A new computer program has been developed to help consumers spot fake online product reviews, giving them the power to determine the true nature of a review.
The computer program identifies genuine or fake reviews by determining whether one-star to five-star ratings distribution appears natural or not.
The computer program was developed by computer scientists at Stony Brook University in New York.
Using an adapted graph for wild animals, showing how their numbers fall and rise depending on predators and food available, the scientists made some interesting observations. When adapted to the number of one-star through five-star online reviews, some distinctive graph shapes emerged.
However, the use of fake reviews skews the graphs. This in turn allows companies to scan for paid for glowing reviews. They therefore pay people to write them glowing reviews that don’t necessarily present a truer picture.
Whereas review sites have done whatever possible to eliminate fake posters, their efforts haven’t been as successful.
The researchers made interesting findings using 4000 hotel reviews from TripAdvisor and more than 700,000 Amazon products.
Unusual graph shapes raised some level of suspicion. For instance, whereas there are cases in which one-time consumers left rave reviews for certain products with low average star ratings, there were other one-time reviewers that left rave reviews for products with the highest average reviews. That might signal individuals who basically logged in to leave a review for money and left.