Slashdot Mirror
First Evidence That Social Bots Play a Major Role In Spreading Fake News (technologyreview.com)
An anonymous reader writes: Researchers from Indiana University in Bloomington provide an answer for how social bots play a major role in spreading fake news. MIT Technology Review reports: "At issue is the publication of news that is false or misleading. So widespread has this become that a number of independent fact-checking organizations have emerged to establish the veracity of online information. These include snopes.com, politifact.com, and factcheck.org. These sites list 122 websites that routinely publish fake news. These fake news sites include infowars.com, breitbart.com, politicususa.com, and theonion.com. 'We did not exclude satire because many fake-news sources label their content as satirical, making the distinction problematic,' say researcher Chengcheng Shao and co. Shao and co then monitored some 400,000 claims made by these websites and studied the way they spread through Twitter. They did this by collecting some 14 million Twitter posts that mentioned these claims. At the same time, the team monitored some 15,000 stories written by fact-checking organizations and over a million Twitter posts that mention them. Next, Shao and co looked at the Twitter accounts that spread this news, collecting up to 200 of each account's most recent tweets. In this way, the team could study the tweeting behavior and work out whether the accounts were most likely run by humans or by bots. Having made a judgment on the ownership of each account, the team finally looked at the way humans and bots spread fake news and fact-checked news.
'Accounts that actively spread misinformation are significantly more likely to be bots,' say Shao and co. 'Social bots play a key role in the spread of fake news.' Shad and co say bots play a particularly significant role in the spread of fake news soon after it is published. What's more, these bots are programmed to direct their tweets at influential users. 'Automated accounts are particularly active in the early spreading phases of viral claims, and tend to target influential users,' say Shao and co."
'Accounts that actively spread misinformation are significantly more likely to be bots,' say Shao and co. 'Social bots play a key role in the spread of fake news.' Shad and co say bots play a particularly significant role in the spread of fake news soon after it is published. What's more, these bots are programmed to direct their tweets at influential users. 'Automated accounts are particularly active in the early spreading phases of viral claims, and tend to target influential users,' say Shao and co."
So, basically these "researchers" followed tweets/rts of articles posted by left-wing sources and right-wing sources, and after making a "judgement call", they determined that the right-wing ones just had to be bots.
Color me shocked, university "researcher" confirms his own bias. Film at 11.
Here's a link to the original paper.
A quick scan shows that they... apparently use every type of data representation. You'll find a scatter plot with non-linear regression, a joint distribution image, a histogram, line charts, diffusion networks, and a triangular distribution thingy where color indicates the log of the number of articles (what the heck is that called?).
It plots the Gini coefficient of the various tweets. (I'll save you the trouble.)
The paper goes from the introduction directly to the results, then the discussion, then talks about the methods afterwards. Is that typical? I always thought the methods section comes before the results.
As near as I can tell, there is no supplemental data that indicates what their data is. Their entire paper relies on the correctness of "hoaxy" and "botometer".
Taking "botometer" as an example, looking for an estimate of accuracy, I could find no papers in established journals about this service. There are a couple of conference papers though - will those do? None has an indication of how accurate the service is.
So if "botometer" has an accuracy of 80%, which would be pretty good, and "hoaxy" had an accuracy of 80%, which would also be pretty good, then the results of the cited paper would be... how good?
Looking at the paper, I have to wonder if it's an elaborate hoax.
There's a *lot* of... um... surprising things about this paper.
OK, humoring you for a moment... name ONE credible conservative fact-checking site. Seriously. Just one.
Does Just Facts count? They're a little different than the sites listed in TFS, but they're conservative and typically accurate.
He's getting rather old, but he's a good mouse.