[grep2]

Who is interested in building the inverted echo chamber of the internet? Take a standard recommendation algorithm, apply it to news, let's say, and invert the similarity measure for the cluster the article belongs to.

In the best case, the end result would be a page filled with "relevant" (in the sense of read by people with a similar background) articles, but challenging your current position on an individual topic. Think having a "small government" believer receiving an article about the success of nordic social systems.

I, personally, would rather read such news than the "more of the same" you get with a standard similarity measure.