How do you tell spam from ham? How to tell a forgery from a real painting?
The answer is: classification is now for some problem domains a tractable problem where before it required human intervention. Telling fake news from real news requires a body of data (doable), a classification algorithm (mostly a solved problem) and the will, time and other resources to actually do it.
To me it sounds very much as something worth doing and possible, that Google with their vast resources appears to drop the ball here (see Google News, which gives equal billing to ham and spam alike) leaves room for an outsider to go and tackle that problem in a decisive way.
I can see all kinds of possibilities and this to me seems like a very worthwhile task.
That's a very good question, which is why I included it. I like one definition I saw circulating which revolved around "news items intended to deceive". I agree that is too hard to pin down to be useful.
How do you know it's not fake?
There's two way I could read this. The first is distinguishing newly emerging news stories for which the facts don't exist and the second is how to automatically check.
Both are valid questions. For the first, I don't know. For the second there is an entire field of computational fact checking, as well as a range of other features.
I think a better question to ask is: how do we structure discourse around content to make our society better immune to the proliferation of low-value content? (Whether that be fake news, clickbait, disguised advertising, etc.) Also, how do we do this without creating political silos, each with their own propaganda?
