AI Thoughts Continued: Grammar v. Content?

[smccormack]

In “Facing My Fears of AI” (June 2024) I blogged about my desire to find ways to integrate artificial intelligence (AI) into my classroom assignments in a way that will enable students to learn to use new technological innovations, while also respecting academic integrity. While still a work-in-progress, I plan to share the ups and downs of my experiences as they develop.

In my summer courses I assigned research projects that required the use of library-based databases and materials. Knowing that students are increasingly likely to use AI, even when discouraged or even prohibited, my assignment instructions included the following statement: “Students should submit their own original work and cite all sources using MLA format. While use of AI is discouraged, any use of artificial intelligence should be cited as a source.”  

We discussed as a class what it meant to submit original work and the importance of using each assignment in our introductory-level history course as a stepping stone to the challenges that will come later in their college careers. The vast majority of students followed my instructions and submitted work that was consistent with that of typical first-year students at my college. Several students came to me for help or consulted our college librarian outside of class for additional assistance. 

Unfortunately, 10% of the students submitted work that was AI-generated and none of those submissions cited any AI-tool as a source per my instructions. I ran all student submissions through an AI and plagiarism detector and in each case it was determined that the work was 90% or more AI-written. 

I reached out to each of the students to discuss the use of unattributed material as an academic integrity issue. Approximately half of the students did not respond to my email and accepted a zero for the assignment with no discussion. The other half denied any use of AI. When confronted with a detailed report from the detection tool, however, they changed their stories. They admitted to having used AI tools such as Grammarly and Brainly, they said, but not software that “created content.” 

I learned a valuable, if not confusing, lesson from this experience: my students see a marked difference between tools such as Grammarly and Brainly, versus Chat GPT and the like. And so here is my quandary: is there harm in allowing the use of grammar-checking AI versus content-producing? Or, do they both create a product that is not the student’s original work? 

My initial feeling was that grammar-checking tools can be helpful in “cleaning up” student work. In speaking with students, however, I realized that in a 3-4 page short essay they were accepting, without much thought, upwards of 30-40 suggested changes by Grammarly or Brainly. In other words: they were changing their entire paper based on suggestions from the AI tool.

As such, I’ve realized that either my instructions are going to need to be more specific or I need to re-evaluate my comfort level with grammar-corrective AI. 

Some faculty reading this blog might ask: “I use Grammarly … why can’t my students?” This question is certainly valid. The difference for me, however, is that many of my students struggle with grammar. A tool that corrects their errors without requiring them to understand the fundamentals behind the mistakes is, in my view, unlikely to progress their writing skills.

I’d love to hear from others in the Macmillan Community about AI policies: how are yours evolving with new technology and new challenges? Please share.