A need for multiple social graphs?

A couple of weeks back Chris Dixon, co-founder of Hunch.com, wrote a widely regarded piece on social graphs; moreover the need for multiple social graphs for differing contexts.

I’d marked the article for comment but just haven’t had time (and still don’t) to really do a response justice so here are a few preliminary thoughts.

Graphs have been around an awfully long time, like hundreds of years but their use in computer science really was a child of the 80s & 90s.  Their current framing is in the social context, a map of all the relationships one has within a social network for example.

However as Chris points out this is limiting.

The thing is ALL my “social” relationships across all contexts, sites, online and offline are my social graph.  So my Twitter followers/following may indeed represent the portion of my social graph that is interests however that context spreads across any number of other properties.

And this is where I think Chris’ thinking butts heads with my own – but only slightly..

Chris talks about “the rising importance of other types of graphs” and gives examples of graphs for Taste, Financial Trust, Endorsement, Local(e).

To me these are all the same graph.  It’s just they represent differing (social) relationship types.  If each were to be represented on separate graphs then the power of graphs in general would be lost – at least without serious jumps forward in semantic technology that is.

What do I mean?  Well, take me for example;

My relationships to my family take on many types (father, husband, son, brother) but they also occur in other contexts such as “financially who do I trust” or “who am I local to.”

To break these into separate graphs would mask the true picture of me.  It would be an administrative nightmare for me to maintain relationships this complex across all and every property in a way meaningful to others.

Far better in my head is an overarching graph that contains all my relationships (this does not mean centralisation of everything as distributed graphs are fine) marked up appropriately with context.

That’s all I have time to write now but this does need more thought.