The "Computer Says No" Myth

There’s a common fallacy that computers – because they have no souls – turn down loan requests automatically, whereas bankers – some of whom may have souls – are more likely to look at all the facts and say “yes.” This myth popped up again last week, when SME business owners in Birmingham, UK, told the UK’s Parliamentary Commission on Banking Standards that banks need to end the “computer says no culture” that freezes lending. (“Computer says no,” by the way, is a reference to a recurring sketch on the comedy show Little Britain.)

Dumb computers, those that require prospective borrowers to tick a series of boxes in order to get a loan, may say no more often than people do. But smart computers, those relying on predictive analytics, don’t. In fact, they’re more likely to say “yes.”

Here are three good reasons why:

1. Information overload. Predictive models that score loan applications weigh dozens or even hundreds of variables simultaneously. People can’t. In fact, studies show that the human brain can only hold seven to nine pieces of information in short-term memory at a time. If one or two of those pieces of information are negative, the human brain will likely view the application as negative (see point 3 below). Now that Big Data and stricter loan requirements have made more data available to analyze for each loan application, analytics have an even greater advantage over people.
2. Data balancing. When predictive models are built, each piece of information is considered in combination with other pieces of information — this is multivariate analysis. Modelers make sure that a derogatory bit of data isn’t “double-counted” because of its impact on another bit of data. The human brain doesn’t have this capacity.
3. No fear. People over-react to bad experiences — the phenomenon is known as negativity bias. So when lenders have had a run of losses, or have had a few borrowers default, they’re likely to view even healthy borrowers in a negative light. Predictive models can learn from experience, but because they don’t have feelings and emotions they don’t over-weight negative information.

This isn’t just theory. When FICO introduced the Small Business Scoring Service in the early 1990s, the banks that adopted it not only approved more loans, they approved them much faster.

That’s not to say people have no place in the loan approval process. Building a relationship is important for business owners and consumers, and people can consider data that the model isn’t built to analyze. But if borrowers really want a fair shot at credit, they should be glad those soulless computers are crunching away in the background.

Comments

[Michael Allen]

It is not a fallacy that (1) computers have no souls, nor (2) that they turn down loans automatically. Computers obviously can filter large quantity of data faster than humans, but they lack qualitative judgment. They approve that assistant manager of the fast food restaurant making $10k a month just as quick as they decline the Intel programmer input incorrectly as making $10 per year rather than $10K per month or that is $30 recent charge off for a parking ticket. They are not the panacea of wisdom or approvals as presented. Further, in most underwriting situations, they approve the 'easy' lowest risk applications and it's the ones that are more borderline that are sent for human review- why would that be?

I don't believe the discussion is being approached properly- it is not an us versus them debate. Any underwriting system that relies solely either humans or automated analytics is disadvantaged. The future should be improved integration of the unique qualities each brings to the process. FICO is, I believe, misjudging the big picture and overly impressed with their contribution.

[Daniel Melo]

As a matter of fact, we agree more than you realize. As I state in the last paragraph of my post, we believe that both the computer and human judgment have a place in these decisions. And that it’s to our disadvantage to rely solely on one or the other -- or to simply to scapegoat either one in favor of the other. So your message that we need both to succeed is absolutely right!