Subject:
|
Re: Personality test vs. Religion
|
Newsgroups:
|
lugnet.off-topic.debate
|
Date:
|
Tue, 2 Nov 2004 19:06:54 GMT
|
Viewed:
|
2692 times
|
| |
|
|
In lugnet.off-topic.debate, Dave Schuler wrote:
> But I sense that your underlying question is this: If a non-scientific
> system can be shown to have a great deal of predictive accuracy, would it not
> be valid to make use of that system? I guess it would, but so what? Show me
> such a system in the real world, and we'll discuss it.
Awesome. The only question that remains is "how much of a 'great deal' of
accuracy is needed?", and to show whether or not M/B does or doesn't follow.
> To be fair to both of us, the intent of my argument has been to demonstrate
> that the test is of no value in making claims re: professional potential. If
> we both accept that point, then I'll stop belaboring it.
I think we've agreed on that from the start-- I would argue that while SOME job
potentials might be slightly benefitted from the traits measured in MBTI, the
traits that are *truly* needed for most (if not all) jobs are NOT measured by
MBTI.
> However, I haven't really focused on whether or not it's good at spotting
> trends. I would say that the test, when used as it is designed to be used,
> has no greater ability to identify trends than normal interpersonal
> interaction. That the test applies jargon-eque acronyms makes no difference;
> it's the same as making the casual observation that "Bob's a jerk" or "Bob's
> a great guy."
Ahh, excellent! I would agree. The test is in fact made up of queries based on
personal/casual observation, hence I would argue its usefulness is only as good
as the observations made by the person taking the test; which is why it's error
prone.
> Actually, casual observation is in fact more useful than the test. The test
> will only return results like "Bob's a great guy," "Bob's a swell guy,"
> "Bob's a nice guy," "Bob's a fine guy," etc.
I think that's a matter of interpretation. The test DOES try and put a positive
spin on things, but regardless, I personally would count "feeler" and
"judgemental" as negatives, when in extreme.
But really, I think you're expecting too much out of the test. It's closer to
being a test for, say, "blond/black hair" than "ugly/pretty hair". It's intended
to measure things that aren't necessarily good or bad, just different. Now, some
of the categories some people might interpret as good/bad (myself for instance),
but it's the goal of the test to try and make them sound better so as not to
offend. And personally (though sadly), I do think that IS the right way to go,
but wouldn't be if people were better capable of accepting their own faults.
> > Again, I keep bringing this up, it IS possible (I think) to get a negative
> > result. If it scored me as an ENFJ instead of an ISTP, that (I would hold)
> > would be a negative result for me.
>
> That's true only if you change the nature of the test, and then we're not
> arguing about MBTI anymore, are we? Built into the test is the ability to
> massage the results to get them to match reality, and this is the fatal flaw
> I mentioned previously. If you take away this post hoc tweaking, then you
> haven't completed the test as it is designed.
Ok, forget the self-change then-- suppose I rated myself as an ISTP (after
tweaking results), and 100 psychologists independantly rated me as an ENFJ? I'd
still say that's a negative result, and as such, would make the test
falsifiable.
> > > I think that's how it was meant--there should be a bell-curve distribution
> > > of each trait, rather than of each four-character personality descriptor.
> >
> > If that's how he's meaning it, he'll have a hard time doing it without
> > incremental levels per dimension. I mean, if the bell curve is smack dab in
> > the middle for one trait, you'll get a 50/50 distribution in your two
> > categories,
>
> I think that we're thinking of this differently. The way I'm breaking it
> down, the critic is expecting a bell-curve representation for each of the 16
> (or hypothetical 625 or 1,089 or whatever) traits. Is that how you're
> reading it? I have the sense that you're basing your (pixellated!) curve on a
> strict binary reading; EITHER Intravert OR Extravert. I don't think that
> you're wrong to do that, but I don't think it's what the critic's describing,
> either.
I'm not really sure-- I'm basically saying that:
- if it's a 50/50 valuation, you're not going to get any sort of normalization
anyway.
- if it's based on the 16 types, the author already showed that normalizations
existed by showing that there were 6 generally found types, and that 12% (not
1/16 or 1/6, but about 1/8) actually fell into a particular type.
> Now we're getting somewhere, but I think that you should describe it as
> "testimonial evidence" instead of "empirical evidence." At best, the test
> provides *inferential* evidence of correlation. The leap from a subject's
> response of "open" or "close" to deciding that the subject is therefore
> "Intravert" or "Extravert" cannot be corroborated except by testimony and/or
> inference, and this is dangerously close to wishful thinking.
I definitely agree with the testimonial evidence part-- which is what I think
pretty much the entirety of psychology is based on. Subjective evaluation of
testimonial evidence; not empirical data. Which is why I've been repeatedly
flogging the term empirical-- basically, how can you believe anytyhing in
psychology if you disregard all but the empirical?
Is the leap to introvert/extrovert substantiated? I would guess so, but not to
any perfecting degree. I can't say for sure, but I would assume that many tests
have been done on M/B before its acceptance, by people rating themselves, and
seeing how close they match "educated opinions". I would also assume that
questions that gave better approximations were retained over questions that
yielded less reliable results. Back to the tea leaves? I'd think so. You can't
prove that the conclusion will always be accurate, all you can say is that it's
got a good track record.
> > And as such, you're right, it only invalidates
> > the test portion moreso. But it does nothing to the credibility of the
> > dimensions themselves.
>
> But, again, the dimensions can only be assessed by inference rather than
> correlated empirically to the test or the test subject, and to that end the
> MBTI is of no real value even in simply identifying those dimensions. You
> can easily apply those dimensions (or approximations of them) to people by
> yourself without need of a test to guide you.
Yep! I'd agree. What the test does for me in this case is provide a collection
of data for analysis. If nobody took the test, and just evaluated themselves on
their own, we'd have no common ground to compare and contrast people. Now,
admittedly, the tool has flaws, but it's better than nothing. And certainly, in
general, most people are the most qualified at making such judgements about
themselves (short of perhaps psychiatrists in some instances). So if you want a
valid assessment, don't use the test. But the test IS useful to people like me
who are interested in seeking trends which would not otherwise be apparent.
> A committee of psychologists could get together and say "A person who
> demonstrates traits A, B, C, D, E, F, and G can be described as "stubborn."
> Thereafter they can assess a subject and determine whether or not the subject
> is, according to the scale, "stubborn."
But you would accept that a trait called "stubbornness" exists, and would feel
relatively confident in your assessment of, say, me, as being stubborn, assuming
you had made that assessment?
> Where does that leave us? Well, the MBTI folks decided that people who give
> a particular range of answers can be described as an Intravert. This label
> may or may not intersect with the mainstream definition "introvert." If it
> does intersect, then proponents of the test will claim that it's accurate.
> But if it does not intersect, then proponents of the test will claim that
> results can vary. This is, of course, a demonstration of non-falsifiability,
> which demonstrates that the test doesn't provide useful answers except by
> inference.
Again, I may be assuming too much, but I would guess that its general acceptance
as a tool has in part come from its coincidence with independant evaluation.
> > What would have to be done to convince you that I'm stubborn when it comes
> > to, say, my beliefs in relative morality?
>
> Why go there? You're already demonstrated to me that you're stubborn re: the
> MBTI! 8^)
Heh, this is nothing :) At least here I'm willing to be wrong if there really
looks to be data in that direction-- Relative morality is much more fun, cuz
there isn't any data anyway, hence no data that can change my mind!
> I don't know that I can give a clear articulation of the way I determine that
> someone is stubborn, or generous, or gloomy, or charismatic, etc. These are
> subjective terms that do not lend themselves to concise definition. In fact,
> that's part of the reason I reject the MBTI--it seeks to give soundbyte
> labels to complex systems of personality.
But let's say, for a moment, that someone invented a test that judged
stubbornness, and it matched your prior assessments on 50 people at 84%
accuracy?
> Well, let me ask it this way. If a friend of yours took the MBTI and his
> answers revealed him to be a total jerk who really doesn't like you,
As you correctly point out earlier, however, it won't point that out, since it
doesn't measure jerkiness or what sorts of people they like-- no negatives :)
> would you stop hanging out with him? I'd hope not, because then you'd be
> subordinating your own interpersonal assessment to an arbitrary and (frankly)
> poorly-founded instrument.
Certainly-- because I value my own personal assessments above the test's. And as
I've said, it's DEFINITELY capable of making mistakes.
> The point here is that the test is seriously
> impaired by its shortcomings, and I see these shortcomings as fatal, while
> you do not.
I think perhaps its the view on the accuracy or relevance of the personality
dimensions where we differ. I think it's generally accurate (maybe 75% or so),
and that each of the dimensions is helpful and distinct. You seem to either
think that the dimensions are completely useless (except insofar as they make
you feel better), and/or that the test is no better than random at making that
determination.
> But you're risking a mistake to ask this list of me. You're getting very
> close, intentionally or not, to claiming that, if I can't propose a better
> system, then I have no basis for objecting to this existing system, and
> that's a fallacy. I believe that this isn't your intent, but it's only a
> little further down the road in the direction you're currently heading.
Sorta, but not quite. My intent was to say that if you accept "depression" or
"cheerfulness" or "moron" or "stubborn" as a valid qualifier, why not the MBTI
dimensions? If you feel that you could rate someone as stubborn/not stubborn,
why not thinker/feeler? Why not introvert/extrovert? Which (if any) MBTI
categories do you feel are valid? How are the MBTI different from "stubbornness"
which assumably you allow to be a valid character trait?
> Being "off by a letter or two" is the same as being wrong, especially when
> you're talking about a field of only four letters! To suggest that 25% or
> 50% inaccuracy is acceptable is to succumb to the practice of "shotgunning:"
Not necessarily. Again, my own problem with the test is this idea of 50/50,
which I think should be a gradual scale. But let's say for a moment that someone
measures a 5/10 on I/E, and there's an accuracy margin of +/-3. Theoretically
half the time they take the test they'll go one way, half the time the other.
The real test with the presented data would be to see which traits flip
consistantly in which people. If someone takes the test 10 times, and comes out
with:
I N F P
I N F P
E N F P
E N T P
E N F P
I N T P
I N F P
E N F P
I N T P
E N F P
It looks to me like they're almost assuredly ?N?P, close to halfway on I/E, and
borderline F/T, closer to F. The test's ability, however, to measure N and P
consistantly show that at least it's testing SOMETHING with more consistancy and
accuracy than the 50% accuracy attributed to the test by the article you
presented. Compare if instead they got:
I N F P
I S F J
I S T P
E S F J
E N T J
I N T P
E S T P
E N F P
E N F J
I S F J
Here EACH of the categories is all over the place, even though between
successive runs there's only a 2-letter difference. Sure, it's possible that
this person is 50/50 on each category, but it's not helping test's reputation as
a dataset. If THIS is the sort of data that's coming back repeatedly, I'll put
less faith in the MBTI. As it is, I don't think that's true.
However, as I've suggested, a scale helps a lot. You might have a scale of:
Ii-eE, Nn-sS, Ff-tT, and Pp-jJ wherein you might see that 1st person as:
i N f P
i n f p
- N - P
e N t P
e n f P
- N - p
i N f P
- N f P
- n - p
e n - P
And suddenly it's not the same error percentage as it was before, but is clearly
milder. On the other hand, if it revealed:
i N f p
I n F P
E N - p
- N T p
E n F P
I N t p
I n F P
e N f p
- n T P
E n - P
Then its ability to measure E/I and F/T is clearly in question, while its
ability to measure N/P would still be pretty valid.
DaveE
|
|
Message is in Reply To:
 | | Re: Personality test vs. Religion
|
| (...) I'm sorry, but this hypothetical example doesn't interest me. Suppose I posit a car that delivers infinite gas mileage--wouldn't you buy it? Heck, yes! But what's the point? It's not difficult to create examples that have no relation to (...) (20 years ago, 2-Nov-04, to lugnet.off-topic.debate)
|
53 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
