15

I've encountered repeated references to fake news on Facebook influencing various elections. Example from a often quoted Facebook critique in Time Magazine:

As a result, when confronted with evidence that disinformation and fake news had spread over Facebook and may have influenced a British referendum or an election in the U.S.

There are of course many other references to the same in the press.

So, the question is:

Is there any systematic trust-worthy (e.g. performed by qualified people on large data sets, and preferably peer-reviewed) research that quantifies the influence of fake news on Facebook on US 2016 elections? Absent that, is there any data at all that these claims have or have not any factual basis, beyond mere conjectures driven by political beliefs of media writers?

For the purpose of this discussion, "fake news" are defined as content masquerading as news or news commentary, purposely written with the intent to deceive or mislead, or references to fictitious (satirical, fiction literature, just plain invented out of thin air) content as factual, for the purpose of convincing people to vote one way or another (or abstain from voting). It does not include sincere media news publications that proved to be erroneous later, or clearly marked opinion pieces that somebody considers wrong. This also does not include clearly attributed sincere political ads on social media or outside, even if those are getting some facts wrong.

StasM
  • 473
  • 4
  • 12
  • 1
    related: https://skeptics.stackexchange.com/questions/35850/did-mike-pence-advocate-conversion-therapy-for-homosexuals – DavePhD Jan 17 '19 at 20:49
  • 3
    The strong connection between the 2016 presidential election and the Brexit referendum is the participation of [Cambridge Analytica](https://www.nytimes.com/2018/03/19/technology/facebook-cambridge-analytica-explained.html), a political consulting firm which worked for the Trump campaign and reputedly was involved in the [Brexit campaign](https://www.newyorker.com/news/news-desk/new-evidence-emerges-of-steve-bannon-and-cambridge-analyticas-role-in-brexit). – Daniel R Hicks Jan 17 '19 at 22:07
  • The problem with this question is that: "It does not include sincere media news publications that proved to be erroneous later" excludes almost everything. There is nothing so outlandish that someone won't believe it sincerely and repeat it. Also why would you think fake news (by whatever definition) would be less effective than non-fraudulent advertising, which has been shown to be effective both by peer-reviewed study and by market behavior? – antlersoft Jan 17 '19 at 23:03
  • @DavePhD how it is related? – StasM Jan 18 '19 at 00:09
  • @antlersoft It doesn't exclude everything. There's a lot of difference between sincere mistake and fake. I am not saying it excludes people that sincerely believe fake news. I am saying it excludes legitimate news reporting that was, for example, based on an error. It doesn't exclude fakery that somebody sincerely believed. Also, there's a good reason to suspect political ads aren't that effective either: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3042867 but anyway I'm not interested in conjecture but actual data. – StasM Jan 18 '19 at 00:14
  • StasM, It’s just one example of fake news that influenced the election. Not a duplicate of your question or an answer to your question. – DavePhD Jan 18 '19 at 00:20
  • This is an essentially precise duplicate of https://skeptics.stackexchange.com/q/42606/3943 which doesn't have an answer so can't be registered as a dupe apparently. – matt_black Jan 18 '19 at 00:54
  • @matt_black no, it's not a duplicate - that one talks about ads, this one is about fake news (which is completely different from legit ads). – StasM Jan 18 '19 at 00:59
  • @matt_black the other question is asking about ads (which could have fake info, real info, or just opinion). That's much different than fake new, if the audience knows it's an ad. The only grey area is stuff like this https://skeptics.stackexchange.com/questions/42407/is-china-placing-advertisements-seeking-to-influence-the-upcoming-american-mid-t where the ad tries to look like news. – DavePhD Jan 18 '19 at 00:59
  • @DavePhD how do we know it influenced the election? Any data exists about it? – StasM Jan 18 '19 at 01:00
  • @StasM no, I didn't mean it as an answer. I can delete it if you want, let me know. – DavePhD Jan 18 '19 at 01:01
  • In this context can't fake news be reduced to people talking to other people, or does it have to be an image and text posted by a non mainstream organisation? – daniel Jan 18 '19 at 06:32
  • I don't think it's possible to answer this question, at least not without going back in time and trying again without fake news. I do know it had an effect, because even here in Germany some of my colleagues were affected by fake news about the US election, but whether the overall effect on the election is statistically significant is impossible to tell. –  Jan 18 '19 at 09:22
  • @DavePhD The other question wasn't just about ads: it was about any targeted content (the title says ads but that is because as a non-facebook users I have no idea how to distinguish between different methods of getting content onto the platform). The body text clearly stated that the question was about all content. I don't mind if you merge the two as good answers would cover both. – matt_black Jan 18 '19 at 10:05

1 Answers1

7

According to Less than you think: Prevalence and predictors of fake news dissemination on Facebook Science Advances 09 Jan 2019: Vol. 5, no. 1:

Some have gone so far as to assert that such content had a persuasive impact that could have affected the election outcome, although the best evidence suggests that these claims are farfetched [reference 2].

where reference 2 is Social media and fake news in the 2016 election J. Econ. Perspect. 31, 211–236 (2017)

which concludes:

the new evidence we present clarifies the level of overall exposure to fake news, and it can give some sense of how persuasive fake news would need to have been to have been pivotal. We estimate that the average US adult read and remembered on the order of one or perhaps several fake news articles during the election period, with higher exposure to pro-Trump articles than pro-Clinton articles. How much this affected the election results depends on the effectiveness of fake news exposure in changing the way people vote. As one benchmark, Spenkuch and Toniatti (2016) show that exposing voters to one additional television campaign ad changes vote shares by approximately 0.02 percentage points. This suggests that if one fake news article were about as persuasive as one TV campaign ad, the fake news in our database would have changed vote shares by an amount on the order of hundredths of a percentage point. This is much smaller than Trump’s margin of victory in the pivotal states on which the outcome depended.

DavePhD
  • 103,432
  • 24
  • 436
  • 464
  • 7
    This fails to take into account the targeted approach supposedly used by Cambridge Analytica, where a small number of people on Facebook were "targeted" based on their apparent "waveringness" and their presence in a pivotal district. – Daniel R Hicks Jan 18 '19 at 01:58
  • 3
    Did CA distribute fake news? I thought they delt with legitimate political ads. – StasM Jan 18 '19 at 07:39
  • 2
    @DanielRHicks TV, Radio, newspapers all target relevant districts when appropriate, but districts only matter for the House and in Nebraska and Maine for electors for president. Sometimes it's better to target supporters so they turn out, instead of actually trying sway people. If you have good info, add another answer. – DavePhD Jan 18 '19 at 11:41
  • @DavePhD - The NYT and NewYorker articles I referenced earlier describe things in some detail but further research would get more technical than I care to deal with at this point, plus much of the info is not readily available to the general public. – Daniel R Hicks Jan 18 '19 at 12:56
  • 1
    @StasM - It's not clear. I gather that CA mainly shared opinion pieces, so the "fakiness" was deniable. – Daniel R Hicks Jan 18 '19 at 12:59
  • 4
    I feel that this is a flawed analysis. It would not surprise me at all if one fake news article had more affect on an individuals decision then one campaign ad, even significantly so. Campaign ads all start to look the same and get ignored quickly, there is little expected news in them either. Where as if someone read (and believed) that a candidate kicked puppies and burned down orphanages that could make a far more lasting affect. Put another way the ability to *lie* in fake news allows presenting more concerning, and thus vote swaying, news then campaign adds that are nominally true. – dsollen Jan 18 '19 at 15:28
  • @dsollen even my references could be a form of fake news. Fake news, fake fact checking sites, there's no end. – DavePhD Jan 18 '19 at 15:33
  • 5
    I'm categorically baffled by studies that imply major corporations (as well as candidates) are stupid in spending money on advertising. – Andrew Lazarus Jan 18 '19 at 20:59
  • 1
    Your summary omits some of the qualifiers the authors use later in the paper: they could be underestimating the articles seen and the impact of each article. The error bar certainly comprises an order-of-magnitude; at the high end, it would be enough to sway the results in some of the swing states. Interesting paper though, thanks for the link. Also, data gathered by online surveys would not be representative-- bots and paid trolls that enthusiastically shared fake news would be unlikely to respond to surveys about their sharing behavior. – antlersoft Jan 19 '19 at 00:49
  • @antlersoft bots & paid trolls also would be irrelevant for counting the influence of the fake news (as opposed to raw distribution number) - however many bot accounts shared the fake, it's irrelevant until it gets to the target - the real voter. That's where influence counting starts. Judging by the paper quoted, they said it's hundredth of a percent. If they were wrong by order of magnitude (unlikely, as it means fake news are 10x better than legit ads) it's still tenth of a percent. I don't think margins of victory are this narrow in presidential election anywhere. – StasM Jan 20 '19 at 03:29
  • I'll wait a couple of days in case somebody else wants to add some data but if there's not anything better I'll accept this one. – StasM Jan 20 '19 at 03:30
  • @StasM - No, their methodology for deciding the reach of fake news was dependent on the extent that it was shared, and they relied on survey data to determine how much it was shared. If the survey data on sharing was wrong (and it was if much of the sharing was by bots and the like) then their reach estimate is also wrong, so the number of impressions is wrong. The study really is an order-of-magnitude estimate at best. – antlersoft Jan 21 '19 at 22:58
  • @antlersoft Wait, if their data relied on survey of real people - as bots do not usually answer surveys - I don't see how sharing by bots changes survey-derived outcomes. Of course a lot of sharing is done by bots - but the point is not sharing, the point is end impact on voters. And yes, it may be imprecise, but imprecision in sharing numbers does not really matter if influence numbers are still right, and these as I understand come from surveys, not from sharing numbers. – StasM Jan 21 '19 at 23:32
  • @StasM - The way they estimated the reach of the fake news (how many people saw it) was based on the sharing data from the survey. If most of the sharing wasn't reflected by the survey, their reach numbers are way off, and so will be their estimates of influence. – antlersoft Jan 22 '19 at 00:26