Glad to find a kindred spirit in terms of applying socionics to real people when you learn it
but I'm pretty against visual identification. Even if it worked (I think it does not), I like that I can only really know a person's socionics type once I get to now them. Its almost the last thing I learn about them, and it is a way to check if everything I learned about them is consistent. If I'm not sure of a person's type, that means I don't actually know them well.
But even if VI is controversial, It would be interesting to put it to the test. If we had a data base of people and their type, assuming people could agree that the database was accurate, we could show a picture or video clip of a person and test the accuracy of each person's ability to visually identify type.