Taxis by algorithm -- streamlining city transport with graph theory

I recently stumbled across this fascinating preprint from last year by some computer scientists and applied mathematicians. They show how the taxi system in NYC could be made roughly 40% more efficient (fewer miles driven, lots less pollution, etc) with a taxi sharing system that would help people coordinate their trips. Making it work requires data and algorithms; the 40% improvement could be achieved, in principle, while introducing no more than a 5 minute delay to any person's trip.

Looks like great idea, right in line with the spirit of using information to improve coordination. I've written a short article on the research at Medium.