Normally I’d say you should seek out more clarity on what their position is before critiquing it much. But I don’t know how. There’s no one to give official or canonical answers to questions. And I’ve read https://www.readthesequences.com (if you haven’t, you can get some info there – even just sticking it in AI and having the AI search it for relevant parts) as well as https://hpmor.com and https://equilibriabook.com I’ve tried to look for some textbook and academic paper type stuff but I’ve had a hard time finding good ones and it’s especially hard to find things that address Popperian type questions about their premises and omissions.
yes
I wrote about something similar recently. For MCDM, you can bias results by using factors that overlap (not fully independent/deduplicated). But judging overlap and getting a complete, perpendicular set of factors (or using math to factor out repetition) is a really hard problem. It’s related to the problem of judging similarity between things which comes up in induction. Whereas with CF it’s a non-issue: you can have tons of duplication/redundancy in your binary pass/fail factors without affecting the conclusion.