Today, most people when forced to make a decision between options interact with some review mechanism. This may be the result of an active decision, such as by seeking out review sites, or a passive decision, such as by being shown ratings under products or using platforms with recommendation algorithms. The success of rating-based system in lowering search costs for customers and consumer is why it has been applied in many areas: pieces of art (GoodReads for books, RateYourMusic for music, Letterboxd and IMDb for movies), products (Amazon, eBay, most clothing stores), gig worker performance (Uber, Lyft, DoorDash), physical locations (Google, Yelp), hospitality (TripAdvisor, AirBnB).
Ratings on review sites don’t just tell us things about the reviewed, but also the reviewer. One can measure how disagreeable and independent-minded a person is by how far their ratings deviate from the average ratings on review sites. If a person rates items similar to other people on review sites, they tend to be more agreeable and conventional-minded. Reviews from these type of individuals provide less information, and hence should be weighted less on review sites.
The average ratings of a product, location, experience, or piece of art influence how likely people are to consider buying, using, or consuming it. Given the amount of choices that users and consumers are faced with today, most people stick to the “safe choice” with high ratings. This, on average, saves people both money and time. Reviews are also likely to influence most people’s perception of the item’s quality. For example, people might quit watching–or not watch at all–a movie with a bad IMDb score earlier than a movie with a good score.
The problem of circularity of review sites is caused by peer influence as well as the path dependence of most review aggregation algorithms. Something that has been rated well in the past is more likely to be given a chance–as well as more leeway–from users. Amazon customers are unlikely to even consider purchasing products with a rating lower than 4 stars. In fact, Amazon removed all filters for customer reviews except for 4 stars and up.
Items that received good reviews early on are likely to become popular and receive favorable reviews in the future. This can be make-or-break for a business, piece of art, or gig worker. Often, the only way to get rid of a bad rating is to “start over”, that is, going bankrupt, recreating an account, or rebranding the product. One way to resolve this is to make ratings time-dependent, so that recent ratings are weighted more heavily than older reviews. But this is not perfect because individuals are still subtly influenced by the current rating. Another method is to hide the rating, and then allow for some perturbations in the recommendation algorithm. YouTube is a good case study: Google officially stopped showing the dislike count on videos in 2021. This decisions received significant negative reaction at first, but was eventually accepted by users. For YouTube, forcing users to form an opinion of a video by watching it, rather than looking at the like bar, is beneficial because it increases the users session duration, and thus Google’s advertising revenue. Companies with sales- rather than advertiser-based business models tend to try to reduce time spent searching, so will likely make greater use of reviews.
In general, the circularity of rating systems favors customers in the short-term, but may harm them in the long-term. For the owners of the platforms, reviews “discipline” the products to focus on providing good quality and service. Reviews can be seen as a form of non-monetary gratuity. But the accumulative nature of reviews on many platforms may also create winner-takes-all effects. As with monopolies among businesses, these winner-takes-all effects may prevent new entrants from entering the market. This decrease in competitive pressure may lead to a slight decline in quality over time. However, since ratings are to some extent self-correcting, the more likely effect is for quality to stagnate–rather than decline–as incumbents don’t have a strong incentive to innovate and improve on their already highly-rated products.
Another big problem with review sites is the incentive to game the system. Such deceptive strategies may range from outright fake reviews to slight nudges to get one’s customers or users to leave reviews. Here, again, algorithmic changes can mitigate problems by lowering the weighting of users that are unverified, exhibit abnormal review behavior, and have a less extensive review history.
The wide adoption of rating systems is a sign that they are generally well-liked–whether by the customer or the producer. They will continue to be used, which is why it is important to think about ways to mitigate the negative side effects of their use.
As an individual, freeing yourself from depending on reviews–or at least being conscious when relying on them–can help you develop a stronger perception of your personal taste. At first you might think that liking an item that is poorly rated means that you have “bad taste”. This may be the case, but it’s also important to note that ratings are averages. Products may have high variance so that they polarize people into giving 1 star or 5 stars. A certain product or piece of art may also uniquely fit your current needs or experiences. Our tastes are subjective and changing. There is no objective ground truth of things that will always be good or bad–even though review sites might seem to provide such a ground truth.