IDEA No 56
COLLABORATIVE FILTERING
Every day there are more webpages to visit, books to buy, music to listen to and films to watch. Search helps us to find stuff we know about – but what about the things we don’t know about?
Collaborative filtering makes predictions about an individual’s interests based on the interests of similar people. The underlying assumption is that if person A has the same opinion as person B on a particular issue, A is more likely to share B’s opinion on a different issue than the opinion of a person chosen at random.
An early adopter of collaborative filtering was Alexa. In 1996, when a visitor arrived at a website, the Alexa toolbar provided a list of websites that had been visited by others who had viewed the same page. In 1999, Amazon acquired Alexa and adopted the technology, introducing its own form of collaborative filtering shortly afterwards.
For each book Amazon sells, it creates a ‘neighbourhood’ of related books based on other people’s purchase history. Whenever you buy a book, Amazon recommends another book from that book’s neighbourhood. This approach is shared by social bookmarking services such as StumbleUpon and Digg. They identify users similar to you, i.e. users who have bookmarked similar pages to you, and suggest pages you might like to visit. Last.fm does the same for music. It builds a detailed profile of each user’s musical taste based on the tracks that user listens to. Netflix does the same for films.
Collaborative filtering is now a fundamental part of our web experience. It helps people to find things they might otherwise miss, and helps online retailers to increase sales through cross-selling. Malcolm Gladwell describes collaborative filtering as ‘a kind of doppelgänger search engine …. If you and your doppelgänger love the same ten books, chances are you’ll also like the eleventh book he likes.’
As the Web matures, our willingness to divulge personal information is growing. In an increasingly connected world, this data is shared across sites and applications. The more personal data we provide and the more it is shared, the more recommendation engines will improve. Collaborative filters will develop tailored experiences based on our demographics and interests as well as our purchase history. Instead of our having to seek out information, collaborative filtering will bring it to us before we have even asked for it. I am sure other people like me would like that too.■
If you like 100 Ideas that Changed the Web, you might like these books too.