Revising Historical Writing Using Generative AI
An Editorial Experiment
Despite widespread fear surrounding their arrival, generative artificial intelligence (AI) tools like ChatGPT are quickly being fused into many of the platforms that we, as historians and teachers, regularly use. As these tools become even more integrated into our world through Microsoft Word, Canva, Google, and the like, we must explore how they can make historical research and writing more efficient and intelligible, as well as how we can teach students how to use them ethically.
As a wordy writer, I often struggle with crafting concise historical writing. And so generative AI caught my attention when I saw a social media discussion about how it could help shorten and strengthen one’s writing.
Because most generative AI tools use your input data to improve their models, it was important to me to find a tool that respected my privacy and intellectual property. Luckily, one of my favorite web-based platforms, Notion.so, had just added an AI tool and specified it did not collect user data either for training or other purposes. I have long used Notion for writing, organizing, list-making, website publishing, and more. Teachers and students can even get a free education plan, though its AI tools require a monthly subscription fee.
Like many similar AIs, Notion’s features provide several tools to “edit or review selection”: improve writing, fix spelling and grammar, make shorter, make longer, change tone, and simplify language. There are also features to summarize, translate, or even to make to-do items from a passage. I found these features to work best at the paragraph level. The resulting process is slow and fiddly, because it requires me to repeatedly process text in small increments, but it has produced the most useful results.
AIs provide several tools to improve writing, fix spelling and grammar, make shorter, make longer, change tone, and simplify language.
I have personally found the process of pairing the “improve writing” and “shorten text” functions to be especially useful. For example, I input the following sentence: “Through these narratives of a romantic past, projects like the Old Spanish Trail highway contributed to larger patterns of adopting Spanish and Anglo fantasy pasts to bring together residents in common understandings of growing regional identities throughout the US in the early 20th century.” Following the AI’s suggestions, I was able to craft the following more effective sentence: “In the early 20th century, projects like the Old Spanish Trail highway adopted Spanish and Anglo fantasy pasts to help create regional identities in the US.”
Despite the great benefits of this process, AI-aided editing still requires a tremendous amount of author oversight and judgment. Accepting all changes the AI makes would result in a terrible loss of meaning and oversimplification of the writing. The “shorten text” tool, in particular, removes much of the nuance and detail so important in historical writing. It frequently paraphrases or eliminates quotes and specific examples. It also often seeks to remove or shorten historiographical notes, e.g., it cut phrases like “The historian So-and-so has shared. . . .” While I often ignored these suggested edits, the tool did help me question if a quote could be paraphrased, when there were too many examples in a paragraph, and where the historiographical notes were superfluous. By exploring the suggestions one change at a time and determining which edits to keep and which to ignore, I found my writing improved in clarity, length, and overall quality. As I worked through my writing paragraph by paragraph, sentence by sentence, this provided a structured process to interrogate my own writing, helping me to find missing or weak transitions, remove redundant examples, and reorganize as I edited.
Initially, I had ethical questions about using AI to revise my writing—was I cheating by using it to improve my writing? Where should I draw the line in the appropriate use of generative AI in writing and editing? I concluded that editing using this tool can be not too different from working with a very patient and thorough editor who makes suggestions about your prose. It’s also far cheaper—even with a monthly subscription fee—making this a cost-efficient approach for students and faculty without many resources. Additionally, this process still requires a tremendous amount of oversight by the author, deciding which changes to implement and which to discard—not dissimilar, again, to working with an editor. The process continually required my expertise, assuaging my initial concern. It was far from an easy process of inputting text and exporting a clean draft.
Editing using this tool can be not too different from working with a very patient and thorough editor.
In addition to simple improvements to my prose, I have realized using AI can improve my writing by identifying recurring problems while helping me to hone my craft. This in turn helped me see that training students to use generative AI to edit their own writing could similarly benefit them. For example, I could ask them to use a generative AI function like “improve my writing” or “shorten” on a paragraph they wrote, and then have them share the suggested changes and why they think the AI suggested them. Such an exercise helps students identify issues in their writing such as passive voice, wordiness, redundancy, and run-on sentences. Asking students what suggestions they would or would not integrate also helps to teach them about writing and editing as a process. For example, this could foster a discussion about the value of direct quotation versus paraphrasing by providing concrete examples from the student’s writing. This type of editing and discussions about it can strengthen students’ writing skills while also potentially reducing some of the instructor’s burden in editing student work.
Besides the pedagogical benefits, embracing generative AI in the classroom provides instructors with an opportunity to explore issues of AI directly with students. This kind of open discussion can help students disentangle when these tools can be used appropriately in the classroom and when they constitute plagiarism. Discussing how generative AI can augment work, rather than replace it, can offer fruitful conversations, as well as prompt discussion of how technology is not always a magical fix. As our understanding of ethics and appropriate citation of AI use evolve rapidly in the coming years, we should engage students in these discussions instead of lamenting students’ abuse of generative AI from our lofty ivory towers.
This article used both generative AI and human editors through its revision process.
Lindsey Passenger Wieck is an associate professor of history and graduate director of public history at St. Mary’s University. She tweets @LWieck.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. Attribution must provide author name, article title, Perspectives on History, date of publication, and a link to this page. This license applies only to the article, not to text or images used here by permission.
Please read our commenting and letters policy before submitting.