Why another test?
Many existing personality models focus on dichotomies - introversion vs extroversion, feeling vs thinking and classify people as strictly falling on one side of the spectrum. While there is some truth to that, people generally don’t exhibit such clear division of abilities, most are in the middle of such scales. Hence these dichotomies can come across as arbitrary division of people.
Also, most of them rely on behaviour based questionnaire, which assumes that people have fixed behaviour regardless of their current occupation or societal demands. This is hardly the case. Further, the questions are not so well defined, resulting in people who take the tests to interpret them according to their understanding. These questionnaires then resort to asking multiple questions for the same aspect to average out these aspects, but most of the time they don’t work that way. Instead, the errors gradually add up and in the end they don’t measure what they intend to measure. And the errors are large enough to show up as statistical anomalies, even if we don’t know the personality types of everyone who took these tests. We will list here some of the glaring anomalies that are found with these tests.
Lack of retest reliability
This is the single most issue of all the previously existing tests. If these tests measure something inherent, like one’s personality the results shouldn’t change if the test is administered again the next day, next month or next year. But this is not the case with most of the existing tests like MBTI, they produce the same result in just over 60% of the cases. This is not the measure of accuracy, which could be even lower.
Statistical support for the distribution based on dichotomies
If people can be divided into dichotomies based on their scores, according to statistics the distribution of scores should follow what is called a bimodal distribution - a distribution with two peaks than a normal distribution, with only one peak. However the test results fail to confirm this observation, as the scores follow only a normal distribution and not a bimodal one. This also means that a cut-off has to be relied for splitting the distribution into two, and this generally runs through the middle, where most people score. This reliance of an artificial cut-off means that a minor score difference for most people will result in assigning of different types, thus affecting the accuracy and the reliability of these tests.
Factor analysis
Factor analysis is a statistical procedure that measures the correlations between different questions that make up the test. If there are four different dimensions, according to factor analysis the questions that test the same dimension should have high correlation among each other and the questions that measure different aspects should have little correlation between them. However, research has shown that this is not the case for tests that are based on dichotomies, like MBTI. Also research has found that there is high correlation between SN and PJ scales. Actually a quick look at the published frequency distribution of types should provide enough evidence for the interdependence between SN and PJ scales.
In spite of these criticisms, the type tests have not changed much, reducing the credibility of the type theory in general. So we bring a test that addresses these issues. Our test is different because:
- Tests are based on cognitive functions and not on dichotomies.
- Questions modelled on attitude, not behaviour.
- We don’t rely on itemized scores. We use bayesian inference for arriving at the right type.
- Minimal number of questions so not relying on averaging to arrive at types.
- We don’t use the PJ (perceiver/judger) classification, which in any case was not originally proposed by Jung but was a later addition.
References: