No theory or model will ever have complete 100% correspondence to reality. Inevitably, the only thing you can do is predict aggregate tendencies and patterns in nature that fall within certain standard deviations. And any responsible scientist is a constant skeptic who knows that all theories and models are tentative and should always be subject to revision impending new evidence. Rationally, we can't accept something as 100% true because we know that we don't know everything yet. Nevertheless, as our methods of probing the universe around us improve, our subsequent theories allow us to take in a greater breath of that universe with greater precision. But that doesn't mean these theories are not useful. Newtonian Mechanics was superceded by General Relativity and Quantum Mechanics. We know now that the universe doesn't *really* operate in the ways NM describes, but NM is a deciseively useful and pragmatic approximation of the universe. You can build perfectly functional airplanes and skyscrapers without any knowledge of time dilation or Heisenberg Uncertainty. However incomplete some theories may be, they do appear to be at least describing something that is there.
So it is with Socionics. I find it a useful model for explaining and predicting a given individual's behavior, in parts. I don't think it is a matter of my perceptions being reframed by the model either, because I was very much aware of these patterns in people well before I ever knew about Socionics, or MBTI for that matter. I could observe and see

in action in people that I knew well before I ever knew that such a thing like

might exist. Same with other functions, like

and

. These psychic structures do appear to have some kind of tangible existence. So overall, Socionics is really nothing more for me than a reference frame for describing and predicting things I was already well aware of beforehand. And I mold it more according to my own experiental data than I do semantic data compiled from reading materials out there. Parts of Socionics I can confirm with my own experience (like the functions), parts of it I can't (like the relationships... duality, supervision, etc.). I take what fits, I dismiss what doesn't.
From a purely scientific POV, Socionics would fail *pathetically* as a valid theory. There are far too many holes and inconsistencies at this point. As a personal tool for my own sake however, it's very useful and that's as far as I take it. With time, Socionics may evolve into something better or be superceded altogether by something with more accurate and comprehensive explanatory power. In the meantime, I wish people would quit

-ing Socionics to death like a fucking logic proof, especially when they try to type others. It doesn't work, STOP it. But that's another rant for another time.