Independent Online

Thursday, November 30, 2023

View 0 recent articles pushed to you.Like us on FacebookFollow us on TwitterView weather by locationView market indicators

Amazon algorithms boost vaccine misinformation, says study

The logo of Amazon is pictured inside the company's office in Bengaluru, India. File picture: Reuters/Abhishek N. Chinnappa

The logo of Amazon is pictured inside the company's office in Bengaluru, India. File picture: Reuters/Abhishek N. Chinnappa

Published Jan 29, 2021


New York - Amid growing concern that e-commerce platforms are amplifying vaccine-misinformation, a new study by researchers at University of Washington has found that Amazon hosts a large number of misinformative products belonging to categories such as books, ebooks, health, personal care and apparel.

The results led to a dataset of 4,997 Amazon products annotated for health misinformation, said the study published at preprint repository

For the study, Prerna Juneja and Tanushree Mitra, both from The Information School at the University of Washington, conducted two-sets of algorithmic audits for vaccine misinformation on the search and recommendation algorithms of Amazon.

Their aim was to empirically determine the amount misinformation returned by its search and recommendation algorithm.

They also investigated whether personalisation due to user history plays any role in amplifying misinformation.

So, they did a systematic audit search-results belonging to vaccine-related search-queries without logging into the platform -- unpersonalised audits.

They found 10.47 per cent of search-results promote misinformative health products on the platform.

"We found that search results returned for many vaccine-related queries contain large number of misinformative products leading to high misinformation bias," the researchers said.

Moreover, misinformative products are also ranked higher than debunking products, the results said.

Next, they analysed the effects of personalisation due to account-history, where history is built progressively by performing various real-world user-actions, such as clicking a product.

"Our study also suggests presence of a filter-bubble effect in recommendations, where users performing actions on misinformative products are presented with more misinformation in their homepages, product page recommendations and pre-purchase recommendations," the researchers said.

The findings come amid criticism that Amazon has not been successful in properly regulating health-related products on the platform.

But in a recent blog post, Amazon said that during 2020, it reviewed almost 10,000 product listings and removed over two million products for violating its offensive or controversial guidelines, VentureBeat reported.


Related Topics: