My claim isn't about rationality recognition per se, it is simply this: psychology has shown that verbalizing can screw us up when dealing with a process that isn't normally done verbally. And a lot (if not most) of our inferential processes are not done in this explicitly verbalized manner (verbalized doesn't necessarily mean spoken aloud, but just 'thinking through in words').

My claim is that there are known ways to get good at verbalizing non-verbal processes, and they involve training on paradigmatic cases. It is only after such training that one can start thinking about edge cases and the borderlands without worrying that the process of discussing the cases is corrupting their thinking about the cases.

Before we can advance rationality by discussion, we must first learn to discuss rationality.

Understood. Thanks for the clarification. Going back and rereading the article after these comments made a few more lights click on in my head.

So, where do we start?