Balkinization  

Sunday, February 01, 2009

Search Engine Secrecy and the Public Sphere

Frank Pasquale

Should we worry about search engine bias? Consider some Republicans' fears that Google, a culturally liberal company, is skewing search results to favor Barack Obama and marginalize the right. Fox News yesterday reported conservative discontent at Google's rapid response to manipulated search results mocking Barack Obama, after its glacial efforts to defuse a "google bomb" aimed at George W. Bush:



In 2003, President Bush's detractors successfully gamed the Google search engine by arranging to have countless Web sites link the words "miserable failure" to Bush's official biography on the White House Web site. The result was that when someone typed the search term "miserable failure" into the Google search box, Bush's bio rose to the top of the search results. And that's how it stayed until 2007, when Google developed an algorithm to detect what became known as "Google bombs" and re-directed the term "miserable failure" to non-political pages.


Unfortunately for Obama, "miserable failure" reverted back to his bio when he moved into the White House. The new president was also Google-bombed with the phrase "cheerful achievement." But this time, Google stepped in quickly, rectifying the situation in a few days, instead of four years. The difference in time did not go unnoticed. "You let this go on for the entire Bush administration," a reader named w3bgrrl wrote on a Google blog. "But since you bought the White House for Obama, you don't want your candidates harmed . . . And your claims not withstanding, even liberals know you're liberal."


There are many good reasons for the difference in treatment; search guru Danny Sullivan discusses some of them in the same article. Google may dismiss such manipulation as a silly prank that really shouldn't be its concern. In NRA patois: Google doesn't produce biased results, google bombers produce biased results.

I still think that political google-bombing merits some attention. As I've noted in blog posts and an article, campaigns are a struggle for salience. As more people form an image of candidates from search results (or related Google properties), we might worry that allegedly neutral, algorithmic representations of authority and popularity are really being influenced by a hidden agenda.

For example, Cory Doctorow's short story Scroogled imagines a Google tightly integrated with DHS and quite willing to use its control of personal information to influence politics. (In Doctorow's story, the company "cleans up" results relating to "members of the Senate Commerce Committee up for reelection.")

I doubt anything like that is happening now, but I'm worried about the fact that no one can verify that it's not. Google's search engine algorithms are a tightly guarded secret, defended in cases like Gonzales v. Google. Theoretically they could be disclosed in a protective order--but what cause of action (or standing?) would a citizen have to sue a search engine over its presentation of data about a given person or entity? One key question for technology policy is whether we will permit Google to assert trade secrecy to the point that we cannot determine whether a scenario like the one envisioned by Doctorow has come to pass.

If you think of Google as analogous to a newspaper, that would not be a problem (and perhaps that result might be mandated by some combination of Miami Herald v. Tornillo and NAACP v. Alabama). However, I think it's better to analogize search engines to phone or cable companies, and to expect commensurate levels of transparency and regulation.

Search engines have some good reasons for keeping their algorithms confidential--if they were public, manipulators of results could quickly swamp Google users with irrelevant results. However, just as Comcast can't avoid net neutrality regulation by saying all its traffic management and spam-fighting methods are trade secrets (or unregulable opinions), Google should not be able to act as an unchallenged "Lord of the Memes" simply by hiding behind these legal doctrines. Moreover, there are ways of developing a qualified transparency that would let a trusted third party examine a search engine's conduct without exposing its business methods for all the world to see.

Compare our dilemmas here to those posed by national security law--another area where we struggle to balance the values of openness and confidentiality. Just as the FISA Court has the right to review even sensitive national security data to assure the rule of law, an analogous institution should be developed to enable regulators at the FTC or FCC to comprehend how dominant search engines' algorithms are developing--and to detect untoward manipulation. Danny Weitzner has highlighted the importance of such a panel in the privacy context, based on his experience in open standards communities:

In the 1990s, the FTC under Christine Varney’s leadership pushed operators of commercial websites to post policies stating how they handle personal information. That was an innovative idea at the time, but the power of personal information processing has swamped the ability of a static statement to capture the privacy impact of sophisticated services, and the level of generality at which these policies tend to be written often obscure the real privacy impact of the practices described. It’s time for regulators to take the next step and assure that both individuals and policy makers have information they need.


A trusted advisory committee within the FCC or FTC should be formed in order to help courts and agencies adjudicate coming controversies over search engine practices. Qualified transparency here is the only chance we have to develop what Christopher Kelty calls a "recursive public"--one that is "vitally concerned with the material and practical maintenance and modification of the technical, legal, practical, and conceptual means of its own existence as a public." Questioning the power of a dominant intermediary like Google is not just a prerogative of "the left." Rather, it's a prerequisite for assuring a level playing field online.


Older Posts
Newer Posts
Home