Debugging my code and teaching with ChatGPT

As both a scientist and teacher, artificial intelligence, or AI, has reshaped how I work. Large language models, or LLMs, such as ChatGPT have streamlined time-consuming tasks and, more importantly, made me a more effective researcher and educator. By spending less time debugging code or drafting practice problems by hand, I have more bandwidth to pursue new research questions and design active learning exercises for my students.

As a computational biologist, I use ChatGPT to write and debug bioinformatics pipelines and translate scripts between programming languages and operating systems.

Tasks that once consumed hours, like tracking down an error caused by a stray character, can now be solved in seconds with a prompt: “What is causing the following error in this script?” 

Once, after nearly an hour of fruitless searching, ChatGPT spotted the culprit immediately: an invisible character. Even better, it explained the fix, all within seconds.

Mount St. Mary's University

Assistant professor William Mills works with undergraduate students, Nancy Moluh, Madison Garcia, Oluwaseyi Femi–Falodun, in his research group at Mount St. Mary's University this past spring to analyze publicly available sequencing data that may provide insights about how microRNAs regulate cell biology and disease.

As an instructor, I use ChatGPT to plan course schedules around holidays, generate problem sets and create practice exams. Just as important, I teach students how to use it for self-directed learning.

For example, I show them how to ask: “Can you give me 10 practice problems about transcription and translation?” or “Why is answer choice A incorrect in this question?”

In a bioinformatics course I taught, more than half of the students said ChatGPT had been one of the most helpful tools for their learning.

Still, I remind my students — and myself — that it is a tool, not a replacement, for learning.

Instead of asking ChatGPT to write code, I encourage them to use it to learn to code. Instead of demanding an answer, they can ask it to explain the reasoning behind correct and incorrect choices.

When I was in graduate school, tools like ChatGPT didn’t exist. I taught myself to code through long hours of trial and error. I spent many late nights alone in my thesis lab, staring at endless lines of an incomprehensible code and wondering if I would ever graduate. That struggle, though slow and frustrating, gave me the foundation I still rely on today.

Because I learned coding from the ground up, I can now spot errors in AI-generated code and write precise prompts to improve it.

But the rapid rise of AI, and its near-universal adoption, has also created risks. Studies link LLMs to declines in memory, attention, academic performance and even changes in brain connectivity.

As educators and learners, we must use these tools wisely: not as shortcuts, but as scaffolds that support true understanding. Without that guidance, I fear students will be tempted by shortcuts that rob them of the personal growth that only comes from the hard process of learning.

I call on all educators to guide students into this new era of learning by showing both the remarkable capabilities and the clear limitations of this technology.