Recommendation systems and convergence of online reviews: The type of product network matters!

“The Internet has given word-of-mouth (WOM) a new significance by allowing individuals to express their opinions and thoughts to a global audience, and so, it is an essential aspect of e-commerce [3]. Customers can simply express their opinion in an electronic form of WOM, and others can adopt such opinions in their decision making regarding purchasing products. In this regard, opinion convergence/divergence, as a social phenomenon, has drawn the attention of the stakeholders in ecommerce in recent years. In general, opinion convergence refers to the situation in which customers’ opinions coincide with regard to a product. In contrast, opinion divergence is the situation in which consumers have varying opinions about the product and could be extremely opposite [4]. Whether it is the convergence of ratings, reviews, or comments for a focal product, or the convergence of ratings, reviews, or comments between a pair of connected products in a product network in an online store [5], research has shown promising insights on the importance and utility of convergence/divergence of opinions for business. As suggested by the literature, with having a potential role in consumers’ decision making [4], deriving or diminishing sales [5], and in helpfulness and informativeness of reviews on an online platform [6,7], yet there is only a handful of research studies looking at the phenomenon of convergence/divergence of opinions in e-commerce.

Although there are only a handful of widely used recommender algorithms, all online stores apply them in different ways and often adapt them to their needs [9]. A majority of e-commerce stores frequently use multiple recommender systems to support their customers in their purchasing decision. Amazon, for instance, extended their initial recommender list to include two additional ones (i.e., ‘Frequently bought together’ and ‘Customers who viewed this item also viewed’). Moreover, each recommender system is trained differently to recommend different groups of products. Each recommender’s type or goal is a determinant of the training procedure and the features it takes as input attributes. Therefore, the nature of a product network (i.e., the network of products that are connected with hyperlinks due to recommendations) for each type of recommender system differs, and this differentiation is shown to be a potential determining factor for WOM, such as online reviews in terms of product ratings [5], and, possibly, for the sales quantity of recommended products [10].”