What I learned when my 2-year-old wanted to wear blackface

Adewole Adamson, MD, of the University of Texas, Austin, aims to create more equity in health care by gathering data from more diverse populations by using artificial intelligence (AI), a type of machine learning. Dr. Adamson’s work is funded by the American Cancer Society (ACS), an organization committed to advancing health equity through research priorities, programs and services for groups who have been marginalized.

Melanoma became a particular focus for Dr. Adamson after meeting Avery Smith, who lost his wife—a Black woman—to the deadly disease.

Avery Smith (left) and Adamson (sidenote)

This personal encounter, coupled with multiple conversations with Black dermatology patients, drove Dr. Adamson to a concerning discovery: as advanced as AI is at detecting possible skin cancers, it is heavily biased.

To understand this bias, it helps to first know how AI works in the early detection of skin cancer, which Dr. Adamson explains in his paper for the New England Journal of Medicine (paywall). The process uses computers that rely on sets of accumulated data to learn what healthy or unhealthy skin looks like and then create an algorithm to predict diagnoses based on those data sets.

This process, known as supervised learning, could lead to huge benefits in preventive care.

After all, early detection is key to better outcomes. The problem is that the data sets don’t include enough information about darker skin tones. As Adamson put it, “everything is viewed through a ‘white lens.’”

“If you don’t teach the algorithm with a diverse set of images, then that algorithm won’t work out in the public that is diverse,” writes Adamson in a study he co-wrote with Smith (according to a story in The Atlantic). “So there’s risk, then, for people with skin of color to fall through the cracks.”

Tragically, Smith’s wife was diagnosed with melanoma too late and paid the ultimate price for it. And she was not an anomaly—though the disease is more common for White patients, Black cancer patients are far more likely to be diagnosed at later stages, causing a notable disparity in survival rates between non-Hispanics whites (90%) and non-Hispanic blacks (66%).

As a computer scientist, Smith suspected this racial bias and reached out to Adamson, hoping a Black dermatologist would have more diverse data sets. Though Adamson didn’t have what Smith was initially looking for, this realization ignited a personal mission to investigate and reduce disparities.

Now, Adamson uses the knowledge gained through his years of research to help advance the fight for health equity. To him, that means not only gaining a wider array of data sets, but also having more conversations with patients to understand how socioeconomic status impacts the level and efficiency of care.

“At the end of the day, what matters most is how we help patients at the patient level,” Adamson told Upworthy. “And how can you do that without knowing exactly what barriers they face?”

"What matters most is how we help patients at the patient level."https://www.kellydavidsonstudio.com/

The American Cancer Society believes everyone deserves a fair and just opportunity to prevent, find, treat, and survive cancer—regardless of how much money they make, the color of their skin, their sexual orientation, gender identity, their disability status, or where they live. Inclusive tools and resources on the Health Equity section of their website can be found here. For more information about skin cancer, visit cancer.org/skincancer.