For nearly a decade, I’ve taught writing to undergraduate juniors. One pattern has remained consistent: students struggle with the leap from structured assignments to the messy, open-ended world of real science communication. So, each semester I open with a quote from writer Pat Conroy:
“Good writing is the hardest form of thinking,” Conroy wrote in “My Reading Life.” “It involves the agony of turning profoundly difficult thoughts into lucid form, then forcing them into the tight-fitting uniform of language.”
His words capture what I believe is the central challenge of science writing: transforming the non-linear, chaotic process of discovery into something that reads as clear, compelling and even inevitable.
Using AI as a creative whiteboard
Courtesy of Austin Shull
Austin Shull poses for a photo with former Presbyterian College undergraduate Charlotte McGuinness while she presents a poster on her research at the 2024 American Association for Cancer Research annual meeting in San Diego.
Despite common fears, I’ve found that generative artificial intelligence, or AI, doesn’t have to short-circuit thinking but can actually enhance it. Used well, it becomes a powerful catalyst for refining ideas and strengthening clarity.
Too often, critics see AI as a shortcut. But for me, it functions more like a whiteboard, an external space where I can sketch my messy thoughts and refine them through rewriting, much like Williams Zinsser wrote in “On Writing Well.” “Rewriting is the essence of writing well.”
When I hit a wall explaining a complex mechanism, I now ask AI: “What’s the clearest way to structure this explanation?” It gives me options, and I weigh them against my knowledge. This back-and-forth helps sharpen my ideas.
In one case, I was drafting a grant proposal on cellular stress responses. I had data linking oxidative damage to a cascade of proteins, but no obvious thread. I worked with AI to explore connections. While it didn’t give me some magic answer on the first prompt, the back-and-forth conversational approach using generative AI helped draw on resources and discover overlapping themes from published work that enabled me to think more clearly and recognize a potential common mechanism.
Courtesy of Austin Shull
Austin Shull poses for a photo with former Presbyterian College undergraduate Layne Benson in the Shull lab.
This wasn’t AI thinking for me. It was a tool to help me see the shape of my own thinking more clearly, like turning a flashlight on the corners of a room I didn’t realize were there. Of course, the human writer stays central. I’m the one asking the right questions, validating the answers and shaping the final voice. AI might be helpful — but I’m still driving.
Teaching scientists to use AI well
The future of science communication should not hinge on whether AI will replace us. Instead, we should focus on how to teach scientists to use these tools wisely, to amplify curiosity, not replace it. We have a responsibility to train scientists to use generative AI ethically, effectively and creatively, acknowledging, as historian Melvin Kranzberg said, “Technology is neither good nor bad; nor is it neutral.”
Therefore, my hope for my students as well as for other scientists is that these non-neutral AI technologies can be used to promote clear thinking rather than being used to avoid it.
