In philosophy this phenomena is called Argumentum ad populum, or the bandwagon fallacy. Simply put, the volume of votes do not have anything to do with the accuracy or veracity of a claim. (See wikipedia entry here - http://en.wikipedia.org/wiki/Argumentum_ad_numerum) I've been thinking about this problem for a long time since the most popular forms of value definition on the web are simple multiple-choice voting.
But what's better? Will we see value judgements for accuracy derived from natural language analysis - the same way that we see search engines derive value judgements for relevance from web pages and links? Or will an open reputation standard evolve similar to eBay reputations, which again would be open to the same problem?
Seems to me we're asking computers to do something that they're not very good at doing, except in dead-simple ways like Digg, so we'll continue to get dead-simple outcomes.