An increasing number of consumption decisions happen on two (or more) sided platforms that carry goods and services from a large number of suppliers. In principle, such easy access to a multiplicity of products should lower consumers' search costs and thereby enhance their decision-making. In practice, consumers can struggle with such a large choice set. Recommender systems (RS) are designed to solve this problem by suggesting relevant products based on a consumer's preferences. This can be good both for consumers and for effective competition between suppliers. However, as has been amply shown within the computer science literature, RS do not necessarily have the ability or incentive to carry out this role perfectly. Even if RS are intended to be consumer-centric, they tend to exhibit inherent biases in the recommendations made. These are associated with the choice of RS model design, the data that feeds into the RS model, and feedback loops between these two elements. The key contribution of this paper is in highlighting that, since these biases can be expected to change consumption decisions, they can in turn distort competition between suppliers, potentially creating barriers to entry and expansion, increasing concentration, and reducing variety and innovation. This can happen even in the absence of any malicious intent from the platform, but the situation may be worsened if a platform's own interests diverge from those of consumers and this is reflected in the RS design. This paper identifies and outlines these important effects at a high level and in doing so provides a trigger for future research.
|Number of pages||26|
|Publication status||Published - 24 Mar 2022|