Thread by @JWGrieve

Really enjoyed giving a talk today on #fakenews #linguistics for the Fakespeak project at the University of Oslo today. A quick 🧵...

My basic argument is that there a lots of definitions of Fake News out there and a taxonomy of fake news is therefore very useful to make sense of this situation and to theoretically ground our research...

The taxonomy I propose is based on the concepts of veracity (true/false news) and honesty (honest/dishonest news), which I argue are independent concepts.

If we take these two factors, we get a taxonomy consisting of four types of news: real news and three distinct types of fake news...

Personally, I think it's important to note that dishonest news is most problematic, or what we might call disinformation as opposed to misinformation, a distinction we can map on to this taxonomy...

My main point for the linguistic analysis of fake news is that we can see the opposition of honest and dishonest news as a form of register variation, as there is variation in communicative intent, and we should therefore expect structural differences...

I then use the case of Jayson Blair, which allows us to isolate variation between informing and deceiving in the writings of one journalist.

en.wikipedia.org/wiki/Jayson_Bl…

Timeline of the corpus below, where pink are fake articles and blue are real articles...

Overall, we find, among other features, that Blair generally uses past tense 'said' to introduces quotes, but in his fake articles he also uses present tense 'says', which seems like an inadvertent marker of evidentiality, marking a lower degree of certainty...

We argue that this type of insight is possible because we adopt a theoretically grounded framework based on work on disinformation and register variation, and these kind of insights can be valuable for Fake News detection in #NLProc, which often ignores disinformation...

Thanks for reading! Happy to answer any questions!

And if you're interested @helenawoodfield and I have a book coming out in 2023 w CUP's Elements in Forensic Linguistics that discusses this framework and case study called 'The Language of Fake News'

cambridge.org/core/what-we-p…

Overall, based on an analysis of dozens of grammatical features in Blair's Real and Fake News, we find that he wrote in a more informationally dense style and with greater conviction when he intending to inform the readership @nytimes than when he was intending to deceive.

FTR these discussions here on Twitter from about a year ago were very helpful for this analysis





Thanks everyone!