Content area
Full Text
Ethics Inf Technol (2013) 15:209227 DOI 10.1007/s10676-013-9321-6
ORIGINAL PAPER
Bias in algorithmic ltering and personalization
Engin Bozdag
Published online: 23 June 2013 Springer Science+Business Media Dordrecht 2013
Abstract Online information intermediaries such as Facebook and Google are slowly replacing traditional media channels thereby partly becoming the gatekeepers of our society. To deal with the growing amount of information on the social web and the burden it brings on the average user, these gatekeepers recently started to introduce personalization features, algorithms that lter information per individual. In this paper we show that these online services that lter information are not merely algorithms. Humans not only affect the design of the algorithms, but they also can manually inuence the ltering process even when the algorithm is operational. We further analyze ltering processes in detail, show how personalization connects to other ltering techniques, and show that both human and technical biases are present in todays emergent gatekeepers. We use the existing literature on gatekeeping and search engine bias and provide a model of algorithmic gatekeeping.
Keywords Information politics Bias Social ltering
Algorithmic gatekeeping
Introduction
Information load is a growing problem in todays digitalized world. As the networked media environment increasingly permeates private and public life, users create their own enormous trails of data by for instance communicating, buying, sharing or searching. The rapid and
extensive travelling of news, information and commentary makes it very difcult for an average user to select the relevant information. This creates serious risk to everything from personal and nancial health to vital information that is needed for fundamental democratic processes. In order to deal with the increasing amounts of (social) information produced on the web, information intermediaries such as Facebook and Google started to introduce personalization features: algorithms that tailor information based on what the user needs, wants and who he knows on the social web. The consequence of such personalization is that results in a search engine differ per user and two people with the same friends in a social network might see different updates and information, based on their past interaction with the system. This might create a monoculture, in which users get trapped in their lter bubble or echo chambers (Sunstein 2002, 2006; Pariser 2011b). Social media platforms, search and...