Black History Month, according to Barack Obama, is about “taking an unvarnished look at the past so we can build a better future.” In theory, artificial intelligence (AI) in medicine should be able to do exactly that by analyzing data from the past in order to optimize patient outcomes in the future. Experts, on the other hand, have cautioned that AI models can intensify and reproduce systemic racism. How do we make greater use of AI to achieve racial health equity?
According to McCradden and colleagues in The Lancet Digital Health, AI’s true potential lies in exposing existing biases in order to inspire systemic progress and correct health-care gaps. The disparity in pain control between Black and White patients, for example, is structural, motivated by prejudices with little objective foundation; nevertheless, race and ethnicity are also used as proxies for health indicators. Pierson and colleagues created an AI model to assess the severity of osteoarthritis in patients with knee pain, and discovered that AI predictions were more closely related to patient-reported pain than radiologists’ diagnoses, particularly in Black patients. The algorithm’s ability to minimize unexplained differences embedded in the data’s ethnic and socioeconomic diversity was highlighted in the report, which was reported in Nature Medicine. According to the writers, AI could be able to better capture the suffering of underserved patients, potentially redressing inequalities in care access.
While there are more cases of AI exposing disparities in health care, these findings are still uncommon. Furthermore, an algorithm cannot correct past omissions of medical data from marginalized groups or inequities in social and structural determinants of health. The perilous state of diversity in the technology industry could be contributing to these flaws. Chen and colleagues write in Nature Medicine that increasing the diversity of our workforce is critical for treating health inequalities with AI, claiming that “more diverse viewpoints would ensure that the right questions are asked.”
In response to the workforce diversity crisis in the United States, President Joe Biden has named Alondra Nelson as the Deputy Director for Science and Society, reaffirming his commitment to fight discrimination and prejudice in health and society. In science and medicine, Nelson has researched the effects of technology and prejudice. Her research has aided in the understanding of the role of genetics in racial health disparities. Her appointment comes at a crucial moment, as COVID-19 tends to disproportionately increase morbidity and mortality among African-Americans in the United States, amid evidence that African-Americans are more likely to participate in risk-reducing behaviors like mask-wearing.
Despite the fact that Nelson’s appointment to the White House has brought much-needed diversity to medicine and technology, more must be done to encourage and retain Black researchers in senior positions. Timnit Gebru, a prominent Google AI and ethics researcher and co-founder of Black in AI, was recently dismissed, sending shockwaves across the AI community in the United States. Her departure has been related to Gebru’s published research results, which are critical of Google’s AI technology, according to reports. A public letter signed by over 2600 Google employees and 4300 academic, business, and civil society supporters says Gebru’s departure “heralds danger for people working for ethical and just AI — especially Black people and People of Color — across Google.” Google has a duty to set AI research standards as a pioneer in the area of AI and health care. All research must adhere to international ethical standards, and AI researchers must be aware of the diverse needs of those who depend on their algorithms. Similarly, Knight and colleagues explain the critical role of data scientists in combating prejudice and discrimination in health care, emphasizing the importance of generalisability, openness, and reproducibility for ethical and race-sensitive data-driven insights.
AI has the power to aid in racial health equity. Growing diversity in the workforce that develops health-care algorithms, on the other hand, is vital to creating AI that can eliminate health-care inequalities. To ensure that the next generation of AI in medicine represents us all equally, companies and research institutes must invest in women and people of color.