Collecting debts with a ‘box full of robots’

AI is coming to the business of collecting bad debts. Companies think they can get what’s owed at less cost. But for borrowers, the practice could herald a dystopian future.

By Shelly BrisbinMay 25, 2023 2:47 pm, ,

No one looks forward to hearing from a debt collector, but what if those repeated aggressive phone calls didn’t come from a person but from an AI chatbot instead? The implications for borrowers aren’t yet known.

Debt collectors are looking toward AI as a means of getting more from those who owe money, at lower cost. Corin Faife, who wrote about the trend for Vice’s Motherboard tech site, says the use of AI in collections is likely to exacerbate the industry’s historic problems with racialized practices and harassment of borrowers. Listen to the interview above or read the transcript below.

This transcript has been edited lightly for clarity:

Texas Standard: Tell us a little bit about how AI collection works, or might work in the not too distant future. Is it more like a sophisticated robocall? 

Corin Faife: There’s really a mix of things. So definitely, a sophisticated robocall. I think that’s one way you could put it. The robocalls we’re used to are very simplistic. They call up and basically just record a predetermined message. But the thing is, with generative AI, it’ll be using a kind of a chat system. So it can already generate very realistic text that seems very natural and human. And then you can take that text and run it through another system that will do text-to-voice. So then it’s actually speaking to you in a way that sounds pretty human as well. And basically you end up having to have a follow-on conversation. Except rather than a person at the end of the line, it’s basically just a whole box full of robots.

Would the person on the other end know that they were talking to a robot or not?

I don’t know the answer for sure, because I think that it depends whether there is any obligation for the companies to announce that there’s an automated message.

Different states have different laws around that sort of thing.

Right. I think the technology is good enough now that if it didn’t tell you for sure, you could definitely get somewhere into the conversation without realizing it was a robot.

What is in it for debt collection companies? Do they really think AI is going to be more effective than human collectors?

I think what they probably imagine is that it will be roughly as effective, but it will cost much less. So they’re just trying to use AI in the way that so many other companies are now. They’re just trying to be able to cut costs and they’ll be able to do more with less. They’ll be able to do something that would take a team of people a day or a week to do – with an AI system, it could happen much faster. It could be minutes or, in some cases, maybe even seconds.

What’s the impact on those who might receive these calls? Will there be much of a change or how do you see it?

There’s only a few different things going on. So firstly, I think, as you said earlier, no one wants to get one of these calls in the first place. But maybe if you’re going to get one of those calls, you would rather it be from a person, and you try and negotiate with them than know that it is just a completely automated system. That definitely doesn’t sound very fun to me.

I think the other thing is that because of the way these systems are trained, this problem – that the data that we feed into them that comes from the world that we live in and emotionally that world is very biased in many different ways already. So we’re kind of feeding them this data that contains a bias and it ends up just replicating this. It can be racist, can have a whole bunch of other sorts of stereotypes and things like that go into it. And unless we’re able to really push systems carefully, we just don’t know how that is going to manifest. And so I think one of the concerns is always that when these companies build proprietary systems, they don’t really give access to researchers or other people that might try and avoid and detect the bias. And so we just don’t really know exactly how it manifests.

You write that this move toward AI comes at a time when more and more people are in debt. I mean, we’re seeing inflation wreak havoc on people’s household budgets. Can you say a little bit more about it?

I think that that really has been slowly mounting up for many years now. I think the pandemic certainly hasn’t helped. And maybe even if we’re in a recovery period now that has led to some people spending perhaps a bit more generously than they would have otherwise, putting on more credit card debt. And so Americans just acquire more and more debt every year, which I guess is why it’s such a big business actually, to be trying to collect that debt. And why the idea that you could optimize it and do it faster sounds very attractive to these companies.

Is this on the radar for regulators at all, from what you can tell?

To some extent. I think I don’t know about the idea of AI in debt collection specifically, but I got a comment from the Consumer Financial Protection Bureau. They said that they are looking into any general risks AI can pose to consumer finance. I think that probably debt collection is included. But these things always move so quickly that sometimes it just takes regulators a while to catch up.

If you found the reporting above valuable, please consider making a donation to support it here. Your gift helps pay for everything you find on texasstandard.org and KUT.org. Thanks for donating today.