Misinformation, Disinformation, and Media Literacy in a Less-Centralized Social Media Universe

Knight First Amendment Institute, Reimagine the Internet 

Great panel today; more to come the rest of the week and they will shortly post the video. 

Francesca Tripodi (UNC) shared her amazing research about how conservatives use textual interpretation techniques to interpret information and reject journalistic interventions. Conservatives then use and trust Google’s top results, believing that Google top results reflect reality, which seems a bit contradictory to me. The problem is that our keywords are ideological, so Google searches confirm one’s worldview: searching for “illegal aliens” gets you right-wing sites that confirm what they already believe, while “undocumented workers” produces very different results. And it’s not just Google—DuckDuckGo is better for your privacy but returns the same type of results based on ideological keywords. Google suggestions create the possibility of parallel internets that are invisible to outsiders. “Data void”: limited/no content is available, so it’s easy to coordinate around keywords to guarantee that future searches are directed to content that includes these terms—this is what happened to “crisis actor.” Search engines are not designed to guide us through existential crises or challenge our beliefs—the notion of relevance is subjective and idiosyncratic as well as unstable and exploitable. Knowing/understanding audience concerns and amplifying key phrases allows conservative media to drive users to search where their beliefs will be reinforced. Like Council of Conservative Citizens reaching Dylann Root in his searches for black on white crime. They encourage viewers to “do the research” while highlighting phrases that lead to the preferred sources. So Google started autofilling “Russian collusion” with “delusion,” a phrase promoted by Roger Stone. In impeachment proceedings, Rep. Nunes used his opening remarks to repeat a few names/phrases and tell us that we should be paying attention to those—which, when searched in Google, linked to Fox, Daily Caller, and even more right-wing sources. Urged constituents to do their own research. Nelly Ohr: a perfect data void/litmus test. She used to work for Fusion GPS and is part of a conspiracy theory about Russia investigation—the search exists in a vacuum and was curated by conservatives as a dogwhistle about election fraud.

What can we do? How can Google fix this? It’s important to stop thinking about a fix and focusing on Google. Misinformation is not a bug in the code but a sociological issue. The only way to circumvent misinformation traps is knowing the kinds of Qs people seek answers to, knowing how they interpret information, and knowing how political actors exploit those things. [Easy-peasy!]

Barbara Fister, Gustavus Adolphus College: In practice, students are treated as information consumers who need to be educated to examine claims. At universities, they are often treated as needing help finding information in the walled garden of the library, focusing on information that will help them satisfy professors. Libraries have felt compelled to emulate Google and create single-search boxes. But the results don’t help you navigate the results, so it’s no wonder that students come up with workarounds. Students have trouble getting themselves situated. They adopt a strategy and stick to it; look for “safe” sources; often don’t really care about the topic because it’s been assigned. Follow the news, but don’t trust it; don’t think college does much to prepare them to ask questions of their own. Feel both indignation and resignation about algorithmic systems invading their privacy. Students feel that they’re in a very different place than professors; they’re used to different sources. “We grew up with untrustworthy sources and it’s drilled into us you need to do the research because it can’t be trusted.” Students are already being taught “media literacy” but more of the same won’t necessarily help, because people who believe misinformation are actually quite “media literate” in that they understand how these systems work and are good at manipulating them. Qanons understand how media/info systems work; they interpret media messages critically; they feel passion for discovery and enjoy the research b/c they feel like they’re saving the world. Alternate authority structure: trust yourself and trust Trump/“the Plan.”

What is to be done? Deep-seated epistemological differences: if we can’t agree on how we know what’s true, hard to see common ground. So what’s next? Recognize the importance of learning to trust, not just to be skeptical; get at why to trust rather than what to trust—saying “peer-reviewed research” doesn’t help; explore underlying values of knowledge systems, institutions, and practices such as journalism’s values; frame learning about info systems as education for democracy: you have a role to play; you should have an ethics of what it is that you will share. Peer-to-peer learning: students are learning from each other how to protect privacy etc. Students are concerned about their grandparents and about their younger siblings—interested in helping other age groups understand information.

Ethan Zuckerman, moderator.

Fister: Further reading: Information Literacy in the Age of Algorithms—what students are interested in that doesn’t come up in class: knowing that Google works by using the words we use rather than as a neutral broker would be very important! Alison J. Head (January 5, 2016), Staying smart: How today’s graduates continue to learn once they complete college; Project Information Literacy Research Institute, Alison J. Head, John Wihbey, P. Takis Metaxas, Margy MacMillan, and Dan Cohen (October 16, 2018), How Students Engage with News: Five Takeaways for Educators, Journalists, and Librarians, Project Information Literacy Research Institute.

Tripodi: People would say “I don’t trust the news” and she’d ask where they got candidate info; they say “Google,” without acknowledging that Google is an aggregator of news/taking content directly from Wikipedia. We’re not in a new time of epistemological fissures or polarization—we have always been in a place of big differences in how we seek truth, what are sources of knowledge, how we validate knowledge. What’s changed: we can connect from further away and we have an immediate ability to determine what we think is right. Focus on keywords is something that work on filter bubbles hasn’t yet considered—it’s not the tech that keeps us in the filter bubbles; we are the starting point for that closure.

Zuckerman: the people who find hate speech on YouTube are the people with hateful racial attitudes—so the polarization argument may not work the way we thought.

Fister: the power to amplify and segment market messages is way more pronounced now. But it was deliberate fissure with the rise of Fox News, talk radio. Amplified by platforms that like this content b/c controversy drives attention. Far right white supremacists have always been good at tech—used film early, used radio; they are persuasion machines designed to sell stuff. They are earning money while using the platforms, which has changed the velocity/amplitude of the most hateful speech.

Tripodi: There may be ways to figure out the keywords that resonate with people’s deep stories, to find the data voids, by doing more ethnographic work. The narrative that conservatism is being silenced: trying to reshape objectivity as “equal balance.” Rebranding of objective to mean “both sides.” If your return doesn’t show equal weight, it’s somehow flawed/biased/manipulated [at least if your side isn’t dominant]—that’s leveraged in the rightwing media ecosystem to say “don’t use these platforms, use these curated platforms that won’t ‘suppress’ you.” That’s complicating notions of media literacy, which sometimes uses “look for both sides” as an indicator of bias. Propaganda campaigns are now leveraging the idea of “lateral reading”—looking for relevant phrases around the target of interest in a new window—these systems are being deliberately exploited. Thinking about keyword curation may help: you could put a bunch of “Nelly Ohr” all over mainstream coverage of the impeachment. Old fashioned SEO manipulation in a new light.

Fister: discussion of the tautology underneath this: you trust the sources you trust b/c you trust them. People create self-reinforcing webs of trust by consulting multiple sources of the same bent. Students are also interested in talking about how algorithms work, including for sentencing people to prison; tie that to traditional values/understanding of how we make knowledge.

Tripodi: in response to comment on similar dynamic on doctor/patient relations: when people search “autism treatment” they are more likely to see non-evidence-based treatments, because doctors with evidence-based treatments are not using YouTube. Has a student who is trying to create a lexicon for doctors to tell people “research these treatments”—you can’t tell them not to search, but you can give them phrases that will return good quality content. Also important to make good quality content for evidence-based treatments. People are looking on YT; have to be there.

Zuckerman: that requires auditing the platform; YT is not that hard to audit, but FB is when it directs you to content.