INTRODUCTION
Algorithms have come to shape our daily lives and realities. They change the perception of the world, affect our behavior by influencing our choices, and are an important source of social order.
Algorithms on the Internet have significant economic implications in newly emerging markets and for existing markets in various sectors. A wide range of our daily activities in general and our media consumption in particular are increasingly shaped by algorithms operating behind the scenes: the selection of online news via search engines and news aggregators, the consumption of music and video entertainment via recommender systems, the choice of services and products in online shops and the selection of status messages displayed on social online networks are the most prominent examples of this omnipresent trend. Algorithms suggest friends, news, songs and travel routes. Moreover, they automatically produce news articles and messages, they calculate scorings of content and people, and are employed to observe our behavior and interests as well as to predict our future needs and actions. By assigning relevance to certain pieces of information they keep consumers, companies and authorities from drowning in a growing flood of information and online data. At the same time, they mine and construct realities, guide our actions and thereby determine the economic success of products and services. Algorithms form the techno-functional basis of new services and business models that economically challenge traditional industries and business strategies. These economic changes and challenges are accompanied by and interact with significant social risks such as manipulation and bias, threats to privacy and violations of intellectual property rights that compromise the economic and social welfare effects of algorithmic selection applications.This rapidly growing Internet phenomenon is here called ‘algorithmic selection’.
It is a central and structuring bundle of Internet innovations in digital economies. Algorithmic selection is embedded in a variety of Internet-based services and is applied for numerous purposes. Although their modes of operation differ in detail, all of these applications are characterized by a common basic functionality: they automatically select information elements and assign relevance to them. This common feature defines the properties of algorithmic selection and allows a formal distinction from other Internet phenomena such as Web 2.0 (O’Reilly, 2007), the Internet of Things (Ashton, 2009; Mattern and Florkemeier, 2010) and big data (Feijoo et al., Chapter 25 this volume).The development of algorithmic selection is closely related to a number of techno- economic and social trends in information societies, including computerization, big data, personalization, automation and economic optimization. In essence, its diffusion and
395
growing importance is fueled by the combination of ubiquitous computerization and the proliferation of an increasingly mobile Internet. In a growing number of economic and social domains, the spread of algorithmic selection is driven by the diffusion of online information, communication and transactions. Computers and the Internet serve as enabling technologies that provide the infrastructure - the technological and functional precondition for a wide range of applications. At the same time, ubiquitous computerization and Internet use generate additional demand for algorithmic selection, because they result in a massive proliferation of data volumes and a growing need for orientation by selection. This (big) data forms the raw material (World Economic Forum, 2011a) for algorithmic selection, creates economic opportunities and calls for data/reality-mining tools in order to harness the economic opportunities. Altogether, the combination of technological, data-based opportunities and economic demand for selection is a major driver for the establishment of new industries, applications and business models, where automation of data processing plays a central role, and algorithmic selection perfectly supports business strategies, especially in terms of process optimizations.
Automated algorithmic selection advances optimizations in various ways: faster processing of larger amounts of data by automation; cost reductions in production and transmission by automation of data processing; strategic enhancements by increased data-driven, evidencebased decision-making (McAfee and Brynjolfsson, 2012); and personalization by mass customization of products and services that are tailored to meet diverse consumer needs.With a high potential for economic improvement, algorithmic selection services are spreading fast in a wide range of industries. As illustrated by the Organisation for Economic Co-operation and Development (OECD, 2013) for big data, their diffusion is especially high in sectors characterized by a high degree of digitization and high data intensity. Accordingly, it already plays a major role in industries that rely heavily on digital production and online transmission such as Internet search, news, advertising, entertainment and social online networks. Further, algorithmic selection has gained importance in areas such as retail, trade, the stock exchange, banking, insurance, politics, security, intelligence, transportation, logistics, science, education, health and employment (Latzer et al., 2014). Given the combination of ubiquitous computerization, rapidly growing amounts of available data, and economic pressure for optimizations, the trend towards increased algorithmic selection in a rising number of domains seems to be irreversible. This provides the starting point and the rationale for more in-depth analyses on the characteristics, role and consequences of algorithmic selection for markets and societies.
Most social science research on algorithms has focused on search engines (Varian, 2006; Machill and Beiler, 2007; Lewandowski, 2012; Konig and Rasch, 2014) and recommendation systems (Resnick and Varian, 1997; Senecal and Nantel, 2004; Klahold, 2009; Jannach et al., 2011; Ricci et al., 2011; Robillard et al., 2014).
This chapter extends the scope of analysis and provides a comprehensive overview, with a special focus on how to think economically about algorithmic selection. It explores the characteristics and implications of an increasingly adopted technology on the Internet that automates nothing less than the commercialization of reality mining and reality construction in information societies. The following questions are tackled: How can the plethora of algorithmic selection applications on the Internet be analytically grasped and categorized? How does algorithmic selection operate and where is it applied? What market structures and business models are evolving and how do they affect existing media markets? What are the major social and economic benefits and risks of algorithmic selection, and what governance choices are available to minimize risks and thus maximize economic and social welfare?The chapter is organized as follows. The next section offers a typology of applications based on algorithmic selection and provides a basic input-throughput-output model in order to show the functioning and economic purposes of the different types of algorithmic selection. Section 19.3 explains the theoretical perspective applied for its analysis. Section
19.4 presents results from market analyses and shows the different phases of markets for applications using algorithmic selection, explores their structures and explains concentration tendencies. Section 19.5 provides insights into business models of algorithmic selection with an emphasis on value proposition, value creation and revenue streams. Section 19.6 examines selected implications of algorithmic selection for traditional media markets and the incumbents’ profitability. Section 19.7 identifies areas of risk, such as possible violations of basic rights, the effects of algorithmic selection on perceptions of the world and potential impacts of algorithmic selection on human development. Finally, section 19.8 summarizes regulatory challenges and discusses opportunities and limitations of available governance choices such as market solutions, self-regulation and state intervention. Section 19.9 draws conclusions about the economics of algorithmic selection.
19.2