This was a talk at ETech 2008 done by a MIT professor regarding work in mining and making social inferences about location data. They were able to accurately infer social graphs just by looking at location history

Copying livetweets from @madrox

• Talk about to begin on building community around location-aware tools
• How does location add value to social spaces? Contention is brightkite does not add value on top of twitter, only complexity
• When talking social location, answer the Q: "are you taking virtual communities into the real world, or enabling real into virtual?"
• Data portability with location is key; but location needs to be strongly tied with the same multiple-identity concepts that live in OpenID.
• Back to brightkite: too much data entry involved to build a real community; need more options for inputting data
• OH: "is anyone still using dodgeball?" "it's still up? I thought they were dead."
• Once again, "twitter needs a filter." Talk of twinkle and "geofences" for limiting and controlling location broadcasting.
• disclaimer: only one other person in this discussion follows more than 25 people on twitter
• Community talk concluded: consensus is that we need to start thinking about signal-to-noise and robust fussing/privacy controls

http://chaitanyakuber.wordpress.com/2008/05/17/wherecamp2008_maps-community/