Amazon To Use AI To Overhaul Review System
In order to prevent fake reviews and product star ratings, Amazon has turned to artificial intelligence. A new artificial intelligence machine learning system that has been custom built by the online marketplace will help to boost ‘real’ reviews. These include reviews from verified purchasers, reviews that have been noted as helpful and more recent reviews.
Amazon is hoping that the system will allow accurate opinions to be at the top of review pages and these will be used to create the star rating for individual products. Up until now, the star ratings took all the reviews into account and used them to create an overall picture of the opinion of the product. However, this enabled fake reviews to influence the star rating even though verified and real reviews had talked damningly about a product.
Amazon has said that the algorithm will improve over time, as happens with machine learning when computers are refined. The new system was put in place last week but it may take some time for the effects to be seen on the site. Because online shoppers can’t see a product, hold it or try it out before they buy it, they rely on reviews from users and previous buyers to make purchasing decisions. This means that product reviews are highly trusted by online shoppers and good reviews can entice shoppers to make a purchase that they may not otherwise have considered.
Because of this, online sellers and marketers have begun trying to inflate the star ratings of their products, especially if it is new to the market or new to a specific online site. These unscrupulous sellers often post false reviews themselves that praise the product, or they may pay other users to leave the good reviews on the product page.
This practice is called astroturfing, and is common in the media as well as being used by political parties during election time too. Astroturfing is common on lots of websites but Amazon has been hit particularly hard with the practice as it is the world’s biggest online marketplace and many buyers turn to the site when searching for reviews of a product or to compare several similar items.
According to Amazon, its new artificial intelligence will be able to tell the difference between original products and those that have been changed to address issues or complaints from customers when they have been sold through the same listing. Because of this, newer reviews will be given more weight to better reflect the new and improved product, rather than the flawed product which had been sold previously. This means that older reviews will be pushed to the bottom and may not be taken into account when a star rating is being considered.
This may also work in the opposite way- products which used to work perfectly and received good reviews but have been changed and now contain flaws will likely be rated on new reviews, rather than older reviews of the previous model.
Astroturfing doesn’t just happen with online stores and marketplaces- travel sites can be heavily hit by the practice in which hotels and attractions look to boost their online ratings by posting fake reviews, or soliciting others to post them. Speaking to the Guardian, TripAdvisor’s James Kay said “We have a whole team dedicated to the problem of fake reviews. We use both automated systems and a dedicated team to review reviews and weed out fake entries”.
Because the problem of fake reviews and astroturfing is becoming more and more prevalent, online sites that present reviews to users will have to start using intelligent systems in order to maintain user trust. In April of this year, Amazon filed lawsuits against websites that sell fake reviews to companies as it said that these sites were trying to make money from duping the reviews system that users should be able to trust.
Amazon’s new system is likely to be monitored by competitors and other online stores and marketplace in order to gauge its effectiveness. If it proves a success then it may help to end the problem of fake reviews and astroturfing across the world of online shopping.