Absolutely agreed that it's not a 1:1, however I think Scott makes a very good case that legibility is *not* understanding either! If you want to tax, legibility will bias to what can help tax (eg. data about land size as proxy), but you won't understand how people grow food.