There is a big push to open up a lot of data about us - mainly, who and what we like. This push is extremely exciting, and I wholly hope it is successful. However, I think there are some aspects of relationships and interest that aren't being discussed -- namely, that both are highly contextual. First, though, a little background.
Recently, I feel down an interesting rabbit hole: I started to learn about lifestreaming. I've set up accounts on lifestrea.ms and onaswarm. This then lead to a lot of reading about Attention Profile Markup Language (APML), which promises to capture and encode (using an open standard) your interests. However, interests are social, and this has led to even more reading about the social graph, or the computer encoding of your connections to other people.
The three concepts dovetail pretty nicely. Your life stream reflects all the things that you're interested in -- it is a stream of your life. So a service like lifestrea.ms aggregates all your feeds (from an OPML file), your bookmarks (from del.icio.us or ma.gnolia), your twitter stream, your youtube channel, your blog posts, your tumblr posts, your last.fm playlists, your photos, your email, your search history (if you let it) -- everything. This is useful in two ways. First, by collecting your fragmented web 2.0 identity, it becomes possible to share your output and content that's relevant to you from a single location. There are all kinds of ways people do that now (just look at my sidebar there, or people's RSS feeds that are rebroadcast on twitter, applications on facebook). Lifestreaming, however, pulls everything together and then reexposes it to the world. Additionally, since this is the data that we share on social networks already, the lifestreaming services themselves are forms of social network -- I can have a friend on lifestrea.ms, or join a group on onaswarm, and then their data intermingles with my data. If the social graphs were open, when you got to my lifestream, you could create an account and suddenly have friend-level access right off the bat, and I could automatically subscribe to your stream because it would know we know one another. From all this data about me and my friends, your lifestreaming service builds your attention profile; because the system knows a lot about what you read, listen to, watch, talk about, and bookmark, it can (theoretically) infer what you might be interested in generally. This is like what Amazon does when you search for books -- by paying attention to what you look for and at and what you buy, Amazon is able to get an idea of your interests and suggest relevant items to you. Well, with your Attention Profile, lots of sites can do this sort of analysis and present you with more relevant content for you (the details, privacy issues, and basic utility of this idea has been discussed elsewhere). This profile is data that you control -- meaning you determine who sees it and what they see -- and could help reduce the serious information overload problem that we're heading toward (or have hit).
Privacy issues aside, this all sounds pretty cool -- I go to a website, login, and instantly it knows what I'm interested in and who I know. But there's a problem: both interest and relationships are rooted in context. With attention, it's pretty clear -- my interests change pretty drastically depending on my context. During a given work day, my attention profile should reflect that I want to know about nonprofit technology, content management, crime prevention. At home, I still want to know about content management, but I also want to know about photography and local DC bands. To be really useful, my profile needs to reflect this, and the other nuances of my personality and interests -- my context and current projects will have a strong effect on my attention profile. This appears to be reflected in the spec, which is great, until we start thinking about making all this stuff social. Thanks to Justin Thorp, I've been reading up on Object Centered Sociality. The basic idea of Object Centered Sociality is that we don't just have relationships in a vacuum, but instead that our relationships are based around some "social object," such as photos, music, or links. To me, this makes sense -- I tend to have shared interests with my friends, both online and off, and social objects are, in a sense, reified interests. And, looking back at the attention idea, all relationships are not the same -- my work friends are not the same as my college friends are not the same as my friends from #plone. The two concepts -- contextual attention and object centered sociality -- are to me intertwined; nothing I do is in a vacuum, but instead is always inspired by some object. We socialize about different objects, and these objects shape our attention profiles differently.
The key point here is simple: The context of the relationship will need to be captured in the social graph, much as the context of my activity must be captured in my Attention Profile. However, this is not a simple thing. Sir Tim recently blogged that we can open up the social graph with RDF and FOAF. But, FOAF doesn't say enough yet. For both Attention Profile or the Open Social Graph to work, we're going to need to find way to talk about context (why I'm interested in this or why I know that person) that every consuming site will understand, which is a very hard problem - we're going to have to express context in language, and agreeing on a sufficiently rich controlled vocabulary will not be easy (nor is it necessarily the best way to go about this). I have no doubt that it'll be solved, but it's not a bad idea to talk about it now.