4

Capacity Building

TRAINING PROGRAM IN AI AND PRECISION NUTRITION

Saurabh Mehta, the Janet and Gordon Lankton Professor, director of the Program in International Nutrition, founding director of the Cornell Center for Precision Nutrition and Health, and codirector of the National Institutes of Health (NIH)-funded Cornell Center for Point-of-Care Diagnostics for Nutrition, Infection, and Cancer, commented that food prevents and can be a major contributor to managing disease, but the diet–disease relationship is complex, with many endogenous and exogenous factors contributing to risk. Moreover, everyone responds differently to food. He noted that the 2020–2030 Strategic Plan for NIH Nutrition Research¹ states that precision nutrition is a unifying and holistic approach to developing comprehensive and dynamic nutritional recommendations relevant to both individual and population health. It is also a framework for incorporating genetics, dietary habits and eating patterns, circadian rhythms, health status, socioeconomic and psychosocial characteristics, food environments, physical activity, and the microbiome in assessing nutrition status and developing interventions.

Artificial intelligence (AI) will play a key role in precision nutrition because of the complexity and amount of data needed for examining the body’s response to internal and external factors (see Figure 4-1) (Lee et al., 2022). These relationships, said Mehta, are further complicated by the time lag and relevant biological period when considering diet as an exposure and chronic disease as an outcome. The feedback loops omnipresent in biological systems makes the challenge of developing precision medicine even more complex.

Mehta pointed out that advanced training in AI for precision medicine is one component of building capacity. To promote that, NIH issued a request

___________________

¹ Available at https://dpcpsi.nih.gov/onr/strategic-plan (accessed January 9, 2024).

for applications (RFA)² aimed at building a future workforce capable of making pivotal discoveries using an increasingly complex landscape of big data and an array of data tools to tackle complex biomedical challenges in nutrition science and diet-related chronic diseases. One RFA element called for assembling teams of interdisciplinary scientists across nutrition, biomedicine, behavioral science, and computational methods. NIH’s portfolio analysis found that out of almost 2,000 NIH training grants, only 20 focused on nutrition and 28 on bioinformatics or data science.

Pulling together collaborators from four institutions at Cornell, Cornell Tech, Weill Cornell Medicine, and the U.S. Military Academy at West Point, Mehta succeeded in applying for this training grant and establishing the Cornell University Training Program in AI and Precision Nutrition.³ Its goal is to train the next generation of scientists and build a workforce equipped with expertise in AI and ML methods to tackle complex biomedical challenges in nutrition and health using high-dimensional data. The focus, said Mehta, is applying precision nutrition to address challenges in maternal and child health.

The training program has one director and six codirectors with 23 faculty members spread across nutrition, computational biology, neurobiology, medicine, population health sciences, computer and information science, and engineering, in order to maximize the faculty and recruitment pool and provide adequate support to trainees. The program plans to add a faculty member with expertise in ethics and fairness in AI. Mehta noted the fortuitous timing because it coincided with the inception of the Cornell Center for Precision Nutrition and Health, which has three hubs: AI & Precision Nutrition, Evidence Synthesis, and Training & External Partnerships; these can serve as a home base for trainees focused on similar areas. The center can also provide supplemental funding for trainees, particularly postdocs, and provide an umbrella for bringing interdisciplinary faculty expertise together. The program will benefit from an initiative at Cornell that is building core AI capabilities and technology for human engagement.

The plan, said Mehta, is to have four predoctoral trainees and one postdoctoral trainee. The predoctoral trainees will be divided between those going for the Ph.D. in nutrition or a related biomedical field who will minor in computer science and those going for a Ph.D. in computer science who will minor in nutrition. He anticipates that the postdoctoral fellow will already have training in computational fields and want to apply that skill set to problems related to nutrition. A key aspect to this T32 program is that it will require all five trainees to have two mentors, one focused on

___________________

³ Available at https://www.cpnh.cornell.edu/t32 (accessed January 9, 2024).

² Available at https://grants.nih.gov/grants/guide/rfa-files/RFA-OD-22-027.html (accessed January 9, 2024).

nutrition-related issues and the other on AI/computational methods, and complete a capstone course that provides them with real data to analyze. Mehta said that if the program is renewed after its initial 5 years, the goal would be to add a second postdoc with expertise in nutrition who wants to learn to apply computational techniques to nutrition research.

Mehta listed several challenges in developing this training program, starting with the need for flexibility and constant evolution. A major challenge is that computer science courses can present a steep learning curve for many nutrition Ph.D. students. The initial group of fellows may need to lean toward those with quantitative or computational backgrounds. Coursework will need to be supplemented with cocurricular activities to develop interdisciplinary scientists. This will include landscape analyses of, for example, New York City AI startups and their partners at Cornell Tech and arranging practicums, practical experiences, and needs assessment exercises with some of these partners. This T32 program may need to be more flexible in appointing trainees earlier, before their qualifying exam, to allow for full development of a truly comentored research and training plan. Mehta said that responsible conduct-of-research training will need to be reimagined and strengthened to account for ethics, fairness, and equity in AI. “This is still a program being built from the ground up,” said Mehta, “and any suggestions and input are definitely welcome.”

BUILDING INCLUSIVE TEAMS FOR FOOD AND NUTRITION RESEARCH

Angela Odoms-Young, the Nancy Schlegel Meinig Associate Professor of Maternal and Child Nutrition and director of the Food and Nutrition Education in Communities Program and New York State’s Expanded Food and Nutrition Education Program at Cornell University, said that her remarks would focus on two key areas regarding diversity, equity, and inclusion (DEI) and belonging (see Figure 4-2):

• What is the significance of promoting diversity, equity, and inclusion (DEI), along with justice and belonging, in applying advanced computation, big data analytics, and high-performance computing in food systems and nutrition research?
• How to create a just, diverse, inclusive, and equitable training environment to support robust and ethical application of advanced computation, big data analytics, and high-performance computing in food and nutrition research?

She noted that AI could both exacerbate disparities and not be effective at closing the gaps for the many U.S. subpopulations experiencing higher

rates of diet-related morbidity and mortality than the general population (Brown et al., 2022). Although much of the work in this area focuses on race, sex, and income, similar disparities in health outcomes exist for members of the LGBTQ+ community⁴ and individuals with disabilities. Despite strong efforts at the national, state, and local levels, disparities in food

___________________

⁴ An initialism that refers to individuals, topics, and communities who are lesbian, gay, bisexual, transgender, queer/questioning. The + symbol acknowledges that there may be sexual/gender identities not represented in the other terms.

insecurity and related outcomes continue to persist, said Odoms-Young (see Figure 4-3) (Coleman-Jensen et al., 2022).

Odoms-Young provided an example of how African American researchers could bring a different lens to nutrition research and incorporate their lived experiences with the burden of obesity in the African American population (see Figure 4-4). She said that a similar conversation has been happening within the AI field, not just in nutrition, though the two conversations are occurring in their own spheres. She quoted Timnit Gebru, cofounder of Black in AI and a member of Microsoft’s Fairness, Accountability, Transparency, and Ethics in AI group: “There is a bias to what kinds of problems we think are important, what kinds of research we think are important, and where we think AI should go. If we do not have diversity in

our set of researchers, we are not going to address problems that are faced by the majority of people in the world. When problems don’t affect us, we don’t think they’re that important, and we might not even know what these problems are, because we’re not interacting with the people who are experiencing them.”⁵

Diversity in AI research teams is important, said Odoms-Young, because extensive evidence indicates diverse teams can generate more inclusive and relevant research questions, which is important when considering the social and structural drivers that link to diet-related conditions and behaviors (Lorenzo et al., 2017; Rock and Grant, 2016; Wegge et al., 2012; Yang et al., 2022). These diverse perspectives can enhance problem identification, decision making, and problem solving. Diverse research teams can alter the behavior of a group’s social majority in ways that lead to improved and more accurate group thinking and are less likely to be influenced by unconscious biases and stereotyping, leading to more objective and unbiased findings. Diverse team members can bring unique knowledge, skill sets, and subject-matter expertise, which can enrich the process and result in a more comprehensive understanding of the subject matter. In addition, their research is more likely to reach and resonate with a broader audience based on its relevance to various communities and stakeholders.

As speakers noted, biased algorithms can create unfair outcomes that unjustifiably and arbitrarily privilege a certain group. This is important, said Odoms-Young, in that different algorithms can act as gatekeepers to health or economic opportunities. Trust, she said, becomes important when considering the diversity and representativeness of AI training datasets and study participation, which is why it is important to include stakeholder engagement when building diverse teams and that study participants realize the promise of that research.

Indigenous knowledge can offer a different way of considering how AI and people relate to one another (Williams and Shipley, 2021). AI, said Odoms-Young, encompasses a wide variety of tools and technologies that model human learning and decision making with wisdom, and wisdom is linked to worldviews. Diverse teams can bring a different worldview that helps when considering how to apply tools from different perspectives.

Odoms-Young cited numerous barriers to DEI and belonging. One barrier is implicit bias. Researchers and team leaders may have unconscious biases that influence their decision making regarding team composition and lead to underrepresentation in research teams, even when qualified candidates are available. A second barrier is limited networking opportuni-

___________________

⁵ Available athttps://www.technologyreview.com/2018/02/14/145462/were-in-a-diversitycrisis-black-in-ais-founder-on-whats-poisoning-the-algorithms-in-our (accessed January 9, 2024).

ties. Access to professional networks and mentorship is crucial for career advancement in research.

Lack of inclusive hiring practices can perpetuate a lack of representation on research teams, with traditional hiring methods potentially favoring candidates from majority groups, and tokenism, said Odoms-Young, can lead to feelings of isolation and marginalization and limit the potential impact of diverse perspectives. Experiences of microaggressions and discrimination within research teams, the fifth barrier, can create hostile work environments affecting members’ well-being and professional development. Other barriers include the following:

• Inequitable allocation of resources: Individuals may face disparities in access to funding, equipment, and other necessary resources. This can hinder their ability to conduct research and publish their findings.
• Lack of representation in leadership roles: A lack of diversity among team leaders and principal investigators (PIs) can limit opportunities for career advancement and an inclusive voice.
• Culturally insensitive environments: Research teams that do not consider the cultural backgrounds and needs of individual members may inadvertently create environments that are less welcoming and inclusive.
• Limited role models: A lack of visible role models from various backgrounds can make it harder for aspiring researchers to envision successful career paths in academia or the scientific community.
• Stereotypes and preconceptions: Stereotypes about the abilities and interests of researchers can influence team dynamics and opportunities for advancement. Preconceived notions about researchers’ areas of expertise may limit their involvement in certain projects.

Structural barriers also start earlier than graduate school, she noted, and breaking down those barriers should start at the high school or elementary school level.

What is striking, said Odoms-Young, is the degree of underrepresentation of minoritized populations among full-time faculty at degree-granting post-secondary institutions; in 2018, they accounted for only about 25 percent of all faculty of all ranks (Hussar et al., 2020). A similar situation is apparent when examining the percentage of doctorates conferred in science, technology, engineering, and mathematics (STEM) fields by race and ethnicity (Hussar et al., 2020). These disparities, she noted, link to funding, where African American biomedical scientists, for example, have a lower rate of NIH funding compared to White scientists (Hoppe et al., 2019). Funding,

she added, is important for team diversity and an individual’s sustainability in academia.

A National Science Foundation analysis found that female scientists leave the field for different reasons depending on race (Metcalf et al., 2018). Native Hawaiian and Pacific Islander women primarily leave for family, but job availability is the main reason for Black and Alaska Native women. She cited another survey of over 25,000 STEM professionals that found that, compared to non-LGBTQ+ scientists, LGBTQ+ scientists were less likely to report opportunities to develop their skills and access to the resources required to do their jobs well (Cech and Waidzunas, 2021), 20 percent more likely to have experienced some professional devaluation, such as being treated as less skilled than their colleagues, and 30 percent more likely to have experienced harassment at work in the past year. Results from the same survey suggest that LGBTQ+ scientists experience some health problems more often than their non-LGBTQ+ peers because of high levels of stress at work, microaggressions, and systemic, ongoing harassment and discrimination.

Racial gaps in net worth continue to persist, said Odoms-Young, and that can contribute to disparities in student loan debt and educational opportunities (Perry et al., 2021). In 2019, 74 percent of Black individuals with student loans had a current balance that exceeded the original loan (Perry et al., 2021), and educational attainment varies geographically, with fewer students in the South achieving a B.A. or higher compared to all other regions of the nation.

Odoms-Young said that when thinking about DEI and belonging, intersectionality is important because all people, even those within certain subgroups, may not be the same depending on their intersectional identities and exposure to intersectional oppression. She highlighted a paper on how NIH is working to foster inclusive excellence for women (see Figure 4-5) (Ten Hagen et al., 2022) and noted strategies for increasing DEI, belonging, and justice from the Nutrition Obesity Research Center (Martin et al., 2023):

• Improving outreach and recruitment by partnering with historically Black colleges and universities, Hispanic-serving institutions, and tribal colleges, and developing position descriptions that communicate an institutional commitment to DEI
• Focusing on policies, procedures, and people by guiding committee members on how to evaluate DEI in candidates, tracking data on recruitment and retention of underrepresented scientists, and evaluating to identify points of attrition contributing factors
• Creating an inclusive environment by conducting institution-wide cultural safety training, cultural and structural competence, and implicit bias training

• Assembling diverse research teams
• Identifying and removing institutional or systemic barriers through pilot funding and equitable workloads
• Supporting community engagement
• Fostering representation in leadership
• Allowing for diverse research topics
• Evaluating progress

Odoms-Young advises her graduate students to try to implement the world they want to see. This may take time, because it requires a culture shift and for someone to emphasize the need to build diverse research teams. This shift will require people in power to make room for others; collecting data, tracking, and commitment to change; and cross-cultural mentoring to address the needs of students from diverse backgrounds that are not represented on the faculty. “There has to be some intentionality,” she said, noting that equity is everyone’s responsibility, not just that of people of color, LGBTQ+ people, or people with disabilities.

Odoms-Young concluded with a quote from Justice Ketanji Brown Jackson on the recent Supreme Court ruling on affirmative action:

MODERATED DISCUSSION

When asked for ideas on how to train future leaders and researchers in the soft skills critical in making teams inclusive, supportive of all members, and safe from microaggressions, Mehta replied that intentionality in both formal training and in how research teams conduct themselves and interact with one another is key. Odoms-Young said that DEI training that involves

___________________

⁶ Available at https://www.supremecourt.gov/opinions/22pdf/20-1199_hgdj.pdf (accessed January 9, 2024).

people bringing their lived experiences is paramount. It can help people realize they have the will but not the tools to be inclusive and create an environment of cultural safety. It is important to bring in a diverse set of trainees to prompt constant thinking about DEI and have empowered and inspiring trainers. Odoms-Young said that data from medical education has shown that a diverse class raises the bar for the entire class.

Mehta noted the need for a minimum set of measurable criteria to assess progress in the field on DEI issues. However, being overly prescriptive about how to address DEI and belonging will not be the most effective route because people come from different backgrounds and have different skill sets. Odoms-Young said that the only way to standardize DEI and belonging efforts is to make funding structures inclusive, for which metrics already exist.

When asked what needs to happen for children in elementary through high school to ensure they are qualified when they apply to college, Mehta said that it will be necessary to overcome the social structures and lack of resources that often disadvantage underrepresented and marginalized populations. Community engagement can help, but again, it will take intentionality to address this problem seriously and dedicate the necessary resources. Becca Jablonski said that a seed grant she received included a requirement to have the training program reviewed by a science team at the university. “I think more funders should really consider doing that, because so often the evaluation is on the outcomes of the research, which is core, but so is this training component,” she said. Odoms-Young found that idea extremely interesting. “What if that was part of all awards, that you have to have some study of how are you contributing overall to the field, to knowledge, to building the field, or building teams versus just looking at research outcomes?” she wondered.

Carmen Tekwe, session moderator and associate professor of biostatistics at Indiana University at Bloomington, said that she has her students develop a statistical method and apply it to different subgroups to see if it works equally well. So far, it does not, but being exposed to the idea that statistical methods do not always apply to all populations introduces the concept of equity in research. She then has her students try to develop group-specific methods. “Many of my students would not normally be exposed to this concept of health disparities in research, but that actually gives them the ability to think about the work and how they can incorporate health disparities into the work they do,” said Tekwe, who asked Mehta if there was a way to apply this approach in his training grant.

Absolutely, said Mehta. “I think that is in the spirit of the whole precision nutrition effort and trying to understand not only the different biological variability but also the variability that is accounted by the kind of methods we use.” It could fit in the capstone course.

Tekwe asked how the field would know it has become more inclusive. Mehta said that it is necessary to make sure every individual feels they can thrive in this area and be comfortable with their skill set. As far as how to assess that, he said that there is no easy answer.

In closing, Rodolphe Barrangou noted the long way to go for the field in terms of DEI but there are reasons to be hopeful and ambitious. “I know there are some challenges and I know there is a long way to go for sure, but disruptive technologies like AI are going to enable us to disrupt the world at a faster pace maybe than we have been able to disrupt it until now,” he said. “But I think with the right leadership, the right mentorship, the right training, the right attitude, and the right teams, we are going to get there.

This page intentionally left blank.