I'm at the Graphing Social Patterns conference in San Jose again today. I'm covering the event for Read/WriteWeb and doing a few interviews for Read/WriteTalk. This morning, Tim O'Reilly gave the 'developer keynote' for the conference. The presentation hit on three basic themes:
- Background on O'Reilly Media
- A Web 2.0 Refresher
- What I Want from the Social Graph?
Tim started off by talking about the idea that "new technologies first exploited by hackers, then entrepreneurs, then platform players." A few good examples were the move toward adoption of universal WiFi access or webservices from network wifi groups and screen scraping hacks respectively.
Tim referenced the report on Facebook that O'Reilly Radar just released. You can see our coverage here. Basically, O'Reilly has been aggregating usage data from Facebook for a few months. They also normalized the data after Facebook shifted from reporting number of installs to true engagement. The most important finding from the report is that 87% of usage goes to 2% of applications. In other words, the Long Tail is not in effect for Facebook applications yet. Tim went on to segment this finding across different categories of applications, but it was consistent across all categories.
A Web 2.0 Refresher
Tim started by joking about the overuse of the term 'Web 2.0'. However, he felt it was important to offer a refresher on web 2.0 in light of the discussion on the "social graph" (see our explanation of the term here). He said that really web 2.0 is ultimately about "systems that harness the network effect to get better the more people who use them." In other words, it is about building 'collective databases.' He cited a paper by Dan Brickland entitled 'The Cornucopia of the Commons: How to get volunteer labor' (Dan was creator of the first spreadsheet Visicalc), which cites 3 Ways to Build Collective Databases:
- Organized Manual: Pay people (ex: Yahoo! Directory)
- Volunteer Manual: Get volunteers (ex: Wikipedia)
- Organized Mechanical: Architect systems to get smarter automatically (ex: P2P)
Harnessing Collective Intelligence
Tim commented that "every true web 2.0 company is building a database whose value grows in proposition to the number of participants -- that is, a network-effect-driven data lock-in - with accelerating returns to winners." Tim repeated his oft-quoted claim that the Google Page Rank algorithm was the first Web 2.0 application. He also pointed to some recent examples, like Wesabe and Mint doing this in the Personal Financial Services space.
Facebook Can Be Doing Better
Next Tim transitioned to how he felt FaceBook could do better leveraging collective intelligence. He focused on the confirmation questions that Facebook requests, while that data is already available on the web. For example, he talked about friend requests from employees when data on the web already confirms they are employees. Also, O'Reilly Authors have requested to friend him and Amazon already shows they are friends. The fact this data exists is something we've explored reguarly with our coverage of the Semantic Web.
Questions you should be asking?
Tim wrapped up this section of his presentation by reviewing five questions that everyone should be asking:
- Am I doing everything I can to build applications that learn from your users?
- Does my application get better with more users, or just more busy and more crowded?
- If "Data is the Intel Inside" of Web 2.0, what data do I own?
- What user-facing services can I build against it?
- Does my platform give me and my users control, or take it away from us? (note: I personally spoke about this at the Always On Conference)
Conclusion -- What I Want from the Social Graph
Finally, Tim discussed what he believes we really want from social graph data:
- I want it to reflect my REAL social relationships (mine my phone and email)
- I want it to help me manage those contacts (how to reach them, updated status)
- I want to manage groups of people
- I want it to recognize asymmetry in relationships
- I want fine grained control over what I see and what I ignore. For example, he talked about wanting to see his daughter's flickr photos but not other updates.
- I want to discover interesting people and express the actual relationship.
What do you think? Is this what you want from your social graph data?