Most Internet users have come across a recommender system in one way or another. Imagine, for instance, that a friend recommended that you read a new book and that you subsequently visit your favorite online bookstore. After typing in the title of the book, it appears as just one of the results listed. In one area of the web page possibly called “Customers Who Bought This Item Also Bought,” a list is shown of additional books that are supposedly of interest to you. If you are a regular user of the same online bookstore, such a personalized list of recommendations will appear automatically as soon as you enter the store. The software system that determines which books should be shown to a particular visitor is a recommender system.

The online bookstore scenario is useful for discussing several aspects of such software systems. First, notice that we are talking about personalized recommendations – in other words, every visitor sees a different list depending on his or her tastes. In contrast, many other online shops or news portals may simply inform you of their top-selling items or their most read articles. Theoretically, we could interpret this information as a sort of impersonal buying or reading recommendation as well and, in fact, very popular books will suit the interests and preferences of many users. Still, there will be also many people who do not like to read Harry Potter despite its strong sales in 2007 – in other words, for these people, recommending top-selling items is not very helpful.