the expense of it's opposite.
I agree.

[rmcnew]

There is a test I know of that test by individual attribute and also by function, but it is based on socionics and not MBTI ...

Have a look yourself ...

http://socionics.wsphere.com/socionicstypetest.html

[INTrPosr]

I think the test should be more like an IQ test. It should test each function individually and attempt to measure your capacity in that function and determine your type from that. This might be somewhat more difficult to test I/E and J/P but I think it's possible.Excellent thoughts Mgbradsh. I am not sure that anyone can ever devise a personality test without the testee's immediate emotions dictating answers. That is why many authors tell you to not just test but validate your results. Do you have any suggestions on types of questions to ask?

[iponjs]

The weakest link to any online test will always be the test taker. I've seen two methods worth mentioning.

The Socionics Type Assistant takes an approach where you're basically filling out a spectra of qualities that are descriptive to your behaviour. There are follow-up questions when the score is inconclusive.

There is a test at tap3x.net which attempts to verify a dominant function and gives as a score a (sorted) list of likely functional orders, with MBTI-type matching (this is because order doesn't matter much for the MBTI.)

N=11 S=4
F=10 T=12

So, T N F S - I totally agree with the order of preference and it does help explain my F-ishness when compared to other INTPs... And the fact that I despise S activities...

I've said it before - MBTI data is suspect to me because I have typed as an ESTJ, ISTJ, INTJ, and INTP at various stages in my life. All of that data is part of the CPP data banks. How many others have the same problem? I think INTPs in general are more likely to mis-type (analysis paralysis).

At the end of the day, it is just a questionairre. Helpful but nowhere near 100% accurate. THat's why we do self-selection workshops to back it up.

[s0978]

I agree, I think most tests put them in opposition though. I just think it makes the tests less accurate and doesn't give someone a true sense of their type.
What would the ideal test accomplish? More detailed or precise personality profiles? More than 16 types? How might the profiles be different if we had an alternate (let's say "more accurate") set of underlying assumptions about the cog processes?


edit: or is this all just in the interest of more accuracy for people who are mis-typed... like I might be a esfj who is so dumb she thinks she is an intp? ha ha... but there have been quite a few people who come through and say , 'oops, psyche, I'm no intp, I'm entp,' etc.

[s0978]

Except, that accommodation and egocentricity really are mutually exclusive, unlike thinking and feeling.

True, but I don't know that it makes it any easier to measure, either.

In the case of
40) I am not sympathetic of the feelings of everyone.

Very Inaccurate . . . . . Very AccurateI might choose "very inaccurate" because I think I am typically so considerate, but my mom would beg to differ. The question tests for a self-concept rather than a preference... not that this is necessarily problematic... just thinking out loud.

[mgb]

Excellent thoughts Mgbradsh. I am not sure that anyone can ever devise a personality test without the testee's immediate emotions dictating answers. That is why many authors tell you to not just test but validate your results. Do you have any suggestions on types of questions to ask?

Images might be a good way. If testing in person you could have a person write in their own thoughts/feelings about a picture. On the internet you could list several options. I think a thinker and feeler might have what appears to be a similar reaction to a picture, but how they express themselves might be very different.

[coffeezombie]

I think the "emotions" thing could be mostly solved just by taking a test at multiple times, and then going with what the most common result is.

[mgb]

I think the "emotions" thing could be mostly solved just by taking a test at multiple times, and then going with what the most common result is.

Is thinking and feeling more of a difference in the way we express ourselves though? I mean, even thinkers can feel sympathetic or get angry, feel hurt, or be really happy. I think that they will act/say/think very differently from a feeler in the exact same situation.

[Ka.avik]

can't imagine a person who thinks intuitively just as much as (s)he thinks sensor..uh..ly. Sensationally!!

Lenore Thomson has some choice words about people who appear to test equally strongly between the two choices -- either T/F or S/N. Basically it's not your inclination, and only very uncomfortable circumstances can force you to 'catch up'.

Or you've just left your secondary so far behind that it's back there somewhere with numbers 3 & 4 ....

But it is my policy to never let a setup go past unused. Yes, Mack -- I think you're delusional. :)

[coffeezombie]

Is thinking and feeling more of a difference in the way we express ourselves though? I mean, even thinkers can feel sympathetic or get angry, feel hurt, or be really happy. I think that they will act/say/think very differently from a feeler in the exact same situation.

Yeah, I've argued this before on other boards when it came to Enneagram tests, and have always advocated creating "situational tests," coming up with an imaginary situation and asking the person what their response would be, rather than just asking them what they think about a simple statement. I got the idea from Ultima IV, actually, the way it tests to determine what kind of character you are at the start of the game. I came up with a couple of "sample questions" for people on that board in order to help them determine their type and had pretty successful results after some tweaking, but have yet to make a full-scale test.

[Hexchild]

If you provide the right questions and answers, I'll gladly provide the quiz engine to try them out. I wrote one a few years ago. At the time I tried to get a few test pilots from the lesser INTP mailing list but I didn't get enough volunteers, so this project was put on ice.

For now, feel free to try out this "test quiz":

http://83.250.33.159/cgi-bin/opaque/opaque.cgi/shallow

EDIT: Link updated to reflect my current IP address (as of Oct 2, 2005)


I've completely reset the test results, so the first person to take this quiz, and whoever takes it before the statistics for those answers get stored in the database, won't see any results at all. The database is updated every 10 minutes, so if you go back 10 minutes later to check your results, you should see them.


Consider the above link a temporary one, which you'll have to use until I get my Internet connection straightened out.

[INTrPosr]

I think the "emotions" thing could be mostly solved just by taking a test at multiple times, and then going with what the most common result is.How does that differ from what people generally do with the test presently?

[Hexchild]

By now a few people have taken the OPAQUE test quiz (although not nearly enough to make any useful analysis from).


So far, the statistics engine has come to the conclusion that you're an INTP (and also a rational/NT):

- if you are male
- if you are between 10 and 19 years old, probably not if you are 20-29
- if you prefer blue to black
- if you are tough-minded, probably not if you are sentimental
- possibly, if you sit in the corner at parties
- possibly, if you rely on hunches


There are similar conclusions for the e5 (Enneagram Thinker). The engine is currently very doubtful that anyone belongs to any of the remaining categories (it will remain so until proven wrong).

[mgb]

By now a few people have taken the OPAQUE test quiz (although not nearly enough to make any useful analysis from).


So far, the statistics engine has come to the conclusion that you're an INTP (and also a rational/NT):

- if you are male
- if you are between 10 and 19 years old, probably not if you are 20-29
- if you prefer blue to black
- if you are tough-minded, probably not if you are sentimental
- possibly, if you sit in the corner at parties
- possibly, if you rely on hunches


There are similar conclusions for the e5 (Enneagram Thinker). The engine is currently very doubtful that anyone belongs to any of the remaining categories (it will remain so until proven wrong).


I looked at a few questions.

No offense because I think you are trying to help, but this is a good example of the types of questions that should be avoided. But I also realize it's a sample test.

The type of questions I envision would be a great deal more complicated. Four or five possible answers and questions that test for only one preference.

I'm not sure how to do it right now because there are too many people in the apartment and driving me nuts, but I'll work on some possible questions.

[Hexchild]

No offense because I think you are trying to help, but this is a good example of the types of questions that should be avoided. But I also realize it's a sample test.

No worries, no offense taken. I'm quite aware that the questions in the test quiz are far too biased and minimalistic for a serious personality test. I am testing the engine rather than the quiz itself, after all.

By the way, I threw in the fave color question as an example of an irrelevant question (I sure hope it is anyway) to see how well the stats engine would handle that. It's far too early to tell yet, though. I expect it to start balancing out once all categories have been covered maybe three times each or so.



The type of questions I envision would be a great deal more complicated. Four or five possible answers and questions that test for only one preference.

I agree that we need more complex questions (that's what you mean, right?), especially multiple answers. I also agree with you that there should be questions that test for a single preference.

If we're going to use my engine, then there should also be questions based on preference combinations of varying complexity, and there should be some very simple questions (but still with as many answers as is reasonably possible). It was built with that sort of variety in mind. The more questions we ask (within reasonable limits) and the more varied they are, the better the end result.

[mgb]

No worries, no offense taken. I'm quite aware that the questions in the test quiz are far too biased and minimalistic for a serious personality test. I am testing the engine rather than the quiz itself, after all.

By the way, I threw in the fave color question as an example of an irrelevant question (I sure hope it is anyway) to see how well the stats engine would handle that. It's far too early to tell yet, though. I expect it to start balancing out once all categories have been covered maybe three times each or so.




I agree that we need more complex questions (that's what you mean, right?), especially multiple answers. I also agree with you that there should be questions that test for a single preference.

If we're going to use my engine, then there should also be questions based on preference combinations of varying complexity, and there should be some very simple questions (but still with as many answers as is reasonably possible). It was built with that sort of variety in mind. The more questions we ask (within reasonable limits) and the more varied they are, the better the end result.

I'd say about 100 questions is probably on the high side, but if we make a test that's more accurate, I think it would be worth it.

Can you incorporate images into the test?

What if you had a picture of say, a homeless family, and you test feeling by varying the degrees of sympathy in the answers.

a) That's too bad, I hope the kids are eating.
.
.
.
e) That makes me so sad that people have to live like that.

In this way I think we can build in questions to test say, just the feeling function.

To test thinking (which I think we should try and define) we could even steal some IQ quesitons (although I'm not so sure that is exactly what we are looking for). Although, I think we can provide enough variety to have a few different types of questions.

This is all assuming you are ok with this and up to the challenge and hopefully more people want to add to this.

[Hexchild]

I'd say about 100 questions is probably on the high side, but if we make a test that's more accurate, I think it would be worth it.

Certainly. But then, that probably requires a bit of research.

It is quite possible with the OPAQUE engine to allow the testee to stop answering questions and go directly to the results page at any time. All unanswered questions are completely ignored by the statistics engine (but won't contribute to the results either, of course). Also, if we add some questions at a later date the resulting statistics will be the same as if all previous quiz takers had elected not to answer those questions. The same principle applies to removing questions.



Can you incorporate images into the test?

Anything that works in HTML/DHTML can be used (even Javascript code if need be).



This is all assuming you are ok with this and up to the challenge and hopefully more people want to add to this.

I'm certainly interested in being involved in a project like this, but I'm hardly an expert on psychology, so although I can probably come up with quite a few theories and/or questions, I'm not sure they would be relevant to the results. I don't mind giving it a go, though.

I'll gladly make any necessary adjustments to the engine as long as they are reasonable (I doubt we need any nontrivial changes anyway) and I'll be involved in making sure the quiz content is configured correctly. I could easily host the quiz as well, but my internet connection (8Mbit download / 1Mbit upload) is probably not ideal for serving a public test.

[Hexchild]

I've just fixed a bug in the probability calculation code that caused notable bias on the results page. The test results should make a lot more sense now.

This bug did not affect the database contents (just the presentation of them) so your statistics are safe :)

[wildcat]

It wouldn't be so much about order, I think you could still have someone listed as INTP or ESTJ or whatever, but you could also have a measure of what T is, rather than a measure of T as opposed to F.

I think when people list their percentages of their functions, it's a relative term, relative to how F or S they aren't. I think that's wrong. I think someone could be both a high T and a high F. Their type would be based on which type description suits them most combined with the test scores.

As for online testing, I agree. I think this other type of test could be used in person as well.
You are right, too.

[INTrPosr]

The weakest link to any online test will always be the test taker. Enough said, you are 100% right. Has anyone considered whether they have a pattern of when they usually go test? I know for moi, it's usually when I am agitated and my emotions are out of whack. I admit when taking the MBTI II, I took it at home, away from everyone and in a quiet place. I thought that I honestly answered the questions, however now realize that there was a lot going on with me at the time. I think the best validation of true type will be from self, combined with those who know you best.