The text of the article explains that it’s based on reducing the number of taxis (or cars for hire generally) on the road, reducing parking spots, and increasing carpooling:

In the 2010s, the Senseable City Lab at the Massachusetts Institute of Technology, where one of us serves as the director, was at the forefront of using Big Data to study how ride-hailing and ride-sharing could make our streets cleaner and more efficient. The findings appeared to be astonishing: With minimal delays to passengers, we could match riders and reduce the size of New York City taxi fleets by 40%. More people could get around in fewer cars for less money. We could reduce car ownership, and free up curbs and parking lots for new uses.

This utopian vision was not only compelling but within reach. After publishing our results, we started the first collaboration between MIT and Uber to research a then-new product: Uber Pool (now rebranded UberX Share), a service that allows riders to share cars when heading to similar destinations for a lower cost.

It goes on to explain that it’s a problem of induced demand (same phenomenon that causes highway expansion not to actually help with congestion in the long term):

Alas, there is no such thing as a free lunch.

Our research was technically right, but we had not taken into account changes in human behavior. Cars are more convenient and comfortable than walking, buses and subways — and that is why they are so popular. Make them even cheaper through ride-sharing and people are coaxed away from those other forms of transit.

This dynamic became clear in the data a few years later: On average, ride-hailing trips generated far more traffic and 69% more carbon dioxide than the trips they displaced.

We were proud of our contribution to ride-sharing but dismayed to see the results of a 2018 study that found that Uber Pool was so cheap it increased overall city travel: For every mile of personal driving it removed, it added 2.6 miles of people who otherwise would have taken another mode of transportation.