Fabricated Reviews on Amazon and Information Cascades
Recently in the course we have discussed the concept of an information cascade, a phenomenon in which people make decisions sequentially, and are influenced more by others’ decisions as opposed to their own private information when making a decision. In particular, people infer the private information of prior decision-makers based on their choices. As we have discussed in class, this inferred information can sway an individual’s decision away from using their private information.
Such a concept is extremely pertinent in online reviews. As opposed to the majority-red/majority-blue marble bag experiment discussed in lecture, in which the information cascade has some sense of optimality, as justified by calculating the probabilities using Bayes’ Theorem, the same may is not necessarily true for online reviews.
Zachary Crockett, in his Forbes’ article, 5-star phonies: Inside the fake Amazon review complex, discusses this very issue. It details that although many consumers give considerable weight to public reviews, such a choice can often be ill-informed. He explains that online sellers on Amazon employ various tactics to fabricate improved ratings and positive reviews on their products. One common tactic is to find people on Facebook groups who will purchase the product as well as leave a glowing review, in exchange for a full refund on said product.
This problem is exacerbated by the fact that Amazon’s search algorithm also uses the information of average product review ratings and average seller review ratings to recommend products to prospective customers in a given search query. In this sense, the consumer receives three signals which influence one another – from the Amazon marketplace recommendation, the individual product review, and the seller’s reviews.
This information cascade is novel compared to what we have learned in class. In this situation there are a multitude of dependent events that act as signals, which complicate the analysis of the situation. As such, fake reviews influence an authoritative source, namely Amazon and their recommender systems. This misrepresents product qualities and thereby dupes many consumers.
This is a direct parallel to an information cascade – fabricated, positive reviews for a product lead to people inferring positive signals from previous purchasers, causing an increasing amount of people to purchase said product. This, in turn, drives up the seller’s average review, as well as the product’s position on Amazon’s recommendation given a relevant search. This, however, outweighs a consumer’s private signal about the product, which they can receive from, for example, examining the specification of said product, and hence furthers the effect of an information cascade.
Interestingly, increased awareness about illegitimate reviews may lead to consumers becoming more informed and in turn ignoring public reviews. Instead they may rely on private signals in order to make decisions, which could interestingly break such an information cascade. In the example of this article, fake reviews on Amazon has led to new review websites, that have stronger checks to ensure the authenticity of the reviews, and such websites could still lead to an information cascade, but hopefully one with a more rational basis.