You'd have to realize that these are all just labeling and categorizing of observations... which is the cause of confusion for most people.

For example, you might say "I like eating food", and say that is Si. If you like eating food, then you must be a "Si type" or "using Si". But actually all you did was to relabel "eating" as "Si". Or to categorize anything that is related to eating as Si. So you might say that cooking is Si. Taste buds are Si. Enjoying food is Si. Whatever. Now you're categorizing whatever that might be vaguely related to food as "Si".

So what is wrong with that? Well nothing really, other than that it doesn't really create any new knowledge than just how to categorize things. For example, in physics you don't just observe a ball dropping, then categorize whatever that might be related to that. And it might not even be obvious how they're related. For example, it's difficult to ascertain how a ball dropping to the ground is related to the fact that a moon orbits around the Earth. Unless we try to find out why and how a ball drops to the ground. Newton could make the connection, because he found out "gravity" and figured out how that works. New knowledge in physics is created from knowing why and how a ball drops to the ground, for example.

So we can say that Socionics is kind of working from backwards. It's perhaps saying that a ball dropping to the ground is vaguely related to a moon orbiting around the Earth, but it's not really clear why or how, because we have yet to find gravity. In fact we're not even asking why or how, we're just saying that they're related for some reason. And we may not even be able to make the connection because they're not immediately or intuitively obvious. So there are only correlations and not causation. And those correlations could be wrong.