The Space Between

Mediation and Bias Across the Divide

The turn of the century brought with it a new kind of Web site that offers real hope of bridging the digital divide. The innovations come in the form of dynamic pages and semantic relationships between pages. The content of dynamic pages is generated by software that gathers information as a user views pages on a site. That information about the user’s habits is then used to generate custom pages that are tailored to fit the user’s Web preferences. Not unexpectedly, the most famous example of this comes from the field of electronic commerce. Amazon.com, the world’s largest online retailer, uses dynamic content to make sales recommendations to its users. The site takes in information voluntarily provided by the users, combines it with that user’s history of viewed products and pages on the Amazon.com site, and then predicts which items that the user might wish to buy in the future. Every bit of information gathered the content the viewer will see. No two Amazon.com sessions are the same for every user, even if they view the same series of products.

Outside of commercial uses, dynamic Web pages are growing increasingly popular for other activities, like reading the news. Google, the Web search giant, offers users its news site, Google News. Visitors to the site decide what news will be shown and in what order it will appear on the page. If a user is interested in technology news and not health news, for example, she only needs to click on the “Personalize this page” link. From there the user moves to a “drag and drop” menu that allows them to decide what in what order news items will be displayed, which type of news will appear, how many stories per category, and even whether or not they want she wants to create a “custom” news category. The custom category uses a set of filters that search the Web for news that fit the terms the user decides, an increasingly common practice on Web sites.

There is a distinction to be made between the ways dynamic text is generated for a Web site. Amazon.com’s method illustrates one way, tracking a user’s movements across the Web site to identify patterns and predict likely future behavior. Google News uses a different way: custom filters that display only the content a user wishes to see. Both create dynamic pages but in vastly different ways. The Amazon.com approach is more invasive and is a somewhat touchy subject for privacy advocates; it is also the reason for Amazon.com’s explicit privacy policy. Conversely, the filter method is passive. The user sets her criteria and waits for the appropriate content to find its way through the net. There is less invasion of privacy but more mediation, because while the user can set the filters, she is likely not privy to the inner workings of the software behind the results.

On a Web where millions find information through search engines, those sites’ theoretical and practical impartiality is a hot-button topic. The Washington Post reported in April 2007 that Google posted a sixty percent increase in profits for 2006 and according to the New York Times, the company was valued at $185 billion. Internet search is big business, even spawning subsidiary industries that promise to put their clients’ listing higher in the search engine results, where they will be more visible to consumers.

A study released in 2000 by Steven Lawrence and C. Lee Giles reported that search engines are “indexing a biased sample of the Web,” and that new technologies are “further biasing the accessibility of information on Web.” The search engines rank pages by the number of other Web pages that link to them. Consider for example two hypothetical Web pages that both deal with the French Revolution. One of the pages has 25 links pointing to it, the other 250,000 links. The 250,000-link page will be placed higher in a search engine’s ranking, possibly on the first page of results. The 25-link page will likely show up much further down, and the lower it is ranked, the less likely it is that Web users will find it and read it. If no one finds it and reads it, no one will link to it. The situation snowballs, making it difficult for small-scale Web authors to get their pages read. As Giles and Lawrence write, “This may delay or even prevent the widespread visibility of high-quality information.” Despite the increasing ease of publishing, it is still hard for the ordinary Web user to get his writing read.

On the other hand, statistics point out a different kind of democratization on the Web. Those looking for documents on the Web are having an easier time finding them, according to Deborah Fallows of the Pew Internet & American Life Project. In January 2005, Fallows wrote that “users are extremely positive about search engines and the experience they have searching the Internet.” Readers are finding usable results when they search the Web, and Fallows says filtered search results add to that experience. But Fallows also says it is a reader-beware environment, as many users are unable to tell the difference between sponsored search results—the listings companies have paid for as advertising which often appear above or within other search results—and genuine results. When this sort of counterfeit site or link is used maliciously, it is known as “phishing” and can open the door for identity theft.