Content area
Full text
The online world is increasingly shaped by forces beyond our control. Algorithmic processing agents are used by a wide range of web publishers, online retailers, and social media companies to determine the kinds of stories that are featured to online readers, the advertisements that are targeted to online shoppers, and the search results they see, to name just a few of the ways in which these hidden programs predict the shape and content of our online experience. Beyond the world of advertising, algorithms are used to make decisions about insurance premiums, healthcare, education, housing, and employment.
Recently, information scientists, ethicists, and legal scholars have begun to identify the potential negative effects of algorithmic agents-such as the potential to enable discriminatory decision making-but there is little consensus on how to regulate the increasing use of algorithms. The marketplace has not yet settled on a supply-demand solution, and self-auditing and independent auditing of algorithms have proven to be technologically challenging, hampering the development of self-regulatory codes.
The other potential mechanism for regulating algorithms is through legal means. This article will explore the very different approaches emerging in the two largest legal systems to have addressed these issues: those in the United States and the European Union.
Potential Negative Effects of Online Algorithmic Agents
Many people believe that algorithmic processing tools are harmless robots that rely on objective criteria to render fact-based predictions about consumer likes and behavior. As information science grows as a discipline, however, scientists are increasingly recognizing that algorithm processing agents are not necessarily objective and can be used for discriminatory purposes or have the effect of magnifying discriminatory attitudes of online users.
The New York Times recently reported on a Carnegie Mellon University study that found that Google's online advertising system showed an ad for high-income jobs to men much more often than it showed the ad to women.1 Similarly, Harvard University researchers determined that advertisements for arrest records were more likely to be displayed in searches for distinctively black names or a historically black fraternity.2 The New York Times also reported on a University of Washington finding that a Google Images search for "CEO" produced 11 percent women, even though 27 percent of US chief executives are women.3
As the Times noted, the...