The Executive Order on the development and use of artificial intelligence (AI) issued by President Biden on October 30 is a directive that contains no fewer than 13 sections. But two words in the opening line strike at the challenge presented by AI: “promise” and “peril.”
As the document’s statement of purpose puts it, AI can help to make the world “more prosperous, productive, innovative, and secure” at the same that it increases the risk of “fraud, discrimination, bias, and disinformation,” and other threats.
Among the challenges cited in the Executive Order is the need to ensure that the benefits of AI, such as spurring biomedical research and clinical innovations, are dispersed equitably to traditionally underserved communities. For that reason, a section on “Promoting Innovation” calls for accelerating grants and highlighting existing programs of the Artificial Intelligence/Machine Learning Consortium to Advance Health Equity and Researcher Diversity (AIM-AHEAD) program from the National Institutes of Health (NIH). And the Colorado School of Public Health is deeply involved in the initiative.
ColoradoSPH helps ensure that artificial intelligence serves and empowers all people
AIM-AHEAD is a national consortium of industry, academic and community organizations with a “core mission” to ensure that the power of AI is harnessed in the service of minorities and other groups historically neglected or poorly served by the healthcare system. A key focus – though not the only one – is using AI to probe electronic health records (EHRs), which can be rich sources of clinical and other data.
“The goal of [AIM-AHEAD] is to use this technology to try to eliminate or better understand and address health disparities,” said Evelinn Borrayo, PhD, associate director of research at the Latino Research and Policy Center (LRPC) of ColoradoSPH and Director for Community Outreach and Engagement at the CU Cancer Center. “This consortium is about the inclusion of communities that historically tend to be left behind.” Borrayo and Spero Manson, PhD, director of the Centers for American Indian and Alaska Native Health (CAIANH) at ColoradoSPH, co-direct the North and Midwest Hub of the AIM-AHEAD initiative, a sprawling 15-state area. Both are also members of the AIM-AHEAD Leadership Core.
The hub, which is housed within CAIANH and ColoradoSPH, serves a variety of “stakeholders” who can help to develop AI, including Hispanic/Latino community health organizations, tribal epidemiology centers, urban Indian health centers, and more.
Addressing the shortfalls of AI and machine learning development
Manson acknowledged that the last decade has brought “an explosion of interest as well as investment” in exploring the promise of AI and machine learning (ML) – which uses algorithms to train computers to perform tasks otherwise assigned to humans – and applying that knowledge to improving healthcare.
“There have been substantial areas of achievement in that regard,” Manson said. But he said the work has also revealed “substantial bias” in the algorithms and predictive models as they are applied to “underrepresented and marginalized populations.”
He noted, for example, that the data in EHRs may be incomplete because of barriers to care that people face, including socioeconomic status, race and ethnicity, and geography. In that situation, AI and ML don’t correct for these factors because the technology uses the EHR itself to analyze the data and make predictions, Manson said.
That’s why deepening the reservoir of data in EHRs and other repositories is imperative for the development of AI and ML, he said.
“The idea is to improve healthcare for all citizens, not just those that have benefited narrowly in the past,” he noted.
Improving the diversity of AI workforce
In addition, the workforce of scientists working on AI and ML lacks diversity, while the benefits of research in the field have not yet adequately spread to underserved communities, Manson said.
The North and Midwest Hub has undertaken several “outreach and engagement” projects to meet the goals of AIM-AHEAD, with ColoradoSPH playing a significant role.
For example, two pilot projects aim to build capacity for applying AI and ML to aid communities. In one, Clinic Chat, LLC, a company led by Sheana Bull, PhD, MPH, director of the mHealth Impact Lab at ColoradoSPH, is collaborating with Tepeyac Community Health Center, which provides affordable integrated clinical services in northeast Denver. The initiative, now underway, uses Chatbots to assist American Indian/Alaska Native and Hispanic/Latino people in diagnosing and managing diabetes and cancer.
A second project is working toward incorporating AI and ML coursework into the curriculum for students earning ColoradoSPH’s Certificate in Latino Health.
“It’s an opportunity to introduce students to how using AI and ML can help us understand and benefit the [Latino] population,” Borrayo said. The idea is to build a workforce with the skills to understand the unique healthcare needs of Latinos and apply AI and ML skills to meet them, she added.
“One of the approaches we are also taking is reaching students in the data sciences,” Borrayo said. “We can give those students the background and knowledge about Latino health disparities so they can use those [AI and ML] skills as well.”
Building a generation that uses AI to improve healthcare
Manson also noted that the North and Midwest Hub supports Leadership and Research fellowship programs, which are another component of what he calls “an incremental capacity-building approach” to addressing the goals of AIM-AHEAD.
“We’re seeking to build successive generations, from the undergraduate through the doctoral/graduate to the early investigator pipeline, so these individuals move forward to assume positions of leadership in the promotion of AI and ML,” Manson said.
Borrayo said that she is most interested in continuing to work toward applying solutions for these and other issues in communities around the region. She pointed to the Clinic Chat project as an example of how AI and ML technology can be used to address practical clinical problems.
“I think understanding the data, algorithms and programming is really good for our underrepresented investigators to learn,” she said. “But for our communities, I think the importance lies in the application.
How can we benefit communities that are typically left behind or don’t have access to healthcare in the ways most of us do?”
For Manson, a key question is how members of American Indian/Alaska Native, Latino, and other communities can “shift” from being “simply consumers and recipients” of work in AI and ML and “become true partners” with clinicians and data specialists in finding ideas that improve healthcare.
“The field will be limited in terms of achieving the promise [of AI and ML] until we have that kind of engagement with one another,” Manson said.