AI could make government more efficient, but at what cost?

From American Public Media, this is Marketplace Tech.

I'm Megan McCarty Carrino.

Government services are notorious for being under-resourced and overwhelmed, so it's no surprise

many are looking to AI for an efficiency boost. Chatbots, for instance, could provide

personalized assistance for filling out applications or referring constituents to the services they

qualify for. But implementing AI in domains where the public interacts with government,

like accessing food stamps, unemployment, or disability benefits, could come with tradeoffs,

according to a recent report by the Roosevelt Institute. Samantha Shori authored that

report on AI and public administration. She also teaches at the University of Pittsburgh.

Because AI is now involved in doing the repetitive tasks, we're seeing that there are increased

demands and expectations on productivity as well.

And so the work is getting more difficult and is expected to be done faster.

And these are qualities that really contribute to stress that people experience in the workplace.

And a couple other of the impacts that we see.

And one that I think is really important is around increased risk and responsibility for individual workers.

And so when there's an error, it comes back on individual workers rather than the people who are designing those systems or the organizations that are choosing to deploy them.

Does it seem like workers in public administration are being consulted about how AI is being

deployed?

There's a growing appetite and a growing movement towards incorporating workers as decision

makers alongside the institution of AI systems, but the implementation

and the procurement process can often be very kind of top down. And decisions about how people's

...