Success in social interaction is not about holding more things in your head but often about holding less things in your head.

Sounds like we do need to go into the longer conversation.

I view most of these skills are something like the follows: at level 0, you have no clue what's going on; at level 1, you have a system 2 model of what's going on that's too slow / clumsy to operate successfully in real time; at level 2, you have a system 1 model of what's going on that's fast and good enough to operate successfully in real time.

Most people go directly from level 0 to level 2, with some level 1 help. Most language speakers don't have an abstract grammatical model of their language in their heads, some constructions "just look weird" or don't come to mind, and they often can't articulate rules even if they use them correctly.

For example, in English, why is something "harder" instead of "more hard"? Why is it "more difficult" instead of "difficulter"?* (This came to mind because my mother is teaching ESL classes and had been surprised that there was a simple underlying rule, which I could not successfully identify before the question was spoiled, even though for any particular word I could correctly determine whether 'more' or 'er' was appropriate.)

But there are situations where it seems better to go through level 1. If you're teaching someone a second language, for example, they're much more likely to be able to make use of abstractly stated grammatical rules than children are. If someone has already been a child and yet not developed a 'normal' level of social intelligence, then the normal approach is inadequate, and we need to consider alternatives.

When developing those alternatives, it's worth noting that the right approach for going from level 0 to level 1 (learning more grammatical rules into system 2) is different from the right approach for going from level 1 to level 2 (practicing the grammatical rules into system 1). So yes, someone who is at level 1 would not get much out of holding more things in their head, but someone who is at level 0 would.

(To elaborate even further, I think someone at level 0 probably has some feature of their personality / communication style at a fundamental enough level of their model that they won't generate hypotheses that contradict it, and thus large parts of human interactions will be fundamentally mysterious to them. The point of reading about personality styles and communication styles and so on is that it generates alternative hypotheses at that level--many 'nerds' do not realize that there are people out there who interpret statements as about relationship closeness instead of as about factual accuracy, and pointing that out to them is the fastest way to level up their interaction ability.)

* Single syllable adjectives get an "er" or "est," multi-syllable adjectives get a "more" or "most," at least most of the time.