Took me a bit to figure that out, he's referring to "Ideological Turing Tests" https://www.econlib.org/archives/2011/06/the_ideological.htm... ... use of the term is kind of a dog whistly connection to a number of internet communities where that idea is considered very persuasive.

[The few times I've seen the term used have largely been cases where the speaker is really embarrassing themselves by claiming ignorance on the part of their opponents while demonstrating an absurd amount of their own...]

Experiments that ask people to fill out surveys as themselves, or as they imagine a liberal would, or as they imagine a conservative would, in equal proportion.

Jonathan Haidt did one of these experiments with about 2000 participants. It's social science so it should come with all the usual disclaimers, but the results can be found here and have apparently been replicated since (though I'm having trouble rapidly re-locating a study I found that replicated it, for some reason I can't find the right keywords to surface it).

You can find a writeup here: https://goodmenproject.com/featured-content/sympathy-for-the...