Notice the "We Recommend" section at the bottom right-hand corner of article pages. The articles listed there are supposed to be, "Stories you may be interested in based on past browsing." The more articles you view, the better refined that list should become.
I don't have any knowledge of how the list is populated. What I recommended a few months back was to let users rate whether they were interested in the story. If a user rates a story with the "U.S. Postal Service" as a taxonomy keyword, for example, then attribute the rating to the keyword. Wash, repeat and recommend stories that match the user's average highest-rated keywords.
CNN.com realized that if someone clicks on an article, it inherently means they've expressed interest. There's no need for the rate-this-article button.
Everyone should be doing this. But so far it’s just PegasusNews.com and now CNN.com leading the way.