Search (Non-)Neutrality and Impact on Innovation: Short talk

The search neutrality debate, as a parallel to the network neutrality debate, is raging worldwide, with search engines accused of biasing the ranking of their organic links to provide a competitive advantage to their own content. Search neutrality has been and is still discussed by governments and regulators, such as in the US by the Federal Trade Commission and in a Senate hearing, or in Europe where Google is facing a $6 billion fine. In a recent paper [1], we have designed a model and determined the optimal ranking policy for a search engine as a trade-off between short-term revenue (based on the potential immediate gain from high-ranked links) and long-term revenue (based on the satisfaction of users due to the relevance of the ranking). We here apply this model to investigate on an example whether non-neutrality impacts innovation. We illustrate that a revenue-oriented search engine may indeed deter innovation at the content level, hence the validity of the argument (without necessarily meaning that search engines should be regulated). The search engine (SE) model and the corresponding results (revenue-maximizing ranking policy) are described in full detail in [1]. A summary of the model and a description of how it is used to develop further the content innovation aspect is in [2]. We consider that among 10 pages corresponding to a given request–offered by different content providers (CPs), one exactly (say, page number 1, or CP 1, for simplicity) is served directly by the SE while the others are served by third parties. To consider the impact on innovation, we assume that CP 2 invests in quality and manages to improve its relevance distribution. The summary of our numerical results in [2] are illustrated in Figure 1: • For a neutral ranking, CP 2 logically makes more revenue than the other CPs, since it regularly gets higher ranking. However, when the SE increases the weight of the immediate gain in the ranking (non-neutral revenue-maximizing ranking), CP 1 becomes the one with highest revenue, despite its (stochastically) lower relevance. • We also take the perspective of CP 2, and compute its optimal decision. CP 2 invests in quality to modify its rele-