Google Goes Social: Orkut and Personalization

I received an IM from someone yesterday about a new service, Orkut, released in affiliation with Google.

At first blush, I wasn't really able to tell much, other than that a programmer had designed the site and the community was definitely closed -- an invitation from an insider is required

Fortunately, I tuned in to the Social Software Weblog this morning to see if they had covered it yet and got an even deeper insight into the matter. As they report:

Orkut has implemented a “Karma” system, where you can rate your connections in three different categories Cool, Sexy and Trustworthy.

Make no mistake about it, this is pretty unique at the moment. Most of the other sites have “testimonial” sections where you can write nice things about the person in a free form text field. Orkut however appears to be the first to provide a rating scale that can then be compiled and the statistics used for *other* purposes. The key to the karma system will be…what “other” purposes it will be used for and will Google ask for permission to use it or assume ownership of the data. Just a few of the questions that we’ll have to wait and find out about.

Source: Soical Software Weblog

The most interesting aspect of this system is definitely the ranking/rating system that is implemented. One thing that strikes me is the lack of any negative values. Clearly the rankings shown seem to indicate that one can go only from good to great. On one level, I can understand that the invitation-only nature of the system would lend itself to be a "good aggregator", however that seems somewhat too forgiving. , There are too many naturally occuring scenarios that could require that the rating be adjusted to reflect a negative value. Perhaps there is a way to prune nodes from the tree, essentially keeping it "good" -- though that seems counter-intuitive to a true social network representation.

This leads me to yet another notion mentioned -- personalization. It seems that Google has a new resource for aggregating data on individuals. It's not clear how well this information could be tied together (legally and technically speaking), but the ramifications are vertainly interesting.

In theory, there are many forms of personalization that can be designed. I can roughly identify them as such:

• Direct Personalization

  This form of personalization comes from asking user's direct and pointed questions relating to their preferences. We see this often as the configurable options for most widgets in software today.

• Indirect Personalization

  This form of personalization occurs behind the scenes. This personalization occurs by observing the user's actions and altering the environment as a result of it. The best example I can think of is Amazon's shopping experience.

• Historical Personalization

  This form of personalization relies on a pre-existing repository of data. Some time ago, when designing the TrueResponse Library application, I designed a Historical Personalization engine. To summarize, the data in the application was all categorized. Combining this data with previous content data and statistics on view rates from the current and previous customers, we were able to determine which content a customer was yet to receive as well as to loosely predict which content would be most successful in wooing the customer.

• Peer Personalization *

  Struck by the possibilities of Google's new system, I am wondering if this is not the next form of personalization. Imagine the combination of social networks, peer attitudes, and user profiles to implement the next generation of personalization engines. Clearly, using the values suggested by your network of friends, potentially combined with the values of your your evaluators, can lead to some new and exciting opportunities.

Very interesting indeed. I'll continue with this in a future post.