The Department of Housing and Urban Development released guidance about the potential risk of noncompliance with fair housing laws and the use of artificial intelligence.
Covering platforms providing targeted digital advertising as well as rental tenant screening, HUD’s notices come following a presidential order last year that directed the department to address AI concerns, particularly how automated technology might enable housing discrimination. Federal officials, likewise, appear to be increasing their focus on threats of artificial intelligence and the potential harmful impact to American interests.
“Housing providers, tenant screening companies, advertisers, and online platforms should be aware that the Fair Housing Act applies to tenant screening and the advertising of housing, including when artificial intelligence and algorithms are used to perform these functions.” said Demetria McCain, principal deputy assistant secretary fair housing and equal opportunity, in a press release.
In its guidance for housing-related ads, HUD, which oversees the Federal Housing Administration, warned of the dangers within algorithmic functions in AI tools that may lead businesses to unintentionally engage in discriminatory practices.
“Algorithmic delivery functions may operate to exclude protected groups from an ad’s audience or to concentrate delivery to a protected group — an outcome particularly problematic for predatory products,” HUD said.
HUD noted a platform’s reliance on algorithms may eventually lead to advertising that results in price discrimination, even when businesses make an effort to target a diverse set of consumers.
Campaign outcomes may also cause AI systems to employ a discriminatory ad-delivery process based on the number and type of audience interactions it has, as well as any disparities models may have been trained on.
Similarly, the creation of customized and “mirror” sets of consumers designed to match certain characteristics, such as attendees of an open house, could end up running afoul of laws when algorithms come up with the list of recipients, even when the original data did not eliminate any protected classes.
Among precautions HUD advised advertising platform vendors to take are separate processes for running housing-related ads and choosing audience segments, as well as specialized interfaces. The department also recommended platforms avoid providing target-consumer choices for ad delivery that might be discriminatory.
It additionally called for assessments of the data used to train AI systems and safeguards to ensure algorithms are similarly predictive across all class groups, with adjustments made as necessary.
At the same time, real estate related businesses that advertise, such as agents and lenders, should “carefully consider the source, and analyze the composition, of audience datasets used for custom and mirror audience tools for housing-related ads” among different platforms when selecting which to purchase.
Advertisers would benefit from monitoring the outcomes of their campaigns to identify and mitigate risk of noncompliance related to the Fair Housing Act as well.
In other guidance regarding AI use in the rental market, HUD said tenant-screening services and housing providers would be better off with transparent machine-learning models in their selection process.
“If a highly complex model has a discriminatory effect, the model’s lack of transparency could make it hard to prove that a legally sufficient justification exists for the criteria used for a denial decision,” HUD said.
The guidance also arrives after several settlements and lawsuits in the past two years point to the priority that federal departments and agencies are placing on enforcement and elimination of redlining and other forms of housing discrimination.
“Under this administration, HUD is committed to fully enforcing the Fair Housing Act and rooting out all forms of discrimination in housing,” said Acting Secretary Adrianne Todman.