Algorithm-controlled carpooling rather than individual taxi rides could cut the number of vehicles on city roads by 75 percent according to MIT researchers. But it seems to be a classic disconnect between the logical solution and the real world.

The Computer Science and Artificial Intelligence Laboratory crunched the numbers from data about three million taxi rides. The researchers explored whether existing systems such as Uber and Lyft, which match riders with nearby drivers, could be made more efficient.

To do this they tweaked basic ridesharing algorithms to add in features such as working with a fleet of different sized cars and being able to change routes and pick-up points even after a car had begun a particular journey. They also explored the idea of sending cars without any passengers to places with high demand in anticipation of journey requests.

According to the researchers, when they applied the algorithm to the real ride records (which is admittedly a case of retrofitting to the data), they calculated it would be possible for four 3,000 four-passengers vehicles to serve 98 percent of the traffic normally carried by New York’s 14,000 taxis, with the average passenger waiting 2.7 minutes for a pickup. The same system could also cover 95 percent of the traffic by instead using 2,000 ten-seater vehicles.

While the numbers sound great on paper, and the argument that more efficient routing would reduce congestion and pollution, it’s hard to see how the system would overcome real world challenges. It assumes all ridesharing/taxi journeys in a city would be operated by a single company or organization which, even if it were possible, doesn’t sound like a recipe for consumer-friendly pricing.

If the numbers bore out, it would also mean at least 11,000 fewer drivers-for-hire, which would be politically unpalatable to say the least. There’s also an assumption that all paid passengers would be happy to rideshare when in reality not only do many people take taxis because they don’t want to share their journey, but it might be hard to persuade them that a route with pickups of other passengers was most efficient for their personal needs.

There’s also the potential for unintended consequences: if 11,000 taxis really could be taken off the streets of New York, it might ease driving conditions to the point that more people are willing to own and drive a car, winding up with the same number of vehicles on the roads but with fewer of them operating “efficiently”.