Landlords Still Make Exceptions in the Algorithm Age

When your credit score drops or an eviction appears on your record, an algorithm might reject your rental application before a human ever sees your name. But researchers just discovered something surprising: landlords are still making exceptions, just in a completely new way.

A team from Columbia, Duke, and CREST-ENSAE Paris studied how landlords screen tenants in Durham and San Jose. They interviewed 88 people involved in leasing decisions, from small property owners to corporate executives. What they found challenges everything we thought about automated decision-making.

"One of the paper's big findings is that landlords and property managers are making exceptions, but they're what we call systematic exceptions," says Barbara Kiviat, assistant professor of sociology at Columbia. "It's not for one person at a time, but instead for one category of person at a time."

The difference matters enormously. Old-school landlords who reviewed applications personally could bend rules for unique circumstances. Someone with a foreclosure from caring for a sick parent might get approved based on their specific story.

Algorithmic systems work differently. They require exceptions to be programmed in advance. After the 2008 financial crisis, many landlords stopped penalizing foreclosures once they understood predatory lending caused them. But that mercy had to be coded into the system as a blanket rule.

Both approaches involve human judgment. The key question is whose stories get anticipated and whose get overlooked.

The Ripple Effect

This research matters far beyond rental housing. Algorithms now control access to jobs, insurance, loans, and countless other opportunities. Each system decides in advance which circumstances deserve exceptions and which don't.

Cornell sociologist Karen Levy calls the study "a really valuable intervention" into understanding digital-age rules. Algorithmic systems don't eliminate human compassion. They just require it to be standardized and specified upfront.

The challenge is that programmers and executives can't anticipate every difficult situation. Common hardships like the 2008 crisis might get built into exception rules. Rare or complex circumstances often don't make the cut.

Empty units cost landlords money, so perfect tenant records have never been realistic. Whether using algorithms or personal judgment, property owners need to fill apartments. The question is how they decide which flawed applications deserve a second look.

Research like this helps us understand the new infrastructure of American opportunity. Automated gatekeeping now determines who gets access to housing, employment, and financial services. Understanding how these systems work is the first step toward making them work better for everyone.