Then you get back-formation of labels depending on party loyalty, which is a particularly bad problem in two-party systems. Something will get labelled as 'left' because it's supported by the Democrats or 'right' because it's supported by the Republicans, not the other way round. Widespread overuse of "socialist" and "liberal" as terms of abuse rather than descriptors makes this worse.
Then there's the question of whether newspapers should simply report what people say, or make their own efforts to determine whether it bears any relation to reality. Sometimes this is impossible: the greatest piece of anti-libel hedging I've ever seen was an article calling Leon Brittan a paedophile, which was comprised entirely of reported speech.
For clustering, see http://www.ex-parrot.com/~chris/wwwitter/20050415-my_country... : an attempt at reduced-preconception data-driven partitioning of viewpoints. It seems to have held up rather well over the past decade: people attempting to look for common factors driving the Brexit vote have found that the strongest correlate is being pro-capital-punishment. The old left/right axis seems to be subordinate to the modern pro-/anti-internationalist one.
> Hahaha no.
I know "That which is asserted without proof can be dismissed without proof" and you're totally in the clear here, but it does make me wonder how you'd go about testing that claim (both yours and the GP's)? I mean, from my point of view many media outlets do lean somewhat left but partly because "reality has a left bias" or however the saying goes (although I lean somewhat left, so it may just be my filter bubble).
How do you determine it for even one media outlet? You could maybe do it by vocabulary choice. There are definitely some terms that are used primarily by one side or the other. Or the metric of what issues they try to get their readers outraged about? I've noticed sometimes there's bias not in how stories are covered, but in which stories are covered.
Perhaps it's not possible to do in isolation, but some sort of clustering technique (percentage of readers in common with another outlet as a similarity measure?) would reveal obvious groupings.
Regarding "most", would one weight different media outlets based on their viewership? Or does my local town paper count equally to Fox/whatever in adding "most"?
I guess really I'm just trying to have a more useful discussion than "media are biased left!"->"lol no", but it is actually an interesting question I think, and if anyone has resources about this (this question has definitely been asked before) I'd be interested in links.