Human-centric AI news and analysis

This article was published on February 3, 2021


Amazon’s search algorithm spreads vaccine disinformation, study finds

University of Washington researchers called the platform "a marketplace of multifaceted health misinformation"

Amazon’s search algorithm spreads vaccine disinformation, study finds


Thomas Macaulay
Story by

Thomas Macaulay

Writer at Neural by TNW — Thomas covers AI in all its iterations. Likes Werner Herzog films and Arsenal FC. Writer at Neural by TNW — Thomas covers AI in all its iterations. Likes Werner Herzog films and Arsenal FC.

Amazon’s reputation for spreading health misinformation has led critics to call it “a dystopian bookstore.”

Their concerns have escalated since COVID-19 unleashed a wave of anti-vaccination zealots.

According to a new study from the University of Washington, the world’s leading online retailer is amplifying the hysteria.

The researchers found that Amazon’s search algorithm promotes books endorsing false claims about vaccines over those that debunk them.

They also discovered a filter-bubble effect that leads users who engage with misinformative products to be recommended more of the same.

The study describes Amazon’s search results as “a marketplace of multifaceted health misinformation” and the platform’s recommendations as “problematic echo chambers.”

[Read: How much does it cost to buy, own, and run an EV? It’s not as much as you think]

The researchers first investigated the amount of misinformation returned by Amazon’s algorithms.

They tested a list of 48 search queries about 10 popular vaccine-related topics, such as “HPV vaccine” and “immunization,” without logging into the platform.

The results were then annotated as promoting, neutral, or debunking health misinformation.

They found that products promoting misinformation consistently outperformed those that debunk false claims.

Study authors Prerna Juneja and Tanu Mitra said over 10% of the results they received promoted misinformation health products.

An echo chamber effect

The researchers also examined whether Amazon pushes users to more misinformation after they’ve shown an interest in it.

They did this by performing typical user actions, such as clicking on an item.

They said this revealed evidence of a filter-bubble effect in Amazon’s recommendations:

Just a single click on an anti-vaccine book could fill your homepage with several other similar anti-vaccine books. And if you proceed to add that book [to] your cart, Amazon again presents more anti-vaccine books, nudging you to purchase even more problematic content.

The study authors describe this process as ” a problematic recommendation loop.”

Figure 1: (a) Amazon homepage recommendations. (b) Pre-purchase recommendations displayed to users after adding a product to cart. (c) Product page recommendations.
Credit: Prerna Juneja and Tanu Mitra
From left to right: Amazon homepage recommendations;, pre-purchase recommendations displayed to users after adding a product to cart; product page recommendations.

The researchers suggest several ways that Amazon can tackle the problem: display informative links alongside the product listings, introduce bias indicators, and modify its algorithms and policies to stop promoting misinformative books.

“There is an urgent need for the platform to treat vaccine and other health-related topics differently and ensure high-quality searches and recommendations,” they said.

You can read their full study paper here.

Also tagged with