While Artificial Intelligence (AI) has promising potential, it’s important to remember that historically, innovation like this hasn’t been enjoyed equally by all Americans. Take the internet for example, a technology that forever changed the way we work and live. Even though the internet has been around for several decades, minorities and those with lower income and education are still less likely to have access.

AI can inadvertently magnify the biases of its creators. Numerous studies have uncovered algorithms that perpetuate racial biases in facial recognition and even in job applicant screenings.

This article focuses on AI equity challenges for Black America in education, healthcare, and economics, and suggests ways to create more equitable outcomes.

AI Economic Challenges for Black America

Artificial intelligence can promote economic inclusion by increasing accessibility to goods, services, and opportunities. In a perfect world, AI algorithms would break down economic barriers to create more equitable economic prospects. Unfortunately, Black Americans often find that AI exacerbates the situation rather than improving it.

One particularly troubling area is mortgages and bank loans. According to a recent article in Wired magazine, studies found that the AI algorithms used to screen apartment renters and mortgage applicants tend to disproportionately put Black individuals at a disadvantage. The article attributes this to historical patterns of housing segregation that negatively influence how the algorithms are built. So, the historical data used by the AI to determine mortgage eligibility is severely flawed.

The innate bias of these algorithms coupled with a lack of representation in the development of AI models can lead to the creation of AI systems that do not adequately address the economic needs and experiences of Black communities.

AI Healthcare Challenges for Black America

AI advancements in healthcare are aimed to benefit all patients, but biases can have serious implications on Black Americans’ care. The National Institute for Healthcare Management (NIHCM) cites several examples of racial bias in clinical care algorithms:

• The Heart Failure Risk Score used by the American Heart Association assigns three additional points to “nonblack” patients, thereby inaccurately classifying hospitalized Black patients at a lower risk of death. This bias can raise the threshold for utilizing clinical resources for those Black patients.
• Specialist referrals for kidney transplants can be delayed when AI equations estimate that Black patients have higher glomerular filtration rates (eGFR), which assumes they have better kidney function.
• The Vaginal Birth after Cesarian algorithm predicts a lower likelihood of success for a C-section when the patient is African American or Hispanic. Vaginal delivery offers greater health benefits than C-sections, but data like this perpetuates higher C-section rates in nonwhite women than white women.
• The STONE score used in urology predicts the likelihood of kidney stones in patients who come into the ER with flank pain. This AI model adds points to that score when the patient’s race is identified as “nonblack”. As a result, clinicians following these recommendations may be less likely to properly screen Black patients for kidney stones.

Healthcare AI relies on large datasets to make predictions. Faulted historical data and inaccurate algorithms can lead to misdiagnosis or delayed diagnoses for Black patients.

 AI Education Challenges for Black America

Edtech tools leverage artificial intelligence for efficiency and personalization. For Edtech companies to design for Black and Brown learners, they must understand these demographics and champion their needs.

Issues arise when educational AI inadvertently allocates resources and opportunities in a way that disenfranchises Black students. For example, if the AI uses grading algorithms that unfairly penalize or underestimate Black students or relies on inaccurate success predictions for college admissions. Scholarship opportunities and other financial support factors may also be influenced by such faulted AI models.

A recent study found an advising software used by numerous public universities was labeling Black students as “high risk” to not graduate in their chosen major at 4x the rate of White peers.

Rigorous and intentional design, ongoing oversight, and accountability are necessary for ensuring educational AI and machine learning models don’t amplify existing racial biases or introduce new biases through assumptions.

Building Inclusive AI for Black Americans

Bias in AI technology is an ongoing problem. These concerns must be addressed for Black Americans to truly reap AI’s benefit, but the issue is complex and multilayered. Each AI application must pinpoint the solutions specific to its use cases.

For example, healthcare mdoels can pull from successful AI algorithms to improve healthcare access, disease diagnosis, and personalized treatments. AI-driven technologies that help address healthcare disparities can be applied in instances where algorithms are proving to be problematic.

In education, algorithms should be designed with the input of impacted communities and then carefully monitored and documented for disparities. And when it comes to fair housing and financial representation, defective data sets should be acknowledged so new precedents can be set for algorithmic fairness. Addressing biases, promoting fairness, and bridging the digital divide can be tackled through:

1. Requiring diversified input throughout the AI lifecycle so the ideation, design, development, deployment, and subsequent monitoring of an AI algorithm is inclusive and has continual oversight.
2. Supporting an environment of transparency when selecting investments and capital for AI projects. Teams must scrutinize what is being designed and how, and think critically about who will be impacted by the new technology.
3. Equipping underrepresented communities with the tools and skills to understand how AI affects their work and day-to-day lives.
4. Supporting and working collaboratively with organizations that are actively working to remove the barriers faced by Black people, such as Black in AI, to include more Black people in the AI conversation and build more inclusive AI opportunities.

Artificial intelligence is changing our lives at a pace that can feel unsustainable. As more AI applications are embedded into our society, we must use our collective conscious to prevent historically underserved populations like Black America from getting left behind. Diversified input, thoughtful investments, and grassroots support are great starting points for creating AI equity, but there is still much work to be done.