Anyone want to help make socionics scientific?

[Lao Tzunami]

@golden, are you talking about how any psychometric test, like let's say a 1-5 scale likert test (strongly agree, agree, neither, disagree, strongly disagree) people have a tendency to rate themselves in the middle regardless of the questions? I remember My statistics teacher mentioned that and said you should never give people an odd numbered option for that reason. That all has to do with experiment design. I don't know what best item scheme would work for socionics - that's something we'd have to figure out. That's one reason I think we need to evaluate socionics statistically, so we can measure small incremental improvements.

[Exodus]

@golden, are you talking about how any psychometric test, like let's say a 1-5 scale likert test (strongly agree, agree, neither, disagree, strongly disagree) people have a tendency to rate themselves in the middle regardless of the questions? I remember My statistics teacher mentioned that and said you should never give people an odd numbered option for that reason.

That seems silly. If you genuinely can't rate yourself as being one or the other, you're forced to arbitrarily choose one. Good way to get bad statistics.

[golden]

@golden, are you talking about how any psychometric test, like let's say a 1-5 scale likert test (strongly agree, agree, neither, disagree, strongly disagree) people have a tendency to rate themselves in the middle regardless of the questions? I remember My statistics teacher mentioned that and said you should never give people an odd numbered option for that reason. That all has to do with experiment design. I don't know what best item scheme would work for socionics - that's something we'd have to figure out. That's one reason I think we need to evaluate socionics statistically, so we can measure small incremental improvements.

No. I’m not talking about problems in making scales or in research subjects using the scales. I’m talking about the proposal that for any given personality trait, the strength of that trait will show normal distribution. My understanding — based on reading I did several months ago — was that it’s been consistently found that traits distribute this way. Therefore, if you use a dichotomy, you may slice across the center of the curve where the greatest number of people are clustered.

[squark]

No. I’m not talking about problems in making scales or in research subjects using the scales. I’m talking about the proposal that for any given personality trait, the strength of that trait will show normal distribution. My understanding — based on reading I did several months ago — was that it’s been consistently found that traits distribute this way. Therefore, if you use a dichotomy, you may slice across the center of the curve where the greatest number of people are clustered.

Yes, that's true. For example regarding introversion/extroversion as a trait, most people are actually in the middle. Very few people are on the ends of extroversion or introversion, and it does form a normal bell curve. Other traits are similar, and it's one reason why I'm finding Big 5 very interesting to look into, because it doesn't assume there's a division, but instead looks at each trait along a scale, which is more normal and true-to-life. Dichotomies are a shitty way to describe people and the biggest downfall imo of MBTI because most people actually should be between types than solidly in any. (It's probably also why the vast majority of people test as a different type each time they take the MBTI)