It is actually science, although mostly statistics. If the input is bad, the output is. The science here is getting a true/false correlation significantly over 50% from a short input like a tweet. They probably used Trump for publicity and because there is publicly available true/false data for all his tweets. But that's what data scientists do.
Yeah ok, I see your point there. But publicly available true/false data being probably from Poynter/Politifact or somesuch, I just don't put much stock in something like that. Like you said, bad input bad output. I hadn't considered that tacking Trump onto it might get them more funding for their research, but it makes sense.
publicly available true/false data being probably from Poynter/Politifact or somesuch, I just don't put much stock in something like that.
It's not about if the fact checking is biased. It would be interesting to see a study on how twitter linguistics influence fact checkers though.
BTW:
Using tweets fact-checked by an independent third party (Washington Post)
ah, gotcha
(post is archived)