A proposal of a relative weighted online 5-star rating system as a way to mitigate online reviews biases

Abstract

Online ratings and reviews can be considered an electronic word of mouth regarding the quality of goods, products, or services. Reviews provide crucial information for customers, therefore significantly influencing their behavior, and they enable feedback to businesses from their customers, increase visibility, drive sales, help in developing a brand and building trust and reputation among consumers. However, the current 5-star rating system currently used on many Internet platforms such as Amazon or TripAdvisor suffers several drawbacks (biases) such as sentiment bias, polarization bias, non-discrimination bias, or vocal minority-silent majority bias. Therefore, the aim of the paper is to propose a new (weighted) relative 5-star rating system which takes into account reviewers’ history (in the form of the average and variance of the past reviews) and transforms absolute aggregate ratings into relative ones, thus providing less biased information for consumers and businesses. In particular, the proposed system reduces sentiment bias and non-discrimination bias. Moreover, the proposed approach enables to reduce the influence of ratings made by bots or dishonest evaluators-humans. The real-world application of the proposed approach dealing with ratings of selected attractions in Madrid area is provided as well.