Tuesday, June 18, 2013
Search Neutrality and Search Bias: An Empirical Perspective on the Impact of Architecture and Labeling
Posted by D. Daniel Sokol
David A. Hyman, University of Illinois College of Law and David J. Franklyn, University of San Francisco School of Law discuss Search Neutrality and Search Bias: An Empirical Perspective on the Impact of Architecture and Labeling.
ABSTRACT: As Google has moved from providing “ten blue links” to “universal search,” controversy has erupted over whether Google is favoring its own specialized search results over competing specialized results offered by other entities. Google’s competitors have complained about “search bias,” and demanded that antitrust enforcers should ensure “search neutrality.” The U.S. Federal Trade Commission included search bias in the issues it considered as part of its multi-year investigation of Google, but it closed the investigation in January, 2013 without taking any action on that issue. However, the European Commission recently identified “the way Google displays links to its own specialised search services” as the first of four “competition concerns” it had with Google’s business practices. Google recently proposed to settle this dispute by providing prominent links to three rival specialized search services (architectural remedy) and more clearly labeling its specialized search results (labeling remedy). We empirically examine the significance of search output architecture and labeling on consumer knowledge and click-through behavior in two online surveys/simulations. We find that the architecture of the search results page is far more important than any labels that might appear on that page. User awareness of labeling is low, and even labels far more explicit than those currently employed do not have much impact. Consumers appear to have quite sticky expectations about how search results are presented, and their click-through behavior tracks those expectations, irrespective of how the search results are labeled. These findings suggest that the impact of architectural remedies will depend greatly on their design features, while labeling remedies are unlikely to have a significant impact.