The future of product recalls: AI and Amazon

Photo (c) gustavofrazao - Getty Images

A new technological approach could benefit consumers

Consumer goods can be recalled for any number of reasons, but the process is primarily used to protect people from potentially harmful products. 

Now, a group of researchers from Boston University School of Medicine have upped the ante. The team has created an artificial intelligence (AI) program named BERT (Bidirectional Encoder Representation from Transformations) that can identify items that have yet to be pulled from shelves but may require being recalled. The system is based partly on Amazon customer reviews.

“Health departments in the U.S. are already using data from Twitter, Yelp, and Google for monitoring foodborne illnesses,” said researcher Dr. Elaine Nsoesie. “Tools like ours can be effectively used by health departments or food product companies to identify consumer reviews of potentially unsafe products, and then use this information to decide whether further investigation is warranted.” 

Changing the game

BERT fits into the consumer landscape as another tool to be used by regulators to keep tabs on items that could be potentially harmful. The system could be a boon to an agency like the Food and Drug Administration (FDA), which is tasked with informing consumers about defective products.

The beauty of technology like BERT is that it can quickly pore through databases and obtain necessary information that could save a lot of consumers from buying or ingesting products that shouldn't even be on shelves. 

To test the effectiveness of BERT, Dr. Nsoesie and her team utilized customer reviews on Amazon after allowing BERT to get a better understanding of words that would indicate a recall is necessary, such as “rotten,” “sick,” “ill,” etc. 

After going through nearly 1.3 million product reviews that were cross-referenced with FDA reports, BERT was able to pick out nearly 75 percent of products that needed to be recalled. The system also picked out additional products -- over 20,000 -- that contained words or terms that were associated with recalls but hadn’t yet been pulled from store shelves, emphasizing that many products may be flying under the radar and negatively affecting consumers. 

The researchers believe technology like BERT can spread the word of recalls more efficiently and help save consumers undue contamination from harmful products. 

“Our approach can improve food safety by enabling early identification of unsafe foods which can lead to timely recall thereby limiting the health and economic impact on the public health,” the researchers wrote.

To stay up to date on the latest recalls, click here

Take an Identity Theft Quiz. Get matched with an Authorized Partner.