Leaning into the Unknown of Artificial Intelligence

Three UMR faculty are digging deep to learn how emerging Artificial Intelligence (AI) technology can support students’ learning and career training to improve health care outcomes.

Tim Doherty, Rachel Doughty and Xavier Prat-Resina recognize that their vital role as educators is to prepare UMR graduates for a changing world — a world that now includes AI. 

“Rather than shying away from AI, we are leaning into it to help students discover for themselves where it is useful and where it isn’t,” said Dr. Rachel Doughty, lecturer of chemistry in the Center for Learning Innovation (CLI).

“Often the focus is on how AI will weaken education by allowing assignments to be completed by AI,” said Dr. Tim Doherty, senior lecturer of chemistry in the CLI. “There are many reasons for concern there, but there is also an opportunity to use AI to enhance education. I am interested in exploring those opportunities while finding ways to mitigate the downsides of AI in education.”

The Emerging Technology Fellowship, a U of M internal fellowship, gave Doherty the support to implement AI lessons in General Chemistry 2 Lab (CHEM 2336), which he co-teaches with Doughty. They are using AI primarily to help students with scientific writing. 

“Students are iterating prompts, editing output and writing their own to put together classic sections of lab reports. As students do this, they are learning what is required for these writing tasks and where AI can be helpful,” said Dr. Doherty.

“We know many students have already begun experimenting with AI, so we have opted to require students to use it for a variety of writing styles,” said Dr. Doughty. “As they use AI to assist their writing of the different sections of a lab report, students see firsthand how useful AI is for writing more general, wordy sections and how much revision and repetition is required for more specific, data-heavy sections. Just like any other tool, students need to be able to recognize an appropriate time to use AI in their writing.”

For example, AI can generate a relatively mistake-free introduction with very little prompting. It is much more challenging for AI to write a discussion or conclusion section because the software does not know what the student did or what their data looked like. “The student must provide AI with a summary of their main findings, a rubric of items that must be discussed and sometimes even an example discussion section on a different topic before AI produces anything acceptable,” said Dr. Doughty.

The assigned AI submissions are graded according to the same rubric used for student-written submissions, so it is up to the student to ensure AI follows the rubric and provides high-quality work.

Overall, said Dr. Doherty, “The students are doing a great job integrating AI for the useful pieces while not using it in places where AI isn’t helpful.”

For chemistry professor Dr. Xavier Prat-Resina, it was his own use of AI that inspired him to include it in his courses. “When ChatGPT3 launched in fall 2022, it completely changed how I do computer programming, so I decided to revamp my small independent study and allow students to use it. I’ve been paying attention to every new advance and how it can affect my students’ learning.”

He insists that AI can help facilitate the work of the expert, but it is essential that the student first learns to do the tasks without the AI tool.

“I allowed the use of ChatGPT for my students who were just learning computer coding. Given that the course was at the very introductory level, most of my students could not identify good answers or could not write the correct prompt to ChatGPT in order to obtain the desired outcome.”

A potential problem is that students at the novice level try to use AI for tasks they don’t yet have the skills to do by themselves or even to assess the quality of the product. “Then there is a mismatch, and I believe this is creating so much mistrust among educators,” added Dr. Prat-Resina.

The biggest potential benefits of AI in education will occur when educators find the right AI tool for the student’s level of expertise, said Dr. Prat-Resina. Currently, the AI tool that identifies an error in a novice programmer’s computer code (like spellcheck for computer code) can be useful.

“I think we will very soon realize that AI cannot yet substitute for human expertise. It is a great tool that helps speed up and automate tasks.” But Dr. Prat-Resina sees two potential risks. “On one hand, I see educators who deny the usefulness or at least the necessity to live with AI. To these educators I would say that they have to let our students be exposed to it. On the other hand, some people have too quickly jumped on the bandwagon and claim that today’s students do not have to learn to be an expert on the tasks that AI can already do. I don’t think that’s true either.”

Dr. Doughty sees the same benefits and pitfalls. “AI can be used as your private study buddy or tutor. It can generate flashcards, define unfamiliar terms and search the syllabus for the date of your next exam. It is an incredibly useful tool as long as it doesn’t become a crutch that prevents you from learning to do things on your own. No matter how adept AI is, there will always be situations where it is necessary to be able to do it yourself.”

Broadly speaking, said Dr. Doherty, technological improvements over time have changed the required skills from memory recall to the ability to use information, which is where AI comes in. “Some tasks will be automated, knowledge recall requirements will be reduced, and a premium will be placed on interpreting and using information while working with others.”

Read more stories from the Fall 2024 Alumni Magazine: The Kettle.