The Politics of Statistics
Chris Anderson has posted an absurd piece called The Probabilistic Age in which he suggests that the reason people aren’t comfortable with Wikipedia and Google is that they are systems that operate according to the laws of probabilistic statistics, which exist on some higher plane that human minds cannot comprehend. Most of the comments on the post focus on Anderson’s incoherent claim that Wikipedia somehow operates “emergently.” (This is a claim that Jimmy Wales himself disputes, by the way.) But what really concerned me was this line:
[Google] makes connections that you or I might not, because they emerge naturally from math on a scale we can’t comprehend.
There is absolutely nothing “natural” about Google’s search results. Google’s (and Yahoo’s and Microsoft’s and everyone else’s) algorithms are designed by human scientists and engineers. These scientists and engineers make specific choices about which algorithms they will use, and which they will not. They decide how the various parts of these algorithms will be weighted. They decide how they will define fuzzy concepts like “spam” and “relevance.” Each of the decisions reflects the values and preferences of the decider, and these values are reflected in the search results we see. It isn’t “alien logic,” it is human logic, and to believe otherwise is to cede control to those who write the algorithms–something I’m frankly surprised Mr. Anderson is willing to do.
When I saw Sergey Brin speak at UC Berkeley this past fall, I was very concerned when he revealed that he himself has fallen victim to, or at least wishes to propagate, the belief that his algorithms are “natural,” saying that the link structure of the web reflected the intrinsic importance of the documents linked to. But documents have no intrinsic importance–they only have importance in the context of a particular query-maker at a particular time. Sergey’s algorithms don’t reveal some truth about what is important–they encode decisions about what should be considered important. Both Mr. Brin and Mr. Anderson need to come to grip with the fact that search engines are inherently political. If people are concerned about Google, or Yahoo, or Wikipedia, then pundits like Chris Anderson should be starting discussions about what we value and how our technologies do or don’t reflect those values, not turning off their brains and blathering on about statistics and the mind of God.