Is it really "cheaping out" when you computerize a system. CBTC systems cost in the billions, and as such NYC's subway is almost entirely running on 1930's relays. It's a solid system, but it requires the drivers to be on point 24/7. Very little pieces of the system enforce speed limits, as this has to be done mechanically with trip cocks, and while trip cocks will often pull the brakes if a driver runs a red signal, there's usually nothing that prevents the driver from going the wrong way on a one way track [1][2].

But ignoring that, computerizing the system can lead to better headways. Right now most systems, and almost all of the NYC subway, use a fixed block system to enforce braking distances. That means it doesn't matter if a train is 1001ft away or 1999ft away, the following train can't enter the next block until the train completely clears it. It's a safe system, but it means trains can bunch up in a stop and go pattern. Modern computerized systems can behave essentially like our self-driving car dream, where they match their speed safely to the train in from and keep the breaking distance block moving. For people in the Bay Area, BART uses fixed-block signaling (but is actually computerized), and Muni uses moving-block signaling. The difference is especially notable on the BART, which tends to speed up and slam on the brakes when there's train traffic ahead.

Also, it's federally mandated that we computerize trains, anyway, [3] mostly because of incidents like the Metrolink collision.

[1] http://newyork.cbslocal.com/2014/08/25/mta-a-train-operator-...

[2] http://nypost.com/2015/04/06/rookie-subway-operator-reversed...

[3] https://www.fra.dot.gov/ptc