There's an attitude I'll caricature as dividing all claims into either sanctioned certainties or mere opinion. Reasoning is a competitive game of justifying those conclusions that'd benefit your side, on the basis of arguments within the blessed circle.

Against this, you might pit an ideal bayesian reasoner tracking a degree of belief in all possibilities. But that's too hard. It is possible for humans to roughly track a few possibilities; if you pick simple ones, none will be exactly right, but you can learn from how they go wrong or not. It's like "try the simplest thing that can possibly work" in programming: you expect there's a good chance it won't, but it can be the most productive way forward to figuring out what will.