In New York, Virtually Every Taxi Trip Can Be Shared

Uber, Lyft, and other ridesharing apps are often described—or describe themselves—in terms of technological inevitability. Taxi drivers and regulators can keep crying about these forces of "sharing," but what's the use? This is just what a connected society is. It's natural.

But the true sharing concept is becoming still even deeper than car service networks and "community drivers." The next wave is peer-to-peer, less in the sense of passengers and drivers and more in the sense of passengers and other passengers.

Unlike the  similarly-minded Uber Pool service, this sort of sharing doesn't discriminate between taxis, car services, random dudes, and otherwise; it just cares that there is more than one seat in whatever variety of vehicle happens to be chosen. Simply: if two people are looking for a ride from the same area, an app might facilitate those two (or more) people going together.

This adds a new dimension of efficiency, potentially cutting down on overall trip miles, which means cutting down on emissions, road wear, and the rest of it. In research published this week in the  Proceedings of the National Academy of Sciences, a team of researchers based at MIT examined 150 million trip records collected from more than 13,000 New York City cabs with the goal of seeing just how many trips/miles could saved under these next-generation sharing systems using a newly-crafted, but not terribly Earth-shattering algorithm. The answer, in brief: 30 percent of all road miles.

More specifically, if passengers were willing to accept up to five minutes in additional travel delays per trip, 95 percent of all trips could have been shared. In the optimum sharing situation, the researchers found savings of up to 40 percent in total travel times. The optimal ride-share, however, requires quite a bit of advance planning and synchronization, so the team set the bar quite a bit lower: only trips beginning within a minute of one another could be shared. This is where you get 30 percent.

"We think that with the potential of a 30 percent reduction in operational costs, there is plenty of room for redistributing these benefits to customers," Paolo Santi, the paper's lead author, said in a statement. "Because we have to offer them lower fares; to drivers, because we have to incentivize them to belong to this system; to companies; and of course, there is a benefit for the community."

The results are perhaps even more impressive than that 30 percent might suggest. The authors of the current study looked at four measurements of any given trip: the time and GPS location of the pickup and drop-off, the set of other trips that overlap with this trip, and, finally, the resulting delay from combining the trips. Sticking within the delay threshold of five minutes, the researchers found that on average any given trip is sharable with 100 other trips. Granted, this is New York City, but still: wow.

The researchers further tested their algorithm using an unnamed taxi-sharing app to see if it would work in real-time. It succeeded quite easily, in fact, with only the processing power of a single Linux box. Crunching data for some 100,000 trips could be achieved in a mere tenth of a second. And with an average of 300 new taxi trips initiated in New York every minute, the algorithm is faster than it would ever need to be.

There are apps that are already tackling the real-time taxi sharing problem. Uber, Lyft and Sidecar recently announced beta tests of real-time pooled rides in San Francisco. Bandwagon, a Brooklyn-based taxi sharing startup, offers an app for matching riders in high-density areas, and has begun a trial integration with Hailo in New York. (My colleague, Motherboard editor-at-large Alex Pasternack, is one of its co-founders.) Hitch is a shared-ride app for the Bay Area. Carma and BlaBlaCar, both from Europe, have developed apps for long-distance carpooling.

For now, the researchers have their own more modest app, a handy taxicab data visualizer called Hubcab, so if anyone is doubtful about these results or thinks they can make an even better algorithm, the raw materials are right here. It's a reminder that ride-sharing is something more fundamental than brands and marketing. Real sharing depends upon a new sort of openness and interoperability beyond what the market currently tends to push. To know the system, you need the data, and the data—be it from a city taxi commission or a Silicon Valley taxi app—needs to be made more available than it is now. Sharing is, after all, the conceptual opposite of proprietary, isn't it?