A stylized story that you could tell about banking since, like, “It’s a Wonderful Life” is: In the olden days, each town had a local bank, and the local banker had a thick set of relationships with the people in the town that allowed him to make nuanced judgments of creditworthiness. If he saw you at church and knew your parents and was impressed by your firm handshake, he’d take a chance on giving you a loan; if he saw you at the pool hall and your parents were deadbeats and you had shifty eyes, he wouldn’t give you a loan no matter how good your collateral.
The key determinant of creditworthiness was character, and the banker’s job was to evaluate character. Then banks became vast multinational entities and could no longer rely on those deep relationships to make credit judgments, and they were selling the loans anyway, so lending became a matter of crude judgments like FICO scores and debt-to-income ratios.
This story is often told with an implication that the old methods were more accurate—that deep local knowledge and conservative banking led to better credit decisions than the crude numeric indicators—though I am not sure how true that is; I have never seen an empirical study on the relationship between firm handshakes and creditworthiness.
And the old methods were susceptible to obvious biases; here is a recent study finding that (human-administered) auto loans tend to discriminate against minority borrowers more than (purely statistical) credit card originations. But in any case, the main point is that the new methods are faster, cheaper and much more scalable: If you are everywhere, it is hard to rely on local knowledge, and so you will rely on algorithms instead.
But while not too long ago “algorithms” meant, like, “divide the loan amount by the customer’s income,” now the word “algorithms” carries the promise of infinite subtlety. Here’s a story about auto lenders who are using machine learning to evaluate borrowers, and the language can sound a little “It’s a Wonderful Life”-y: “What artificial intelligence and machine learning allow us to do is to get much broader perspective on consumers, and we’re going to be able to lend more to people who were invisible” thanks to additional data shedding light on their creditworthiness, she said.
The computer doesn’t just look at crude signals, but probes deeply into the real story behind them, sort of: Instead of looking simply at whether a potential borrower has ever filed for bankruptcy, for example, the machine-learning system helped Prestige consider such factors as when the bankruptcy happened, and analyze that data with other variables, including previous car-payment records and time spent living in his or her current residence. (Amusingly, recent bankruptcies are better, for legal-risk rather than character-evaluation reasons.) If you are nostalgic for the old methods of individualized human evaluation, this is … good-ish?
Lenders are making nuanced wholistic evaluations of a person’s life and character rather than just relying on crude scores.
I tend to be an optimist about this sort of stuff, and I can easily believe that a machine-learning algorithm can end up making better lending decisions than both (1) local humans relying on their gut instinct for handshake quality and (2) simplistic numerical scoring mechanisms. And yet: creepy?
The old-timey banker’s evaluation of your character was opaque and unpredictable and prone to bias, but at least there he was, sitting across the desk from you, shaking your hand. The algorithm’s evaluation of your character is opaque and unpredictable (at least to you), and perhaps prone to bias, and you have no human context to fit it into, no way to say “well I always thought that algorithm was a jerk.”
It’s a story about lending but really it’s a story about postmodern life generally. In the olden days, people lived on a small human scale where they had deep connections to each other and all knew each other’s business.
Then modernity occurred and people became more autonomous and anonymous, with weakening ties to their communities but, at the same time, with more freedom. Then, like, Facebook occurred, and now we somehow have the worst of everything; everyone knows everyone else’s business but in an inhuman, alienating, super-scale way.
(Bloomberg)
