Much has been made of social graphs as filters for discovering new content, to the extent that many are now challenging the view that a social filter is even remotely indicative of interest - "…who you know doesn't always translate into what you like"
I'm not convinced that there's such a thing as an "interest graph" - as suggested in the recent Social Graphs Vs. Interest Graphs - but I do believe there are useful intent or interest models, that can be extrapolated from an individual's behaviour.
Flipboard, ShowYou and others aggregate media based on your friends' articulated sharing; this can result in many false positives, with inferences drawn from a semi-autistic social model; our articualted social graphs are driven by liquid affinities and etiquettes.
What's perhaps more insightful is the notion of aggregating media, based on what you own and use… collected artefacts thoroughly riddled with your tastes, memories, aesthetic choices and emotions.
The recently launched Bandito for iOS is an useful illustration of this - Bandito examines the content of your iTunes collection and suggests news items based on your tastes. Curiously, Bandito is a collaboration between a music licensing marketplace and music data provider - suggesting some novel, emerging value chains for the music industry.
Conversley, the wonderfully serendepitous and sublime Shuffler.fm aggregates music from curated news sources into musical genres which also present the news item in its original context along with each track. "Playing" music blogs as continuous mixes implies a kind of social graph, but that's a contrivance - it's simply a collection of editorial and an act of curation. Curation is a little more deliberate than shotgun sharing.
Whether acquired though piracy, digital stores or signals rippling through your social graph, what you collect - not who you know - defines your media genome. Your social graph - as it stands today - is autistic, lacking the subtlety, nuance and fidelty to articulate what you like; indeed, it simply broadcasts what others like.
I have a hypothesis that time and place can be as influential as a collection… but more on that another time ;)