Hacker News new | ask | show | jobs
by chronic6l 3492 days ago
> I wonder how labelling would work with news like saddam hussain's weapons of mass destruction - reported by cia, backed by the goverment as truth, but turns out it was all fake.

There's no need to talk technical about graph nodes and machine learning. It's very simple - the fake news labels are, and will always be biased in favor of the liberal/progressive/Democrat agenda. Objective news is a lie. No one wants to hear it. We want echo chambers that make us feel good. The labels will shift to accommodate this desire.

2 comments

Indeed we'll keep seeing the mass stream media biased to omit objective news.

Otherwise you would see the US population asking themselves why on earth are their armed forces so deep involved in Syria, just short of military invasion. They get spoon-fed that the country is ruled by a dictator and their population needs to be freed, but so is Saudi Arabia with a far worse regime where people get their heads cut on monthly basis, which are praised as critical ally in the region.

"Fake news" is a way to block the things deemed as dissident from the main stream, effectively blocking the room for logical argument and exposing the facts.

I think there are many people who are interested in truth - whatever it is, without feeling the need to associate their position with any red/blue person (politician).

There was no mention of machine learning (no predictions of any kind) - the idea is to keep track of history of raw data (labelling and it's changes) and report on "reliability score ..." (how many times labelling itself was false).

The thing is it doesn't need to be biased at all. You just keep all data for all sources. There's no "objective truth" per se, there are only true/false labels according to source.

Of course you can create your own "beyond reasonable doubt by xyz" or "government xyz pov" source and assume that this is going to be your objective truth reference. But, just like any other sources, your source will just have reliability rating in context of any other source.