Composing with AI

Analyzing Submission Policies

For the study, I analyzed the content of ten scholarly publisher and journal submission policies on use of ChatGPT and other artificial intelligence. As this is a preliminary study, I chose depth over breadth, focusing on a smaller set of policies. Table 2 lists the policies I read and the online locations where I accessed them.

Table 2: ChatGPT and AI Policies Studied

Accountability in Research (ACL) Conference

https://2023.aclweb.org/blog/ACL-2023-policy/

Computers and Composition (C&C)

https://www.elsevier.com/journals/computers-and-composition/8755-4615/guide-for-authors

Elsevier

https://www.elsevier.com/about/policies/publishing-ethics/the-use-of-ai-and-ai-assisted-writing-technologies-in-scientific-writing

Journal of the American Medical Association (JAMA)

https://jamanetwork.com/journals/jama/pages/instructions-for-authors

Nature

https://www.nature.com/articles/d41586-023-00107-z

Oxford University Press

https://academic.oup.com/pages/authoring/journals/preparing_your_manuscript/ethics#Authorship

Proceedings of the National Academy of Sciences (PNAS)

https://www.pnas.org/post/update/pnas-policy-for-chatgpt-generative-ai

Science Journals

https://www.science.org/content/page/science-journals-editorial-policies

Taylor & Francis

https://newsroom.taylorandfrancisgroup.com/taylor-francis-clarifies-the-responsible-use-of-ai-tools-in-academic-content-creation/

World Association of Medical Editors (WAME)

https://wame.org/page3.php?id=106

These ChatGPT and AI policies were often embedded in larger author guidelines or submission instructions. I attended specifically to the sections that address AI and ChatGPT. This was a convenience sample made up of policies published in the months following ChatGTP’s release in late November 2022. Because I was interested in publishers’ and journals’ initial responses, the study corpus comprises policies I discovered through web searches from January to May 2023. These policies represent a range of disciplines to ascertain approaches to authorship across academic fields. However, the sample is not statistically representative or large enough to draw conclusions about scholarly publishing writ large or about particular academic disciplines. Still, the findings provide insight into ways in which scholarly journals and publishers (implicitly) construct authorship and writing in their policies on generative AI and ChatGPT.

In addition to paraphrasing how each policy defines authorship, I coded for whether each policy allows or forbids listing generative AI as an author and why. I also coded each policy for whether it allows or forbids inclusion of AI-generated text in submissions, under what conditions, and how such use is to be acknowledged. As a researcher interested in ways in which these policies discuss writing, I classified the policies’ references to writing according to James Britton’s (1982) taxonomy of writing. According to Britton, writing can be classified as transactional, expressive, or poetic. These types of writing are distinguished by each’s purpose (and the roles of participant and spectator)¹. Transactional writing, for example, communicates information and can be informative, regulative, or persuasive. Expressive writing facilitates learning and construction of the self. It connects the abstract to personal experience. Finally, poetic writing presents writing as art. It is an object to be enjoyed for its aesthetic properties (p. 155). The policies analyzed for this study privilege the transactional. While perhaps unsurprising given the purpose of academic journals and publishers, this framing is limited.

Using these lenses for analysis, I found policies to limit authors to people, allow inclusion of some AI-generated content under certain conditions and citation guidelines, and frame writing as a textual product performing transactional functions.

¹James E. Kinneavy (1969), of course, included a fourth category in his basic aims of discourse: expressive, referential, literary, and persuasive.