Doug and I have discussed social software's dual-capabilities to evoke knowledge by explicit user input (wikis, blogs, social networking platforms such as LiveJournal and LinkedIN) and also to study the knowledge implicit in users' natural activity (collaborative filtering, social networking platforms such as Spoke and HotOrNot). Doug agreed with my assessment during a recent podcast, for example, that users may begin to react to rapidly-multiplying SN platforms by becoming reluctant to define entirely new profiles and recruit entirely new (to the platform) webs of friends. He has suggested that Friendster's recent fall from grace has been the result of UI and bandwidth issues which demanded too much of users' time. As I noted around that time, AOL has at its disposal the raw ingredients of a dominant social network precisely because AIM encourages users to establish and maintain contact lists implicitly, in the course of daily interaction. And we're not the only ones to have brought this issue up. Soon, if not already, social software developers will need to establish deep understanding of pros and cons that characterize explicit and implicit strategies.
The balance thus far has favored explicit user input, but implicit strategies have already yielded powerful results in a number of areas -- Amazon's recommendations and Eurekster's "search parties" both spring to mind. But technological limitations have constrained some platforms, at times forcing explicit implementations which are clearly less powerful than implicit ones. Dodgeball is a clear example of this -- requiring users to manually 'check-in' via SMS is clunky, and the system would see a lot more updates if today's cellular phones supported GPS for third-party apps.
This week, MIT's Media Lab ever-so-helpfully reminds us that we'll get there soon enough. By cobbling custom phones and a custom Bluetooth solution together with cell tower triangulation, a group of researchers was able to sidestep GPS and gather more than "350,000 hours of data over nine months about the location, proximity, activity and communication" of 100 student, faculty and staff volunteers. Not only did their system become adept at guessing whether two people were co-workers or friends, it could leverage the data to predict volunteers' next move with about 85% accuracy. I'm inclined to call this creepy, but they call it 'Reality Mining'.
Privacy issues abound here, although I think those deserve their own post. To me, the key insight into MIT guys' striking achievement seems to be the seamless ease with which they collected their powerful dataset. No posting, "friending" or checking-in. In future, there will obviously remain powerful and optimal applications of explicit user input (indeed, posting to this blog has already taught me a great deal). But the technology that enables the very best implicit social software gets closer every day, and it's a good idea to understand the unique potential of each.
I think all these service are starting to pull together. As the services all start to communicate togther it will become more knowledgable. Like if my blog and slashdot comments link together for idea the web of links will grow. Cross services communication will play a big part in the coming services.
Posted by: John | August 14, 2005 at 03:46 AM