Deciding Whether to Use Artificial Intelligence
|
|
|
The introduction of ChatGPT 3.5 in November 2022 has led to a rethinking of the process used by engineers and scientists to produce technical documents [1]. Since that release, much debate has arisen on whether, when, and how much engineers and scientists should use artificial intelligence (AI) to assist in their writing. This website presents multiple perspectives on these three questions.
Case for AI-Assisted Writing. Many respected engineers and scientists have begun experimenting with artificial intelligence both to improve specific aspects of their writing and to streamline the writing process [2]. While few professionals are arguing for using AI to write all documents or even to use AI to perform all major steps of the technical writing process (shown in Figure 1), many are finding success in having AI assist on portions of the process. For instance, many engineers and scientists who are writing in a second or third language find AI extremely helpful in revising documents to connect ideas and to adhere to the many rules of grammar, punctuation, and usage. 
Figure 1. Three important stages of the writing process: thinking, drafting, and revising.
As depicted in the analogy of Figure 2, AI is not only helping engineers and scientists write a significant slice of current documents but also helping write other documents that would have been too large or difficult to take on with a traditional approach.
Figure 2. Analogy that AI is assisting not only on a large slice of documents but also helping write new types of documents that would otherwise be too time-consuming to attempt.
Case Against AI-Assisted Writing. Writing done with artificial intelligence has distinct weaknesses. As indicated on the page of telltale signs of AI writing, the biggest weakness for engineers and scientists is a lack of technical precision. Writing generated by AI can be "erroneous, misleading, or entirely irrelevant" [3]. In general, the more technical precision that is needed in a document, the less effective that AI is at drafting the document. The imprecision of AI writing stands as a strong counterargument for using AI to write about technical content.
Another counterargument about AI-assisted writing concerns AI not properly crediting sources. Not crediting a source can severely damage one's reputation in the professional community. Many such examples of damage exist, such as James Watson not properly crediting Rosalind Franklin for her role in discovering the structure of DNA [4]. Properly crediting someone else requires precise writing, which is not a strength of AI. |
Still a third counterargument concerns the learning lost by using AI. This argument is that engineers and scientists should not use AI for technical writing assignments because doing so allows students to bypass the learning of the content for the document. When you draft a document, you wrestle with the precision of your ideas. As Francis Bacon said, this wrestling sharpens your understanding of that content [5].
The Diffusion of AI-Assisted Writing. Since November 2022, AI-assisted writing has begun diffusing into engineering and scientific documents. The theory for the diffusion of innovation helps explain the adoption of and resistance to this innovation [6]. As depicted in Figure 3, when an innovation becomes available, a group called "early adopters" experiments with the innovation and, if they find enough value, they adopt that innovation. Already, many engineers and scientists have adopted AI to assist in researching, drafting, or revising some of their documents. On the other end of the curve, because writing with AI threatens deeply established practices, this innovation has encountered significant resistance [7]. Many not only have publicly decried the innovation but are blocking its use in their institutions, departments, or courses. 
Figure 3. Order of adopters for an innovation. In this theory, when faced with the decision of whether to adopt an innovation such as using AI to assist with writing, people take one of five stances.
For this middle group, an important question is deciding when it is appropriate to use AI to help write documents. This answer is complicated. Much depends on the content, audience, purpose, and occasion of the document. For instance, in many cases, using AI as an assistant is fine for specific stages of the writing process, but not for others. For instance, some argue that AI is not appropriate for drafting new content when you are still learning about the content. In such a case, using AI to draft deprives authors of the insights that naturally arise during the drafting stage when authors convert their thoughts into words on paper.
Purpose of This Website. The purpose of this website is to provide insights about using AI to assist in technical writing. One goal is to help engineers and scientists (and teachers of technical writing) decide whether or when to adopt (or teach) this innovation.
At the site, one webpage highlights examples of AI-assisted writing by professional engineers and scientists. Another webpage discusses stylistic and grammatical signs for when a document is written with AI. Knowing the style of AI writing is valuable for multiple reasons, one being to craft AI prompts that mitigate AI's weaknesses. Yet another page presents quotations from faculty about allowing or not allowing students to use AI to assist in their technical writing. Still another forthcoming page is an experiment to determine how much time using AI would actually save students performing a common writing assignment in a technical writing course. 
|
Photo by pxibay.com (CC0 1.0)
My Position on Using AI to Help Write Technical Documents
Photo by Shanghai Jiao Tong University
|
Photo by PickPic
Photo by Penn State
|
Prompts to Address Weaknesses
of AI Writing 
Drawing from Adobe Stock
|
References
- Berber Jin and Belle Lin, “OpenAI Unveils GPT-5, Its Latest and Most Powerful Model, After Two-Year Wait,” Wall Street Journal (7 August 2025), p. A-1.
- Catherine Berdanier and Michael Alley, "We still need to teach engineers to write in the era of ChatGPT," Journal of Engineering Education, vol 112, issue 3 (July 2023), pp. 583-586.
- Qinjin Jia, Jialin Cui, Haoze Du, Parvez Rashid, Ruijie Xi, Ruochi Li, Edward Gehringer, “LLM-generated Feedback in Real Classes and Beyond: Perspectives from Students and Instructors,” Proceedings of the 17th International Conference on Educational Data Mining, (Atlanta, Georgia: International Educational Data Mining Society, July 2024), pages 862–867.
- Sharon Bertsch McGrayne, "Rosalind Franklin," Nobel Prize Woman in Science (Washington, DC: Joseph Henry Press, 2002), p. 330.
- Sir Francis Bacon, "Of Studies," The Essays (Harmondsworth: Penguin, 1985).
- Everett M. Rogers, The Diffusion of Innovation, 5th edition (Free Presse, 16 August 2003).
- Jessica Grose, "These College Professors Will Not Bow Down to AI," opinion, The New York Times (6 August 2025).
*AI was not used to research, draft, or revise this webpage.