Here's most of a letter I wrote recently regarding my typing of person that was discussed at the Polish socionics forum:
The fact is, all socionists type by association, even those who claim otherwise. And all socionists apply some type of "model thinking" — even those who seem to type by association alone. I used to make more of an effort to present my thoughts and conclusions in "model language," but I don't care enough anymore. It's a kind of intellectual laziness that comes from feeling that the socionics model as it is is doomed anyway. Associations with people of known types are actually a good way of typing (and again, every socionist relies on them heavily) inasmuch as the types of the people being associated with were properly identified and analyzed by the typer.
People would like to think that socionists have some strict algorithm they apply, because this would make socionics easier to understand, apply, and develop further. However, I don't believe I've ever met such a socionist. As soon as a person begins taking responsibility for his typings of others, he finds that "model thinking" alone is insufficient to produce a result. It's like listening to a technical debate between two experts and trying to determine who is right using your emotional reactions alone with no intellectual knowledge of the subject. Emotional reactions can be honed and cleansed of outside influences to the point that they become a fine tool for understanding many things, but they are clearly inadequate to deal with primarily intellectual matters. Likewise, socionics is primarily about how we respond to different types of people on a mostly unconscious level. Using "model thinking" alone (if that were even possible) can get you quite far, but it's not the ultimate arbiter. The ultimate arbiter is the network of invisible psychological-emotional connections between people, which are hard to put into "model language." Once this invisible network becomes evident, you can use that to correct your understanding of the model.
At least that's the way I see things. Many socionists who emphasize a model-heavy approach would disagree with me, particularly those who are unable to feel the nuances of interpersonal interaction on an emotional level. But actually the whole reason of socionics' existence is to explain relationships, and the better it does that, the more useful it becomes. The model is a semi-decent approximation at best.