Social networking and Netflix

Netflix is apparently adding some social networking / community features including:

1. Latest reviews stream that continually loads movie reviews in real time as people post them to Netflix
2. “Members’ Top 10 Lists” widget that displays user-generated movie lists based on what Netflix thinks you will like
3. “Unique in…” area that shows the movies that are uniquely popular in your hometown
4. Selection of strangers on Netflix who share your interests or are most “similar to you”
5. List of your friends’ recent activities with Netflix (what movies they have requested, whether they have been returned, etc.)
6. “Friends’ Quiz” that generates simple questions to test you about your Netflix friends’ movie-renting behavior
7. Friends’ Love/Hated area that shows the movies your friends loved or hated (pretty self-explanatory)

This is what I like about social networking. Forget Facebook, MySpace and Friendster. The really important things going on right now are Wikipedia and stuff like this. There is an entire book ("The Wisdom of Crowds") about how markets are much better at making decisions than individuals. One recurring example in the book is about a "guess the number of jelly beans" jar. The average of everyone's guess is usually pretty close to the correct number, sometimes even closer than any single persons guess.

When Netflix put out their challenge to come up with a new algorithm to recommend movies I thought about it. Being a bit of a movie buff I have some problems with their recommendations. If I said I like "Raging Bull" they would be more likely to recommend "Rocky" than "Taxi Driver" or "Mean Streets," both of which are much more similar to "Raging Bull," thematically and visually, than "Rocky" is.

My thoughts about how the recommendations should be determined was that the user needs to specify some criteria - like what they look for most in a movie. I look at the director, my wife looks at the actors. As far as I am concerned I don't care who is in a Scorsese movie because I know Scorsese will do a good job. My wife doesn't care who directs a Nic Cage movie because she likes Nic Cage. My wife is concerned with what the movie is about, I don't really care what it's about because I see movies as works of art to be experienced while she sees them as stories.

If we each said we liked the same movies Netflix would recommend the same movies to both of us, at least I think it would, while we would actually be interested in widely different films.

To account for differences like these we either need a pretty complex algorithm with lots of user-inputted data, or we need some sort of social network. Ideally would be a combination of both. People who gave similar answers to the questions as you did are likely to like similar movies as you do. Or if not you can at least see what connects the choices of the two people and make an inference based on that. This can get extremely complicated which is why I never went ahead and wrote this, although I have most of it planned out in my head.

My wife suggested that a good idea for a site would be a gift recommender. For example all of the kids out here in NY are crazy about Webkinz. Her family back in California has not heard of them. So punch in 6 year old, boy, and a location and you get customized suggestions for what such a person might like. This would obviously require a strong social component and would need to learn from the results it gives.

Anyway this Netflix idea is great. This is where the power of social networking really lies. Not in stuff like MyFacester, but in learning from crowds and applying that knowledge. This is the core of Web 2.0 in my opinion. The Wisdom of Crowds. Check out the book if you haven't already.