It should not have come as a shock that even before Twitter Inc. removed accounts flagged for suspicious activity - i.e. spreading fake news, and re-tweeting Russian bot BS, we knew that the fake news out of assorted rat warrens, like Reddit, 4chan, FOX News etc. which ended up in tweets, spread faster than the facts.
An analysis of more than 4.5 million tweets and retweets posted on Twitter from 2006 to 2017 indicates clearly that inaccurate news stories - or deliberately falsified ones (e.g. the "Pizzagate conspiracy" of Michael Flynn et al) - spread faster and further on this social media platform than true stories. Further, people - humans - play a bigger role in sharing falsehoods and fake news than the bots.
Those findings were reported in the March 9 issue of the journal Science, and could easily guide strategies for curbing misinformation on social media. Indeed, the report may already have been used by Twitter Inc. to dump up to 6 million resident troll accounts. (WSJ, 'Twitter Subtracts Suspicious Accounts From Follower Totals', July 12) As per the WSJ piece:
"Twitter is removing accounts that had been blocked due to an unusual change in behavior such as the sudden sharing of misleading links, sensitive information and other types of problematic content. Removing the locked accounts will reduce follower counts by 6 % across the service, the company said.."
Of course, Maul Street generally balks at any company changes that might lower profits, never mind they might also cool off some of the polarization and incivility from Twitter trolls. So no surprise Twitter's share price fell nearly 9 percent last Monday after The Washington Post reported Twitter was accelerating removal of problem users, accounts. (Ibid.)
The significance of the Science paper is that most studies hitherto have been anecdotal only, according to Filippo Menczer, an informatics and media prof at Indiana University- Bloomington. In his words, interviewed by Science News: "We didn't really have a large scale, systematic study evaluating the spread of misinformation".
How large scale are we talking about for a worthwhile study? In the cited paper, researchers examined 126, 000 tweet "cascades" to discern fake news mongering on that platform. The cascade was defined by families of tweets composed of a single original tweet and also all the retweets spawned by that original message. All of the cascades focused on one of about 2,400 news stories that had already been debunked by at least one fact-checking organization (e.g. Factcheck.org).
The media scientists at MIT - led by Deb Roy - then investigated how far and fast each cascade spread. They found that discussions of false stories tended to start from fewer original tweets, but some retweet chains ended up reaching tens of thousands of users. Meanwhile,true news stories never spread to more than 1,600 people.
Interestingly, true news stories took about six times longer than false ones to reach 1,500 people. In other words, we have a graphic, real world validation of Winston Churchill's famous quote: "A lie can travel half way around the world before the truth can get its pants on."
The question, obviously, is why? Why is truth, or any factual account, so slow to propagate relative to falsehood? Is there some brain defect or anomaly that causes humans to gravitate more to untruths, or fake stories, and spread them?
A key finding of the Roy et al study was that bot traffic didn't make that much difference to speed of propagation of falsehoods. Initially, the researchers removed all bots from the analysis, but when bot traffic was re-introduced, it was found the bots spread true and false stories about equally. The only conclusion? Humans, not bots, are the primary agents for spreading false stories.
So why are humans more prolific spreaders of fake news? The suggestion of one of the study's co-authors - Soroush Vosoughi - is that the yen for novelty is at the root. In other words, humans are more inclined to spread tall tales because they are perceived to be more novel. Look at it this way, a Reddit user is scanning headlines and spots two stories, one is an account of a newly discovered pulsar with its signal flickering at 30 times a second, the other concerns a 30 year old wife and mother claimed to have been abducted by aliens with sexual experiments on their mind. Well, nine will get you ten the Reddit user will choose to spread the latter story - as an original - on his Twitter feed. You can also be sure in a race with the other story about pulsars - say tweeted by a Brane Space reader- the alien abduction- fake news will spread to four or five times as many via retweets, and a lot faster than the pulsar story.
Pulsars are just not as fascinating or novel as the abduction of matrons by hyper-sexed aliens.
Vosoughi reinforces this POV, noting that compared with the subject matter of true stories, the fake news topics tend to deviate more from the themes users were exposed to earlier, say in the two months before a user retweeted a story. So there's a much greater chance a potential retweeter would have already read or seen the pulsar story and responded with a 'Meh' while the suddenly appearing alien abduction of a human female might have sent him instantly tweeting it out with a "Wow! Let's see now, 'Alien sex fiends abduct MILF!'
Tweet replies to false stories also were found by the study to contain more words than true stories, indicating surprise. Again, no astounding find.. Give an assignment to two users to write as many words as they can think of about pulsars, and about "Alien sex fiends abducting MILF". I promise the latter will be found to have sparked more words, all fictitious and imaginary, but still words. A little element Dr. Pat Bannister also uncovered in her years of research: that liars and confabulators tend to embellish, use more words that truth tellers.
Informatics Prof Menczer, not surprisingly, asserts the analysis on the speed of fake news propagation "provides a very good first step" in understanding what kinds of posts grab the most attention. Well, it appears the result is obvious: those posts that are most novel (fewer points of origin) and can most easily be embellished.
See also:
http://www.smirkingchimp.com/thread/william-rivers-pitt/80204/even-fox-commentators-agree-putin-s-helsinki-hustle-humiliated-trump
No comments:
Post a Comment