I’m not concerned with these particular authors having criticisms. I linked to them for fun, although the underlying issues are real.
These and many other articles and discussions—from material in peer-reviewed journals, to discussions by people knowledgeable in psychometrics, to mainstream infotainment that piggybacks on the more substantive work—all discuss similar problems.
What I have said for years is that most personality type theories are overdeterministic. Socionics and MBTI are trying to do too much. The things the theories observe include actual traits (though these are not always easily distinguished from states, and I’m not sure that all the characteristics that get tied together should be bundled). We do know that these traits show normal distribution, clustering in the middle. This means that the hard lines a 16-type theory lays down are drawn across the messy, blurry, crowded gray zones where most people actually exist. Only a small number of people will neatly fit into the boxes. And a greater number of people will not fit neatly. So ...
* people do not agree
* people do not agree on how to type others
* people do not agree with people’s self-typings
* people do not agree on how to define the types
* people do not agree on their own self-type, meaning they change it or never settle on it to begin with
* self-report tests do not agree with themselves, as people retest with different results
And so on.
If we understand that so much of what has been drawn as hard boxes simply cannot be that neat and simple, if we understand that typology is usually in the realm of -ishness (I am EIE-ish, you are LIE-ish), we are coming closer to reality. But in the case of Socionics in particular, if people were content with -ish regarding their own types, they would have to accept that intertype relations become even more -ishy. The main thrust of the IR part of the theory is prediction of what will happen between people. The less exact and the more complex the interpersonal scenario begins, the less predictable it becomes, simply by introducing more variables and more room for error.