Google Recommends Blogs: Another PageRank?

Greg, Andre, and many others, have written good things about the Google Reader recommender system: if you read your blogs and your news using Google Reader, Google will recommender other feeds to you based on your profile.

I have been a bit more critical. The recommender system they offer is basically flat: it seeks to offer a single top-3 list (or top-30 if you ask for more). I have been arguing for years that recommender systems ought to be multidimensional: the user should specify what type of feeds he wants at a given moment. The past and your observed habits are not always good predictors. When you go to a store, the salesman does not immediately suggests things to you, even if he knows you well: he first finds out why you came to the store, he asks questions, he determines your context. For example, I am already subscribing to many Theoretical Computer Science (TCS) blogs. I do not want anymore. Yet, Google Reader assumes that because I have many TCS is my list, surely I want more. Wrong! Yet, I have no way of telling Google Reader that it is wrong, that I want something else. It could start by offering me a tag cloud instead of a list as Amazon started doing this year.

But if Google gets its recommender right, this could have far reaching consequences. According to some unconfirmed numbers, 35% of Amazon’s sales would be due to its recommender system. Right now, 50% of all subscribers to my blog are from Google Reader. This means that, one day, 1/6th of all subscriptions to blogs could be through this recommender system! The recommender system would become part of the blogosphere the same way the Google Ranking algorithm is part of the Web. Will it be a good thing for the blogosphere?