SOCIAL RISKS

Algorithmic selection and attendant personal data collection have become objects of public concern and have raised questions about their impact on society as well as the need for public policy.

Generally, the assessment of risks is an appropriate method to relate estimated economic and social benefits to risks, for example the benefits gained by search engines in managing information overflow versus the risk to user privacy. A first step in such analysis is generally to identify possible risks and benefits and assess the probability of their occurrence and the number of people affected. For example, how many people or institutions use an algorithmic application, do these people or institutions have a multiplying effect, and how often and how intensively do people/institutions use the application?

The various risks of algorithmic selection applications found in the literature are here grouped in three overlapping categories, which in particular indicate that such analysis not only touches upon cost-benefit calculations but also extends into ethical/moral value judgments: (1) threats to basic rights and liberties, (2) impacts on the mediation of reality, and (3) challenges to the future development of the human species. Overall, eight specific risks can be distinguished that accompany the diffusion of algorithmic selection: (1) manipulation, (2) diminishing variety, the creation of biases and distortions of reality, (3) constraints on the freedom of communication and expression, (4) threats to data protection and privacy, (5) social discrimination, (6) violation of intellectual property rights, (7) possible transformations and adaptations of the human brain, and (8) uncertain effects of the power of algorithms on humans, for example growing independence of human control and growing human dependence on algorithms.

Empirical examples of manipulation are ‘Google bombs’ (e.g., Bar-Ilan, 2007), described as planned massive influence on search results, or the improvement of websites through search engine optimization that improperly tries to increase the attention achieved.

Manipulations have also been identified for recommender and reputation systems for goods and services such as hotel or product recommendations (e.g., Rietjens, 2006; Schormann, 2012). Algorithmic selection is furthermore associated with bias inasmuch as it is presumed to develop an algorithmic reality where content is only visible when it is produced and shaped according to the rules that algorithmic selection prescribes (Zhang and Dimitroff, 2005; Cushing Weigle, 2013). The rules themselves leave out certain aspects of reality and have incorporated specific values that unknowingly discriminate against particular content. Qualified empirical evidence for this phenomenon is rare, but various authors have discussed the self-enforcing mechanisms of algorithms and their biasing effects (e.g., filter bubble; Pariser, 2011), or the creation of a digital divide on a content and usage level (Segev, 2010). Accordingly, the much-discussed media realities that are being formed by the gatekeeping function of traditional mass media reach a new level, leading to discussions about algorithmic realities that follow different, increasingly automated, personalized and commercialized rules (Just and Latzer, 2016).

Constraints on the freedom of communication are also identified as a possible risk of algorithmic selection - an argument derived largely from its technological design. As the name implies, it has a selective element that can be shaped, with differing effects, however. On the one hand it can be used to gain access to relevant content or to protect IP rights (Wunsch-Vincent, Chapter 11 this volume) or to keep children from accessing harmful content (Hinman, 2005). On the other hand, algorithmic selection may be adopted to diminish the democratic potential of digital media by being used for censorship (Zittrain and Palfrey, 2008).

To fulfill their role as information intermediaries and information brokers, algorithmic selection applications have to rely on content produced by third parties and on data produced by consumers. Both sources of information involve certain risks.

It is argued that, without infringing the intellectual property rights of content producers and distributors, many applications such as search engines, information aggregators or recommender systems would have no data basis on which to build their services (Stuhmeier, 2011). This kind of use of third-party content has led to disputes over copyright and other intellectual property rights, and publishers all over the world have sued Google for infringing such laws (Clark, 2010, 2012; Chiou and Tucker, 2013; Quinn, 2014). Moreover, many algorithmic selection applications are personalized/customized applications, that is, applications that use data collected from the users to personalize results. This incorporates great risks concerning users’ privacy and data protection (Brown, Chapter 12 this volume). Today, personal data has become the new oil for the economy (World Economic Forum, 2011b) and operators of algorithmic selection applications are major collectors of such data online. They use these data to customize services and monetize them (as an exchange for other/more data or by selling them directly) - activities that have resulted in various data privacy challenges (Chaleppa and Sin, 2005; Zimmer, 2008; Xu et al., 2011; Toch et al., 2012). Algorithmic applications also raise debates concerning their influence on human cognitive abilities - a pressing object for future research. Current discussions range between questions of whether these applications result in the loss of abilities (Carr, 2010; Henig and Henig, 2012) or whether they are simply helping in allocating cognitive resources more efficiently, like other technologies in history (Sparrow et al., 2011). Finally, there is a general discussion on how the relationship between humans and algorithms can be described and how this human-machine relationship will develop or should be shaped in the near future (Bunz, 2012; Schirrmacher, 2013). This includes questions about the power of algorithms, about whether humans are still able to control them or to what extent they control human behavior and development.

There are economic motives that promote major risks such as manipulation, threats to privacy or the infringement of IP laws. These motives are mainly predicated on efforts to maintain and amplify market power, for example by prioritizing one’s own services in search results and excluding others - a concern that has raised discussions of whether search results should be subject to a search neutrality principle (Lao, 2013), for example. Systematic manipulation is said to be mainly applied where goods, services and information are sold, or where trust in transactions needs to be built (e.g., deceptive recommendations). Major groups affected are search engines and recommender systems. New markets of manipulation evolved around algorithmic applications, provided, among other things, by search engine optimization and marketing agencies as well as web content production agencies. In the meantime, they have become a vital and essential branch of the rapidly growing e-commerce sector.

Altogether, the production of economic wealth by algorithmic selection co-evolves with the emergence of social risks. Algorithmic selection leads to a commercialization/ economization of automated reality mining and construction. The construction of realities - well known from research on traditional media - is not only automated by algorithmic selection and extended to further aspects of life but at the same time increasingly oriented on economic and less on social rationales (Just and Latzer, 2015). As a consequence of these increasingly automated and commercialized mining and formations of realities on the Internet, certain forms of governance seem to be necessary and are being discussed in order to realize the economic and social welfare goals anticipated by algorithmic selection.

19.8

<< | >>

↑

Source: Bauer J., Latzer M. (Eds.). Handbook on the Economics of the Internet. Edward Elgar,2016. — 603 p.. 2016

More economic literature on Economics.Studio

SOCIAL RISKS

More on the topic SOCIAL RISKS: