LogFAQs > #895931565

Yeah, this has been doing the rounds over the past few days.

Having read the study, it's maybe not as convincing or clear about its methods as I'd like. It's an interesting piece of research, but I'm not sure what it is beyond that. To be totally honest, I'm as anti-Trump as they come, and my default assumption is that Trump supporters share a shit-ton of "fake news" and conspiracy theories. I don't think for one minute they're unique in that, but it's definitely the sense I get and I do think they're particularly bad for it. I just wish the foundations of the study were a bit stronger.

On to the study, the "junk news" sites in the list are mostly right-wing (which isn't necessarily a sign of bias if those sites genuinely are less reputable), because they breached at least 3 of 5 metrics they came up with. However, as far as I'm aware, they don't show exactly what metrics each site breached, and how they determined those breaches (I could be mistaken). Also, what sites did they eliminate from the list on the basis that they didn't meet these criteria (i.e. what was the full list of sites before filtering)?

It's also helpful to be clear about exactly what the study states:

On Twitter, the Trump Support Group shares 95% of the junk news sites on the watch list, and accounted for 55% of junk news traffic in the sample.

So the 95% figure that has been banded about is the percentage of the sampled sites that they shared content from, not the actual raw amount as some have stated (which is 55%). Also, is something necessarily junk news just because it comes from a "junk news" site? Note the difference between "fake news" and "junk news". I think their term is a bit more inclusive (i.e. news that isn't necessarily outright false, but put across in a misleading way, with few sources, anonymous authors, overly emotional and partisan headlines, etc.).

The sample is also fairly limited:

From this reduced dataset of 10,691 [Facebook] pages, we collected all posts from the 90 days between October 20, 2017 and January 19, 2018

This reduced the dataset to 13,477 Twitter users.

They had a dataset a few times larger and filtered it down. The time period was the ~3 months before the SOTU.

Correct me if I'm wrong on any of this.

Thoughts?