Play Four: Fake Health News

When researching health topics on the internet, you often have to scroll through a page of false results before finding factual information. This condition is referred to as a “data void,” which happens when a search using a keyword reveals “content produced by a niche group with a particular agenda.” According to Renee DiResta, director of research at New Knowledge, the reason for lack of factual counter-content is relatively simple: No one feels the need to produce it, and even if they did, it is unlikely to go viral. Further, when a plethora of fake news stories result from a keyword search, it isn’t because the message represents the conventional wisdom, but rather is a result of these niche groups having thousands of members who thrive on social media and oftentimes pay for ad campaigns to reinforce their beliefs (“The Complexity of Simply Searching for Medical Advice,” Wired, July 3, 2018; wired.com/story/the-complexity-of-simply-searching-for-medical-advice).

Fake medical news is possibly the most nefarious category of false content, as it plays on peoples’ hopes for good health, and its determination requires a slightly different checklist of evaluation criteria. Medical journals need to be peer-reviewed by colleagues in the same field. Also, the plausibility of the claims needs to be considered. Remember this caveat: If it seems too good to be true, it probably is. In evaluating pharmaceutical news, it is important to keep in mind that prescription drugs must undergo a rigorous testing process before they can be given to humans. Despite news of miracle cures and breakthrough treatments, it could be many years before these products are available to patients.

Clickbait health-related headlines often begin with titles such as “The Secret That Even Doctors Won’t Tell You.” Vulnerable people click on these links, garner traffic to the hosted websites, and inadvertently generate revenue for the content producers. Doctors have little incentive to keep knowledge of effective treatments from patients; indeed, doing so would be in complete opposition to their mission of saving lives and curing disease.

Luckily, there are reliable research tools that librarians can use in helping their patrons search for medical information. HeathNewsReview.org reviews medical news coverage and assigns a validity score to claims based on a 10-point list of criteria denoting credible health news. NHS Behind the Headlines (nhs.uk/news) is a U.K-based site that analyzes the science upon which news is based and offers a conclusion and interpretation of findings for each. It also features links to underlying research and data.

In addition, we can also use our own research skills to make health-related news determinations. In analyzing a health-related story, try doing a Google search for some of the keywords that it contains along with additional terms such as myth , hoax , scam , false , conspiracy , clickbait , and ™ junk science∫ . Your list of results might then contain an article that debunks the science in the article in question. Also, the comments sections of articles can be revelatory. For example, another reader may have commented, “This article contains junk science” and may have also provided links to prove it.

Play Five: Fake Videos

Facebook users, in 2016, viewed videos 8 billion times per day (“9 Insane Statistics on the Future of Internet Video,” Jeremy Goldman, Inc., Oct. 17, 2016; inc.com/jeremy-goldman/9-insane-statistics-on-the-future-of-internet-video.html), but a Facebook executive predicted that the service would be “all video” in 5 years (“Why Facebook Could Be ‘All Video’ in 5 Years,” C. Zillman, Fortune, June 14, 2016; fortune.com/2016/06/14/facebook-video-live).

Further, Facebook pays news organizations $250K for providing 3 months of 20 videos per month. It awarded a $3.03M contract to The New York Times for 1 year of broadcasting live (“Facebook Signs Deal With Media Companies, Celebrities for Facebook Live,” S. Perlberg and D. Seetharaman, The Wall Street Journal , June 22, 2016; wsj.com/articles/facebook-signs-deals-with-media-companies-celebrities-for-facebook-live-1466533472).

And, like it or not, Facebook is a huge provider of news content; Aarti Shahani of NPR News called it “the new front page of the news for more than 1 billion people every day” (“From Hate Speech to Fake News: The Content Crisis Facing Mark Zuckerberg,” NPR, Nov. 17, 2016; npr.org/sections/alltechconsidered/2016/11/17/495827410/from-hate-speech-to-fake-news-the-content-crisis-facing-mark-zuckerberg).

As news evolves from text to video, it seems a natural progression that fake news will evolve into fake videos.

How can you tell if a video is fabricated? The technology makes it almost impossible. In July 2017, University of Washington computer scientists used audio clips and algorithms programmed to manipulate mouth movements in order to produce a lip-synced video of Barack Obama that was so life-like, Obama’s former security advisor, David Edelman, thought it was real (“Obama Tech Advisor: Fun’s Over for AI-Driven ‘Deep Fake’ Video,” G. Nott, Computerworld Australia, June 14, 2018; computerworld.com.au/article/642424/obama-tech-advisor-fun-over-ai-driven-deep-fakevideo).

BuzzFeed was able to do something similar using an artificial intelligence (AI) program called FakeApp. Media fabricated through AI (called “deep fakes”) are a huge concern to both lawmakers and intelligence officials. The U.S. Department of Defense’s Defense Advanced Research Projects Agency (DARPA) has held contests to encourage development of tools to detect deep fakes; the European Union has followed suit. Other companies are working to develop software that uses block chain technology to watermark each change to video content. However, since a lot of these tools are still in development and could possibly be cost-prohibitive when brought to market, librarians must rely on a few rudimentary techniques when analyzing videos.

First, as in the IFLA checklist, consider the source. Can you find the original video creator and its source? Does the video content match the content of its accompanying story? In reviewing the video itself, consider the following questions:

• Do the eyes blink? Human beings blink an average of 17 times per minute. The subjects in deep fakes often do not blink at all.
• Is the mouth realistic? It is difficult to create teeth, tongue, and mouth interiors that look realistic in deep fakes.
• If you can examine the video on a pixel level, you should be able to see blood flow in the faces of the subjects. Also, can you detect the subject’s pulse? If there isn’t any blood flow and they don’t have a pulse, the image may not be that of a real person.
• Can you slow down the video to freeze-framed shots? Deep fakes often have glitches as they transition from frame to frame.

Play Six: Eyes on the Prize

A 2018 Gallup/Knight Foundation poll asked U.S. adults, “Who’s Responsible for Fighting Fake News Online?” Librarians and information professionals were not an answer choice, but according to data journalist Felix Richter, the results were telling: Only 38% of respondents believed that users them selves should take responsibility for being able to tell the difference between real and fake news (“Who’s Responsible for Fighting Fake News Online?”; statista.com/chart/17549/responsibility-for-fighting-fake-news).

As a counterpoint, Pew Research found that 78% of American adults believe libraries “help them find information that is trustworthy and reliable,” with 56% stating that libraries help them “get information that helps with decisions they have to make” (“How People Approach Facts and Information,” John B. Horrigan, Sept. 11, 2017; pewinternet.org/2017/09/11/the-elements-of-the-information-engagement-typology).

Clearly, our constituents are looking to us for coaching and believe that we have the playbook to help them win the fake news game. It is up to us to call the plays, making sure that we remain hyper-vigilant, constantly tweaking and updating our game-day strategies while simultaneously encouraging others to take a page or two from our playbook as well.