Links and power: the political economy of linking on the Web

ABSTRACT

Search engines like Google interpret links to a Web page as objective, peer-endorsed, and machine-readable signs of value. Links have become the currency of the Web. With this economic value they also have power, affecting accessibility and knowledge on the Web.

Links have always been fundamental to the Web. In the last few years their value has become regulated as search engines and other systems that find and define the structures of the Web increasingly index links and anchor text in addition to keywords and page content. In these projects, links are seen as objective, democratic, and machine-readable signs of value. There has been little or no critical discussion about this aspect of links, though link data is heavily used. This article discusses the implications and the power structures inherent in this relatively undocumented but influential change in the structuring of the World Wide Web and is an attempt to scan the field from a critical, humanist perspective.

TRACKING LINKS

A popular though clearly flawed assumption about the Web is that all its nodes are equally accessible. It is true that the Web has no formalized structure or centralized organization other than the rules of the mark-up and scripting languages we use to write and design Web sites. Even those rules are at times disputed: different browsers obey and interpret them in different ways. However, certain Web sites have always been more accessible than others.

In the first years of the Web, most surfers used human-compiled directories listing sites by topic, portals provided by Internet Service Providers (ISPs) and other commercial actors, and search engines that indexed keywords and text phrases in Web sites. After a while, the extensive commercialization and the growing public awareness that highly ranked search results could be bought reduced the credibility of these sites (Introna & Nissenbaum, 2000). Google drastically changed the search engine game by not simply counting keywords but using links as the primary method of determining the value and thereby the deserved visibility of a Web site. Today most search engines have followed Google's strategies and calculate the value of links. Indeed, almost any search engine you use today will use one of only three algorithms to power the search. The algorithm will belong to Google, to the Yahoo! Group (which recently gobbled up AlltheWeb, Altavista, and Inktomi), or to Teoma (Fabos, in press). So much for diversity.

So what was so new about Google's algorithm? Google indexes links between Web sites and interpret a link from A to B as an endorsement of B by A. Links can have different values. If A has a lot of links to it, and C has very few, then a link from A to B is worth more than a link from C to B. The value determined in this way is called a page's PageRank and determines its placement in search results (Brin & Page, 1998; Google, 2004; Marlow, 2001-2002). The PageRank is used in addition to conventional text indexing to generate highly accurate search results. Links can be analyzed more accurately and usefully than traffic or page views and have become both measures of success and dispensers of rank.

Links are increasingly being used in preference to content indexing, not only in search engines but, for instance, to identify communities of Web sites (Flake, Lawrence, Giles, & Coetzee, 2002), or, on a more local scale, to examine social networks and the transferral of memes between webloggers (Marlow, 2001-2002). Google is one of several companies that are developing a map of the Web that identifies connections among individuals, companies, organizations, and Web sites--a map that may prove priceless not only for improving search results but also for personalizing searches and, of course, ads. Sign up for Orkut, the Google-affiliated social networking system, and make all your social relationships machine-readable. Publish a Blogger.com weblog and use your Blogger account to comment on your friends' weblogs--Google owns Blogger and can access this information. Get a free Gmail email account, with half a gigabyte of storage: "don't sort; search," says Google, and now they not only serve you ads based on the content of your emails but they have your personal correspondence in their ever-growing databases.

The extension of search into social networks and personal publication and communication shows that knowledge about the relationships among content is becoming the prime real estate of the Web (Kottke, 2004). Social relationships are simplified in systems like Orkut and Friendster so that machines can process them. Similarly, links between Web sites are assumed to provide an objective measure of value and to be a sign of peer endorsement. This reductive view of links and its implications should be examined more closely.

AN ECONOMY OF LINKS

Links have a direct value on the Web and can be seen as a pseudomonetary unit. A Google search on currency of the Web shows that this is not a novel idea, though it is little theorized. Conventional thinking has assumed that linking from A to B takes value from B and adds value to A. Lawyers have complained that linking to another site's news items, for instance, may be a copyright violation, and companies have sued against those who link to their site (Kelly v Arriba Sort Corp., 1999). Though more sophisticated, Ted Nelson's (1982) concepts of transclusion and transcopyright belong to a similar paradigm where content is value and links are mere mechanics, an outside vehicle for the transmittal of content rather than the item of value itself. In its fully implemented state, transcopyright sees a link from A to B as A using something owned by B, which readers should pay for in the form of a micropayment. This makes perfect sense in a traditional, product-oriented economy where content is king. B manufactured a product that A's readers consumed and should therefore pay for. After Google, it makes no sense at all. The economy of links is not product oriented. It is service oriented, and the service is the link. The link is an action rather than an item, an event rather than a metaphor (Miles, 2001a).