UW study finds Amazon promotes vaccine lies, especially to already misinformed users

Findings from University of Washington researchers’ audit of Amazon algorithms point to a dire situation: The tech giant’s search engine promotes misinformation about vaccines.

The consequences, researchers fear, could prevent the United States from controlling the COVID-19 pandemic.

Achieving herd immunity requires the vast majority of the population to get vaccinated, said Prerna Juneja, the doctoral student at UW’s Information School who authored the study with Tanu Mitra, a UW professor of social computing.

If some people’s reluctance about the COVID-19 vaccine is reinforced through Amazon or social media, she said those companies have a social responsibility to intervene.

Many people don’t realize, Juneja said, that their search results and newsfeeds on websites and apps are personalized.

Companies are constantly farming data from users’ activity and feeding it into algorithms that present content they’ll pay attention to, not necessarily information that is accurate.

On Amazon, every time a shopper clicks on a product, that becomes data that influences future search results and recommended related products, she said.

“When that happens for something as small as a chocolate, that’s fine, I guess,” Juneja said. “But when it starts recommending things that affect the health of a person, you can’t apply the same logic.”

Based on their experiments, even if an Amazon shopper started from a blank slate, having never used the site before, search results still ranked books touting misinformation higher than debunking books, according to the study.

But if Amazon shoppers had shown a history of interest in untruthful health products, their search results would yield even more untruthful books about vaccines. Juneja and Mitra call this the “filter-bubble effect.”

“It’s like a rabbit hole,” Juneja said. “You start believing one kind of misinformation and as you engage with that, the algorithm starts recommending more misinformation.”

She said Amazon must be aware of the problem, because it’s in the media.

“They’ve been ignoring the problem,” she said.

Reporters had been pointing out that a search of the word “vaccine” pulls up books by conspiracy theorists before the researchers’ study was published, Juneja said.

Thursday, a search of the word “vaccine” on Amazon presents a book called “Anyone Who Tells You Vaccines Are Safe and Effective is Lying” within the first 10 recommendations.

The Netflix documentary “The Social Dilemma” also raised awareness about misinformation online, she said.

In the film, former and current social media industry leaders lay out how social media sites make money by getting views on ads.

Algorithms that pick up on what people focus on then present more of that content to keep people looking at their phones, the documentary argues.

The film points to a 2018 study published in Science that found fake news spreads six times faster than real news on Twitter because of this process.

The process is similar with YouTube, the focus of Juneja’s first study in a series of experiments about algorithms.

That study found once a user develops a watch history, a pattern of consuming conspiracies will cause more misinformation to be recommended to them. On Amazon, there’s also the issue of “sponsored” content, Juneja said.

“Now it’s like anybody can pay money to promote a product on the platform,” she said.

A large portion of the untruthful search results yielded in the study were sponsored.

Amazon had not publicly responded to the article’s publication on Facebook or Twitter as of Friday.

In March 2019, after pressure from Rep. Adam Schiff, D-Calif., Amazon pulled some books connecting autism to vaccines.

“Don’t blindly trust the search results the algorithm is showing you,” Juneja said. “You should learn to think critically about them. What comes up might be the most relevant, but it might not be the most trustworthy or credible.”