Laura Bliss is CityLab’s West Coast bureau chief. She also writes MapLab, a biweekly newsletter about maps (subscribe here). Her work has appeared in The New York Times, The Atlantic, Sierra, GOOD, Los Angeles, and elsewhere, including in the book The Future of Transportation.
In real time!
With signal outages, garbage fires, door jams, and a constellation of track repairs, New York City’s oft-delayed subways present a daunting challenge for their six million daily riders. To navigate a system in decline, it takes a village of fellow passengers, conductors, and a phalanx of sharp transit apps that can alert commuters to breakdowns and suggest alternative routes.
This month, a new feature on one such tool should prep New York City riders for a little more success. Citymapper, the UK-based transportation software startup, now automatically interprets the MTA’s service statuses—descriptions of incidents, construction projects, and outages affecting subway lines, posted to the agency’s website—and factors them into its route recommendations.
It does this with a special artificial intelligence, a “bot” that’s been trained to “read” the often long, complicated strings of text the MTA uses to communicate disruptions and delays. For example, if you’re traveling downtown on the F train from 63rd Street, and the MTA releases this message—
“Due to FDNY activity at 23St there is no B, D, F, M train service between W4St-Washington Sq and 42St-Bryant Pk in both directions.”
—the bot would extract the relevant bits of information and offer an at-a-glance route change to avoid the outage. (In this case, transferring from the F to the 6 train.)
Google Maps, Transit App, and other tools often include planned service disruptions (such as long-term station closures) in their route suggestions. But Citymapper says it can read and interpret moment-by-moment updates, in real time.
That’s impressive. The MTA isn’t consistent with the language it uses in these statuses; the same stations can have multiple and abbreviated names, and directions can be “southbound,” “Brooklyn-bound,” or “both directions.” Sometimes places are misspelled. Citymapper’s AI has been trained on thousands of these garbled messages to recognize names and phrases. The app studies these inputs as they come in, and spits them back out in the form of a new, easier-to-read route—rather than leave users with the same directions, only with a worrisome exclamation mark and a link to the MTA’s unintelligible communication.
Citymapper interprets and visualizes this status update—
“Due to FDNY activity at Lexington Av-59St, there is no N train service between Times Sq-42St and Queens Plz, northbound and southbound.”
—which confusingly drops the “boro” from the N station “Queensboro Plaza,” as:
For now, Citymapper’s status-reading bot is only available for commuters in New York City. But that’s a good place to start: Thanks to its branching subway lines and aging technology, it may be “the most challenging city in the world when it comes to disruptions and service changes,” according to the company’s blog. Other large rail and bus systems are getting the same AI treatment; stay tuned for updates.
Meanwhile, transit apps are slowly improving around the world, as advocacy groups work to standardize GTFS data. That’s open-source data about static bus and train routes, stops, timetables, and fares. More than 800 transit agencies around the world release this information to Google (with differing approaches to grammar, capitalization, and abbreviation). Third-party developers use that data to build a growing variety of route-generating tools with all kinds of bells and whistles, providing competition to Google Maps, the OG transit planner. That’s good for users, too.
Of course, the best transit-disruption app would be no app at all, but a perfectly functional delay-free system. Alas. If the train can’t be reliable, at least your directions can try.