Google and several partners recently launched OpenSocial, an open API for accessing social network information. If you haven't read about this yet, check out the NY Times article for a primer.

Social networks aren't the only domain where sharing information between sites is a much-requested feature. Indeed, we've all invested significant time in building (perhaps redundant) databases of our preferences in recommendation systems such as Netflix, Amazon, and MovieLens. Wouldn't it be nice if those ratings would be portable? Then I could sign up for Netflix, and instantly bring my hundreds of movie ratings with me.

Of course, there are a number of issues with ratings portability, such as:

• What is an entity, and how is it uniquely identified?  (e.g. Is the entire first season of the TV show The Wire a single "tv show entity", or several?)  How is an entity's category/type determined?
• How do I communicate my preferences for an entity to the system? Do I use a single 5 star scale, a set of yes/no questions, open text, or something else?
• Why would any large company wish to share it's ratings database? (Yes, Netflix and MovieLens have published ratings data sets, but those ratings have been anonymized)

While these are significant issues, I wonder if an open ratings API (built on a platform such as OpenSocial) would clear the path to more ubiquitous recommendations. We could allow MovieLens users to publish their ratings to OpenSocial. Myspace developers could then create a variety of social- or algorithm-driven recommendation systems based on this new source of data. If we were running MovieLens to make money, we'd probably try to build these apps ourselves, and sneak in viral features to bring more new users back to our site.

In the short-run, there are other nice benefits of OpenSocial for small-ish recommendation sites such as MovieLens. For example, could we do away with our "buddy" feature, and just allow users to import their social network from Myspace?

Max

(Note: some newer recommendation companies like Flixter and iLike are OpenSocial "launch partners").

I was talking yesterday afternoon with several other lab members about Martin Fowler's "Patterns of Enterprise Application Architecture." In his book, Fowler admits that most patterns aren't anything new. "Creating" a software pattern is just naming and describing a software practice already used by some developers. Of course, Fowler presents patterns with amazing clarity and skill, so his contributions are valuable to the development community.

There's a fascinating short article in IEEE Spectrum this month summarizing an interesting debate about Metcalfe's Law.  Metcalfe's Law says that the value of a communication technology grows with the square of the number of users of that technology.  Informally, the rapid growth is because the value grows with the number of users with whom you can communicate.  

The debate started with a July feature article by Briscoe, Odlyzko, and Tilly that argues that Metcalfe's Law overstates the growth rate.  The heart of the argument is that the growth is actually proportional to the value of the users with whom you can communicate, and that the value of being able to communicate with users probably falls off like a power law function.  (Many features of social networks fall off with a power function.  Some people believe that this is a fundamental law of social networks, that separates them from other types of networks in which value falls off more slowly.)  If the value of new people to communicate with falls off as a power law, then the value of the communications network scales as n * log (n) -- much, much slower than n squared.