Here's a sum up of my conclusions on socionics.
First the thing that everyone knows and accepts: socionics is a theory that talks about abstract constructions that are based on arbitrarily discerned patterns. These patterns do not exist in any objectively measured way in reality. (And no, VI is not an objective tool. there is no proven causal link between facial structure and functions.)
But my issue is bigger than the lack of an objective tool. There are several unfounded assumptions before even getting to the point of considering to develop an objective measurement. Here's just some example issues at the conceptualization level.
There is an assumption that information elements determine a big part of communication. Nope. Communication between people is based on something more complex than abstract information elements or functions. The framework of evolution theory works better to explain human communication (too long to get into details here about that).
There is another common assumption, that is, if information elements are a certain way of information processing (which statement is fine on its own, but only until you try to get into details!) then there is very high correlation and in some cases even a direct causal link between the functions and between all the known cognitive elements, personality traits and/or concrete behaviour elements. This is simply not true, either in an absolute or a relative way.
Basically, the concept of these functions is an attempt to group many observable elements together. But in reality, there is just a brain that processes information in a much more complex way cognitively. To elaborate on this further: for the brain there are many processes to do before it results in something that socionics users attribute to the functions way of information processing. These processes are the building blocks for information processing in the brain. There is no guarantee or proof whatsoever that all these will result in the exact functions as organized on the high level in socionics theory.
Why no guarantee whatsoever? Well, sure, there are correlations between elements or even personality traits but weak and there is no direct causal link between these correlations. The causes are factors in the background that is not part of the socionics theory. So what this means is that these background factors cause the things that you've observed as being correlated to each other. The things observed are not causing each other, something else is causing them. I am not saying that the correlations are random, obviously they aren't random but that does not constitute a proper explanation. This means you cannot draw real conclusions from this model.
That's fine, for people-analysing, we do not know that much yet about the brain's complex workings anyway. In practice everyone obviously just uses a few rules for efficient (or not-so-efficient) communication and all the other usual common sense life strategies that are well known outside socionics. As example, here's one generic common sense statement: "people look at the world in a different way". The thing is, socionics itself does not add anything worthwhile to these common sense strategies. It can be a vehicle for some people to hear about strategies and then try to apply these strategies. Basically, people advising other people on psychology issues.
This is again all fine, just not directly related to the theory. Any kind of applied psychology with very different theories behind does the same job essentially. These theories are very different and yet, still work like socionics does or work even better. Also very important here is that these different workable theories achieve this without adding the extra assumption of the notion of certain detailed concrete types, strictly defined communication styles between types including the duality concept and other things. So how about sticking to occam's razor and discard the superfluous assumptions.