As generative AI technologies like ChatGPT have become part of many people’s lives, some Missouri State faculty have looked into the ways this technology might interact with their curricular and research interests.

We caught up with faculty who have expertise in dietetics and nutrition to learn about AI’s potential impacts on their field.

Critical Thinking and Patient Care

Dr. Sarah Murray, senior instructor in public health and sports medicine and program director of nutrition and dietetics, and Jaime Gnau, clinical assistant professor in public health and sports medicine, have both asked students to assess AI-generated meal plans.

Meals Planned by AI?

In one assignment, Murray asked ChatGPT to create a meal plan for a 25-year-old female athlete seeking to gain healthy weight. ChatGPT produced pretty appealing menus — featuring plenty of fresh produce and whole grains along with salmon, grilled chicken and lean beef.

But, Murray challenged her students, did the plan meet the patient’s goals? It turns out that it didn’t. When students assessed the plan with a tool that calculates the nutrients in food, they found that ChatGPT’s suggestions came up short in key areas like overall protein and calorie count.

“Then I wanted them to think about the hypothetical patient as an actual person,” Murray says. “What if she were from a different country or culture? Is this plan based on an American diet, and is that a good fit if the patient is from a different country?”

In addition to thinking about a range of backgrounds and cultures, Murray encouraged students to consider differences in ability and lifestyle. How much cooking and food prep did the plan require, and would this make sense for the patient?

She also challenged students to evaluate the menus through a socio-economic lens. Ingredients like fresh produce and salmon might put the ChatGPT plan out of reach for some budgets. How could it be adjusted for a patient who needed to spend less money?

While these analyses revealed significant shortcomings in the ChatGPT meal plan, Murray says this wasn’t the assignment’s only takeaway. She says she was “trying to give students permission to use AI,” while also asking, “Is this a good use of the tool?”

Ultimately, Murray says, “We have lots to learn about AI. Instead of being scared of it, I’m trying to learn how to use it.”

Making Information More Accessible

Dr. Daniela Novotny, senior instructor in public health and sports medicine, has asked students in her class on metabolism to consider how AI might help explain complex concepts to people with varying levels of scientific understanding. For example, she says, AI can be asked to explain a specific concept in language that’s clear for a fifth grader or a high school student.

Like Murray’s and Gnau’s AI meal plan assignments, this exercise challenges students to look at AI-produced material with a critical eye. While AI’s suggested language can help clarify difficult ideas, it’s crucial to examine it for errors and mischaracterizations, Novotny says.

A Potential Game Changer

Natalie Allen, clinical associate professor in public health and sports medicine and team dietitian for Missouri State Athletics, is researching whether AI can accurately calculate the calories and carbohydrates in a food, just by reviewing a photo of it. Allen says that if this method works, it could have a big impact for people with type 1 diabetes — including kids.

“Simplifying the way they count carbs and dose insulin has the potential to be a game changer,” Allen says. “Taking a pic is quick and relatively accessible. If AI can generate an accurate carbohydrate count of the meal to help a patient calculate the necessary insulin dosage, it’s a win-win — for the patient and the dietitian.”

Undergraduate student Mai Elaraby is assisting Allen with this project, something Allen says has been a huge help. The goal isn’t to replace the education, intervention and guidance diabetes patients receive from health care providers. But, Allen says, “As dietitians, our goal is to make things easy for the patient, and this use of AI may help.”

The research process highlights just how much information AI has to assess to generate this kind of data. “Think about the complexity in the context of a picture of chicken wings,” Allen says. “Are they baked or fried? Is the ranch dipping sauce fat free or regular? How many ounces of meat are on each wing? A dietitian would ask these questions, look at the picture and adjust calculations accordingly. We’ll see if AI can do the same.”

Thoughtful, Nuanced Approaches

“Sarah, Jaime, Daniela and Natalie represent the thoughtful, nuanced approaches that are happening across our colleges,” says Provost Dr. John Jasinski. “During the past year, we’ve encouraged discussion about the use of AI in academic affairs. We’ve also shared literature on this topic and guidance about potential applications of AI in teaching and research. It’s exciting to see where these ideas and conversations are heading.”

Around Campus…

Tom Peters, dean of Missouri State Libraries, says his team is currently exploring a variety of AI technologies with the hope that they might improve the process by which search terms are attached to the library’s books and other media.

These search terms, often called “metadata,” are what help library users find items of interest. For any one book, for example, there might be tens — or even hundreds — of applicable subject headings. On top of this, the process of cataloging includes extensive rules and naming conventions. As a result, Peters says that metadata can be very labor intensive to create — and getting it right is crucial for everyone who depends on the library’s resources.

Peters and his team hope that the right AI tool will make the process of creating metadata more efficient. He doesn’t envision AI as the final authority in the process — more like a helpful head start. Ideally, he says, it could be leveraged the way that some people use AI to create an outline or first draft of something they intend to write.

“A lot of times, it’s easier to start from something than to generate the content from nothing,” he says. “That’s the kind of assistance we’re hoping to find with these tools.”