My intuition agrees, but it has been known to output crap on similar topics. Know of any way to test this claim?

Yes, and there's a problem here, which is that adversarial debate might actually be able to productively hijack motivated cognition to generate the best arguments on both sides. Given that, finding a way to test this would be good. The most obvious way to do so is to have groups of people either discuss some difficult to figure out factual claim or to debate it with randomly chosen sides, and then see after a vote of a neutral audience whether the audience favored the correct one or not. Do this with a few different issues with different groups of people and that might get relevant data.