I think this is correct, and I also think it holds for reviewing human-authored code: it's hard to do the job well without first having your own idea in your head of what the correct solution looks like [even if that idea is itself flawed].