As Artificial Intelligence Grows in Government, Experts Urge Caution

The technology certainly has benefits, but some say they could be outweighed by its drawbacks.
December 2017
(Shutterstock)
Tod Newcombe
By Tod Newcombe  |  Columnist

Have a question about taxes, vehicle licensing or simply need the phone number for a state agency? If you live in Mississippi, just ask Alexa, Amazon’s digital assistant that has been programmed to also work with state and local governments.

In fact, Mississippi is one of several governments that uses so-called chatbots to handle a range of basic but frequently asked questions. In Utah, teenagers can solicit help reviewing questions that might be on a driver’s test. In Los Angeles, residents can get detailed information about city-sponsored events. And Las Vegas’ digital assistant can answer queries about parks, upcoming elections and the status of building permits.

These early forms of artificial intelligence are popular with governments because the technology is both affordable and doesn’t require a lot of custom development. But A.I. could have big implications beyond just cost savings. “Mississippi is a rural state,” says Craig Orgeron, the state’s chief information officer, “which can mean accessibility challenges for residents.” The one piece of (mostly) universal technology among all residents is a smartphone, which isn’t ideal for conducting transactions with the state. But if the state uses a chatbot to intercede, says Orgeron, then it becomes a bit easier for someone to get help, renew a car’s registration or apply for benefits. “It could be the perfect marriage of services, access and mobility,” he says.

A.I. simulates human intelligence and changes the way it behaves without intervention. It can perform both simple and complex tasks. For example, A.I. can schedule meetings for workers, search for documents or even fill in forms by asking users specific questions. In the case of the more complex situations, A.I. can predict the best time for a caseworker to intervene in a child welfare case.

Despite just a smattering of uses for A.I. in government so far, some experts are already urging caution. Hila Mehr, a fellow at the Harvard Kennedy School, recently wrote, “Despite the clear opportunities, A.I. will not solve systemic problems in government, and could potentially exacerbate issues around service delivery, privacy and ethics if not implemented thoughtfully and strategically.”

Specifically, Mehr calls for A.I. to augment and improve services. It should not, however, be viewed as a technology that can simply replace workers, nor should it ever become a substitute for making critical decisions about citizens. In other words, even if A.I. can predict when a foster kid might be mistreated, caseworkers shouldn’t rely on it alone. Mehr also points out that the technology is not unbiased. After all, she wrote, it was programmed by humans.

One of the biggest effects A.I. will have on government relates to jobs. It will almost certainly replace some workers, such as call center agents whose job is to answer questions and help complete discrete types of transactions -- just the kind of task for an A.I.-enabled device such as Alexa. But the workforce that remains is likely to be smarter. “There’s a trend of the government worker becoming the knowledge worker,” Orgeron says.