John Steiner, MD, MPH, has come a long way without artificial intelligence (AI). He’s led research projects, guided trainees into professional careers, and published hundreds of original research papers, editorials, and other writings over the years.
“Pretty much everything that my generation of researchers learned about scientific writing predated generative AI,” says the longtime University of Colorado Anschutz School of Medicine faculty member, who was a full-time professor of internal medicine until 2008.
Now, in retirement, AI is taking up more of Steiner’s brain space as he’s returned to CU Anschutz to teach courses to early-career and mid-career researchers on scientific writing through the Adult and Child Center for Outcomes Research and Delivery Science (ACCORDS), which he helped start more than two decades ago.
Almost all his students see some value in using generative AI to some extent in their writing work.
“They’ve become my teachers,” he says. “Most of these students agree that AI can be helpful in searching the literature or summarizing what’s already been written about a topic, although you do have to be careful about hallucinations and made-up references and those sorts of land mines.”
In a recent editorial he wrote for JAMA Internal Medicine, Steiner recounts his AI-focused discussions with students and how the evolution of technology is influencing scientific writing, underscoring that most researchers don’t have a background in the humanities or social sciences that can bolster writing skills.
“In their clinical or doctoral training, they have been immersed in laboratory or clinical practice,” he writes. “While doctoral students may progressively develop writing skills as they plan and prepare their dissertation, clinician-scientists may only receive formal training in grant writing, the sales pitch necessary to launch their research. Only later do they discover that writing is the gateway to academic advancement.”
AI tools appear to offer an alternative to learning those skills, but Steiner says even with the evolution of AI, there’s value in diving headfirst into the writing process.
‘Research is a creative act’
At its core, Steiner says that research is a “creative act.”
“You start to think of science as this hyper-rational discipline, but it requires enormous creativity,” he says. “It's just a different kind of creativity than painting a painting or writing a short story.”
Creativity remains important in scientific writing, Steiner adds. Word choice, for example, can reveal the biases of the researcher and influence the reader’s perception of the work.
“We proofread our papers to make sure that we do not confuse association with causation or overgeneralize from limited samples. Skillful word choice also reinforces the exactitude that we expect in science,” he writes in JAMA Internal Medicine. “We often appreciate good writing the most when we read bad writing. Sloppily written grant proposals, manuscripts, and published articles reduce our confidence in their authors’ thought processes, research techniques, and findings.”
In one exercise with his students, Steiner and the class worked to edit the introductory section of an abstract that was too long for publication. The class brainstormed new phrasing and cut unnecessary words, reducing 73 words to 46 words.
As an experiment, the class then asked an AI program to shorten the same text. Almost immediately, the program generated a 48-word abstract. Yes, it was much quicker, the class found, but the context was lost.
For example, the AI program asserted that the problem under consideration was “significant,” Steiner says. That word in scientific writing is usually reserved for its statistical connotation.
The experiment was a lesson in intentional choices that come with the writing process and are learned with experience. Over time, it’s possible that AI programming will find a role in scientific writing, but it won’t replace the need for researchers to cultivate writing skills.
“Efforts to improve the quality of scientific writing are not distractions from research,” Stainer writes. “Rather, skillful writing exemplifies the creative process of planning, action, reflection, and communication that remains at the heart of the scientific enterprise.”
Future skill building
As a humanities undergraduate, scientific research was a mechanism for Steiner to return to creative pursuit in many ways. As his career evolved, he also found a love for publishing commentaries and perspective essays.
“It was a hobby I could call part of my profession,” he says.
His hope for his students is that they will also develop a fondness for the skills necessary to write scientific papers, whether they be an abstract describing a clinical study or a short essay about a professional experience.
With five to seven students per class, and already in retirement, Steiner knows his reach is finite, so he’s partnered with other School of Medicine faculty members to carry the torch.
Sarah Brewer, PhD, MPA, assistant professor of family medicine, and Darcy Thompson, MD, MPH, professor of pediatrics, plan to keep the courses going, providing a resource for early and mid-career researchers who want to improve their writing skills.
“It’s important that we help our researchers develop writing skills, and I’m grateful to ACCORDS for sponsoring the work and for those who see the value in it,” Steiner says.
