IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Analyzing the Unintended Impacts of Policy With Data and AI

Through a project launched during a recent SAS Hackathon, Milwaukee County, Wis., is looking to AI to examine and improve ordinances and policies related to foreclosure that may inadvertently promote racial inequities.

Photo of filing cabinet files with one file that reads "Policies" in the center.
Milwaukee County, Wis., is looking to the power of AI to help identify and reduce inadvertent bias and inequities in its policymaking.

Local government entities are becoming increasingly aware of their role in addressing systemic racism. As experts explore how civic tech can reinforce inequities, some local agencies like Pittsburgh are assessing the tech tools and algorithms they rely on.

An April 2020 ordinance officially launched the countywide effort to improve public health through a racial equity lens. This effort followed a 2019 declaration that labeled racism a public health crisis.

According to CIO Lynn Fyhrlund, the county found that in many ways the causes are systemic, pointing to redlining as one issue that has left a lasting impact on society today.

And while this complex issue is without a one-size-fits-all solution, Fyhrlund said that one way the county hopes to address the causes is by analyzing policies and ordinances with AI and machine learning to better understand unintended regional impacts.

The effort began when the county participated in the 2022 SAS Hackathon, a six-week program where organizations come with a business problem and the related data, and using SAS software tools, the team worked to create a minimum viable product to address the problem.

The county’s initial idea was to use the AI and machine learning technology to search for instances of bias in the language of policies and procedures related to foreclosure. But as Fyhrlund learned, the impact of those policies and any bias within them is revealed through the related data. For example, the impact of policy wording and any subsequent inequities may be quantified by the number of foreclosures occurring, the neighborhoods in which those happened and the demographic data of those affected.

The technology used was SAS Viya Suite for analytics and AI, hosted on the Microsoft Azure platform, as explained by SAS head of smart cities partnerships James Caton.

“They found open source data on foreclosure rates in the county,” Caton said. “They use other information — demographic information, ZIP code information, etc. — to then focus in on foreclosure rates by socio-demographic [data] and geography in the county.”

The 2022 hackathon's MKE Equity team, led by Fyhrlund and including members from the county, SAS and SolasAI, worked together with data scientists and a SAS mentor through weekly meetings to analyze various data sets across different county systems, with an emphasis on foreclosure data.

Even so, the results were inconclusive, but Fyhrlund said the idea has traction, and the work is not yet done. To move forward, the county would need to gather data from even more sources.

Another thing the county team realized in their work was that the local government is not the sole entity impacting these outcomes. For example, with foreclosures, the county plays a small role — impacting foreclosures primarily at a tax level — while other players like banks may have a greater impact.

The project generated questions that Fyhrlund said the county should already be asking, noting that it will be a “constant learning journey” to find an effective, long-term solution.

“With cities and commercial organizations, the reality is they don’t intentionally create programs that have bias. Bias is there unintentionally,” Caton explained, noting that technology can be used to analyze any program or policy that impacts the community to ensure that it is having the intended impact.

Fyhrlund’s advice for other entities looking to embark on their own policy analysis project is to focus on the data — know how data is collected, where it is stored and how it is going to be used.

And when it comes to data, privacy concerns are likely to come up, so organizations must focus on sensitivity and ensuring that any data use follows privacy guidelines. He also underlined the importance of making sure the use of AI tools is ethical.

Editor's note: An earlier version of this story misidentified a hackathon participants. The error has been corrected.
Julia Edinger is a staff writer for Government Technology. She has a bachelor's degree in English from the University of Toledo and has since worked in publishing and media. She's currently located in Southern California.