No matter how well you atomize a proof there remains inferential gaps that gets filled by humans agreeing that something is obvious. Some are considered axiomatic, many aren't.
... That's basically what many theists object to Yudkowsky's sequences. "There are inferential gaps".
I'm mostly asking this open question to those among us who are well-versed in developmental psychology (I'm mostly thinking of children) . Although, failing the actual scientific research on the topic, I guess some testable hypotheses would be great too.