According to the MTA, AI technology is being used at multiple stations to measure the percentage of riders avoiding fare. As the agency proclaims it lost nearly $258 million dollars to subway fare evasion in 2022, a May “Blue-Ribbon Panel” report meant to summarize the extent of the problem contained the striking admission.
The report claims that fare-hopping is a literal “insult” to those who pay (according to the testimony of one rider). Therefore, the agency says it needs every tool at its disposal to combat fare hoppers, including artificial intelligence.
The report clarifies the various classifications it uses for different fare hoppers. In addition to the “classic opportunistic evaders” that wait for emergency gates to open and slip through, there are apparently distinct varieties of turnstile fare-hoppers: jumpers, duckers, piggybackers (two hoppers at once), and back-cockers (“pulling back the tri-wheel and squeezing through.”)
The MTA reports claims that more than half of fare-hoppers are opportunists, with jumping landing at second most-popular with 20 percent.
Essentially, artificial intelligence--running 24 hours a day--is now being used to tally up these categories at seven stations. The agency intends to roll AI monitoring out in two dozen more stations over an undisclosed time period.
Insider reported that the specific program being used by the agency was Awaait Artificial Intelligence’s DETECTOR software. As per Awaait’s website, cameras capture “offending” hoppers and send “alerts” to inspectors when infractions occur.
As per the MTA, the need for the technology apparently stems from 10 rotating “human” fare checkers per “quarter” of the subway system not satisfactorily logging the necessary data. So far, the tech has allegedly made the conclusion that fare-hopping “peaks” at the hour of 3 p.m. to 4 p.m., with the morning rush-hour serving as a notable lull.
In what is perhaps a nod to those who may call the technology unproven or risky, the report states that “with the technology providing reliable ‘before’ and ‘after’ evasion counts, it will be increasingly possible to test new approaches in search of what really works.” Whether or not AI fare-monitoring indeed works, it has indisputably arrived.