Your Next Loan May Be Based More on Your Habits than Your Credit Score
Artificial intelligence programs analyze past data to predict future behavior
If you think that artificial intelligence is unlikely to affect you, think again.
Over the next three years, digital lending is expected to grow to make up almost 10 percent of all loans in the U.S. and Europe, reports NPR. Many of today's 2,000 digital startups use artificial intelligence (AI) to analyze the massive amounts of data that are created every day.
And government regulators are having a hard time keeping up with how quickly the industry is changing.
Marc Stein, the head of Underwrite.ai, creates algorithms that can teach themselves, a feature known as machine learning.
"Shop at Amazon, and they use a form of machine learning," he says. "That's how Amazon's recommender system works."
Every correlation in the data teaches the program something new, some piece of information that it can then use to predict whether or not a loan applicant is a good risk for a lender.
"If we looked at the delta between what people said they made and what we could verify, that was highly predictive," Stein says.
And digital lenders are now pulling in all kinds of information, including purchases, public records, and even SAT scores.
Some lenders require potential borrowers to download an app as part of the application process. This app uploads extremely high amounts of information, such as daily location patterns, text message punctuation, and how many of the entries in their contacts list include last names. This is known as "alternative data" in the credit industry and is used mostly to decide about short-term, high-interest loans.
But this will likely change soon.
"In 10 years, there will hardly be a credit decision made that does not have some flavor of machine learning behind it," says Dave Girouard, CEO of online lender Upstart. "FICO and income, which are sort of the sweet spot of what every consumer lender in the United States uses, actually themselves are quite biased against people."
According to government research, FICO scores hurt young borrowers and borrowers from foreign countries because they often have low incomes, and higher-interest loans are targeted at low-income earners.
Girouard believes that the "smarter" data gathered by AI can make lending more fair for everyone.
"The variables we're introducing to this model actually are reducing the bias inherent in most lending systems," he says. "And that's just one way we're actually expanding access to credit."
If AI can tell good borrowers from bad just by their browsing history, it will no longer matter if your neighborhood is low income or if you are a recent immigrant.
But there are also risks to using AI in the lending industry.
The Perils of AI
Computer programs could develop all kinds of unintended bias if humans play no part in deciding whether or not to extend a loan. For example, some companies have found a correlation between late-night Internet use and bad rates of repaying loans. So should night owls have to pay higher interest rates?
Although there is not yet any evidence that such unfairness is happening, regulators are trying to get the jump on the issue before it does.
Jo Ann Barefoot is a consultant and former regulator.
"And of course one of the tough issues therefore is that these new kinds of models have not yet been tested through an economic downturn," she says.
In the end, only time will tell whether AI will help or hurt borrowers.