You should call it black and white. Because that's what it is, black and white thinking.

Just think about it : using nothing more than one bit of non normalized information by compressing the opinion of people who use wildly variable judgement criteria, from variable populations (different people care and vote for different topics).

Then you're going to tell me it "works nonetheless", that it self-corrects because several (how many do you really need to obtain such a self-correction effect?) people are aggregating their opinions and that people usually mean it to say "more / less of this please". But what's your evidence for it working? The quality of the discussion here? How much of that stems from the quality of the public, and the quality of the base material such as Eliezer's sequence?

Do you realize that judgements like "more / less of this" may well optimize less than you think for content, insight, or epistemic hygiene, and more than it should for stuff that just amuses and pleases people? Jokes, famous quotes, group-think, ego grooming, etc.

People optimizing for "more like this" eventually downgrades content into lolcats and porn. It's crude wireheading. I'm not saying this community isn't somewhat above going that deep, but we're still human beings and therefore still susceptible to it.