Housing advocates say automated tools from companies like CoreLogic, used by landlords to screen tenants, fail to capture the complexity of criminal records
But advocates say algorithms can't capture the complexity of criminal records — Mikhail Arroyo had made it out of the coma … Tweets: @mims Tweets: Christopher Mims / @mims : AI redlining, cool cool http://twitter.com/...
Context & Ripple Effects
This 2019 report is an early entry in a fight that has only grown: housing advocates argue that tenant-screening tools sold by companies like CoreLogic compress messy criminal records into pass/fail risk scores, and the Mikhail Arroyo case — an applicant rejected despite circumstances no flat dataset captures — is the human face of that critique. The concern predates the current AI boom but maps directly onto it.
Subsequent reporting has validated the advocates' core claim rather than blunting it: a later investigation found many algorithmic screening services widely used by landlords produce flawed reports, and by 2023 advocates were warning that landlords' growing adoption of AI screening heightens both error rates and discrimination risk. The same pattern of opaque risk scoring was simultaneously being documented in the courts, where a game built on real defendant data demonstrated the shortcomings of COMPAS.
First-order effects
- Applicants like Mikhail Arroyo bear the immediate cost: a criminal record flattened into an automated score can mean a denied lease with little recourse to explain context, while landlords relying on CoreLogic-style tools inherit those errors as their own decisions.
Second-order effects
- Screening vendors face pressure to justify their accuracy and dispute processes — the documented flaws in these services give advocates and regulators a concrete target, and landlords who keep using them after such findings take on reputational and legal exposure.
Third-order effects
- If the pattern holds across screening, sentencing, and patrol-allocation tools, opaque risk scoring becomes a regulatory problem spanning housing and criminal justice alike, pushing toward demands for explainability, contestability, and audit rights over records-driven algorithms.
The trend: Automated risk scoring is spreading from criminal justice into housing decisions faster than accountability mechanisms for its errors, making tenant screening one front in a broader fight over algorithmic gatekeeping.