Open milahu opened 4 years ago
This is just some theory that puts everything into one scale. You can also put all questions in yin and yang for example.
But you can get for example more information if you make additional axis on authority, because you put communism on the conservative site which is only true when you want authoritarian government. At the same time there are people who are liberal but are opposed to technologies like people for environmental protection.
All 8 scales give additional information where you can be extreme in one scale without being in your scale extreme.
you can get for example more information if you make additional axis
yes, but the axis must be orthogonal.
my claim is, the four axis ( econ / dipl / govt / scty ) are much too similar, and do not add much/any additional info
here is a plot of the weight-factors on a 4D coordinate grid: the problem is, most red dots are on the same diagonal
my solution to "four orthogonal dimensions" is
"( econ / dipl / govt / scty ) are much too similar" more precise, the axis pretend to be linearly independent [orthogonal], but the questions do not reflect that. [ to make this test better, we need better questions. ]
just have a look at the weight-factors for ( dipl / govt / scty ). for all questions, they have the same sign ( positive / negative ), with only one exception:
{
"question": "Even when protesting an authoritarian government, violence is not acceptable.",
"effect": {
"econ": 0,
"dipl": 5,
"govt": -5,
"scty": 0
}
},
.... which is a bug in my view, both factors should be negative.
conservatives only do peaceful protests, or passive protest like labor strikes. liberals can have violent insurgencies / revolutions. in conservative culture, violence always is "from above", from police or military. in liberal culture, the majority of citizens have guns, and the weak government is under pressure.
This is just some theory
this whole test in its current version is just a wild theory, from somebody who is obviously untrained in the art of factor analysis.
when you study the MBTI test, for example, 50% is redundant
the questions only appear to be different, cos they use different words and phrases, but if you happen to master the art of translation and abstraction, [ im afraid, only fire is good at this, so only a minority will truly understand ] you easily notice the strong similarity.
outer values | hans eysenck | MBTI 1+2 | MBTI 3+4 |
---|---|---|---|
young age | extravert | iNtuition | Perceiving |
old age | introvert | Sensing | Judging |
feminine gender | neurotic | Extraversion | Feeling |
masculine gender | psychotic | Introversion | Thinking |
the only benefit of redundant dimensions is, you can measure how well the test questions were understood. when similar questions are answered consistent, understanding was good. when similar questions bring opposite answers, understanding was bad.
understanding seems to be more bad for "old age" types, earth and water. these are anti-sensitive, have poor self-value, tend to have false self-images
Thank you for providing all these details. I'm not sure if I misunderstood some things but to me it looks like that you want to make a personality quiz.
I think this quiz is meant to be a political quiz. Maybe one can argue that political values are reflected from the personality but it doesn't mean that people with the same personality can have very different political values.
So maybe you want to say that this quiz is mostly redundant if you want to measure the personality of someone. But for measuring the political values this is not the case.
Which political values do you think are less redundant than the 8 chosen ones? Or do you think most politics can only be devided into conservative VS. liberal?
I'd like to add here that there is some interesting critique of some redundancy in the questions. However, each of the axis has a number of questions which only affect that axis, so it is certainly possible to achieve a unique result on each axis. In other words, although there are questions which affect more than one axis (and which have a tendency to affect those axes similarly across the quiz) there are enough single-axis-affecting questions to allow unique results.
It's also worth noting that the entire reason 8values exists is as a counter to the reductive notion that politics is a single axis, and to provide additional nuance to the often of the left and right wing that other political tests struggle to capture (see political compass).
That said, there is definitely merit in the factor analysis @milahu has done here, and if you can suggest how the test might be improved to reduce redundancy and increase linear independence in the 4 axes, I am all ears and I would love to hear those suggestions.
Actually, that's why having a dataset would be useful. Is the model simplifiable? Doing a PCA or other dimensionality reductions of the results ( or even on the questions themselves ) Would really help finding that
The test is clearly only appicable to US, as progressive vs conservative are terms subjective to country's tradition and can contain different ideologies in different countries.
Certainly, the test is written from and targeted at a western perspective; however, I disagree that it is not applicable elsewhere than the US; I for one am not from the States and yet the 8values test is perfectly applicable to my politics and the political spheres of the country I reside in.
redundancy is a problem with many tests, like MBTI or big-five
in my understanding, all the 8value questions only measure one single scale:
in terms of Milton Rokeach and the classical four elements and Hans Eysenck two-factor terminology (extraversion + psychoticism)
in a two-factor-model with the two orthogonal factors equality and freedom, the two conservative types [fire + earth] share the same diagonal, and the two liberal types [air + water] share the other diagonal.
in theory, there are five types of questions, five scales to measure:
the current 8values test pretends to give four scales, but really gives just one scale.
edit - related to issue #69