If you’ve ever forgotten about a student loan or missed a few payments on your credit card, you’ll be familiar with creditor calls. But it’s not your local bank or even their head office that’s hounding you. The vast majority of financial institutions today outsource their debt collection to third parties. But at this point, they’re still human. Have you considered a future with AI debt collectors?
“Over the past decade, as banks have effectively outsourced debt collection to third parties, this corner of the financial world has slid into an abyss of harassing phone calls, shoddy recordkeeping, and wrongful collections.”
~Penny Crossman, American Banker
It’s distressing — although perhaps not altogether surprising — to learn that debt collection is often biased against minorities. Investigations by ProPublica and reinvestment groups from California, Maryland, North Carolina, and Illinois have concluded that debt collectors disproportionately target low-income and racial minority communities.
Could data science be put to use here? Would AI debt collectors make this situation better — or worse?
In China, a number of AI debt collectors have appeared — startups utilizing artificial intelligence to scrape borrowers’ information from the internet. In some cases, the companies even access geolocation data to determine where borrowers are. While these methods might be effective, the ramifications of bias in the system could be serious.
Garbage In, Garbage Out
The predictive models used by AI debt collectors are only as good as the data they’re working with. If human bias enters the system — either through biased data or bias built right into the algorithms — the results aren’t going to be any better than human debt collectors. This is the dark side of data.
In her book Weapons of Math Destruction (which I can’t recommend highly enough), Cathy O’Neil warns of the dangers of blindly trusting algorithms and statistics that are used to sort, score, target, and monitor all aspects of our lives. There is great risk in the assumption that statistics and mathematical models are fairer simply because they are free from human bias or discrimination.
And those are just the concerns about the data that determines who to attempt to collect from. In addition to that, we have to consider that using AI debt collectors would mean debtors trying to resolve their credit problems could end up dealing with a virtual assistant — not a real person. An AI debt collector is unlikely to have the emotional response a human has. But is that good or bad? Virtual collectors might be slower to frustration and anger with debtors. But they’re also unable to deviate from the procedure for truly unique cases.
So Will AI Debt Collectors Work?
The answer is a cautious “maybe.”
If the systems are designed carefully and ethically, its possible AI debt collectors could be an improvement. But using open algorithms and understanding how existing data might contain hidden bias is crucial. Without a great deal of care, these systems could make inequity in debt collection much worse.
Want to know more about the importance of open algorithms? Need help checking your data for hidden bias? The Datassist team is here to help. Drop us a line now to discuss your needs.