I don't see why this isn't a good model for how human reasoning happens either, certainly as a first-order assumption (at least).