What’s Routine Work, Anyway?

[mimmoore]

I recently wrote about my struggle with AI overload, a post prompted by an email predicting that AI would soon take over the “routine work of scholarly editing and publication.”  What, I wonder, makes something routine in scholarship or academia?

My social media feed, guided by my history of clicks, has lately been flooded with advertisements for AI platforms that promise to clear my hectic schedule by picking up some of the drudgery of my job—specifically, providing feedback on student writing, crafting standard letters, or even putting together meeting agendas.  Apparently, those tasks are “routine” and can be handed over to a trained LLM (Large-Language Model). For a fee, of course.

Except I don’t think of these as routine.  (Well, maybe writing agendas could be seen as routine; still, I think reducing the number of required committee meetings and service responsibilities for faculty who teach four or five courses per term is a more helpful solution than letting a machine throw together a bulleted list of agenda items.  Just a thought.)

I suspect that most of us who teach composition would not agree that response to writing is so “routine” that it should be handed off to an algorithm.  But what about other tasks, such as annual self-evaluations or recommendation letters for students?

Over the Thanksgiving break, I wrote three recommendations for students applying to graduate programs.  The process took over six hours of my “vacation days.”  Surely, one might suggest, I could have saved time by generating a letter template from ChatGPT and tweaking the details, right?

My answer, without hesitation, is “no way.”  It’s not that I produced something particularly masterful in those letters over six hours, nor do I expect that the letters will be the deciding factor for graduate admissions committees.  And given enough detail, could the AI produce something similar?  Perhaps.  But the AI will not have the joy of reviewing, remembering, discovering forgotten moments and sorting through assignments, emails, projects, or random hand-written artifacts from these students, who have shaped the way I teach just as much as I shaped their undergraduate experiences.

For each student, the process began with a review of a CV, followed by a transcript review to find all the classes taken with me, as well as classes in which the student worked with me as a supplemental instructor or writing fellow.  I looked for conference presentations and academic service—noting that two of the three students attended faculty-focused conferences to present research with me.  I thought about my initial impressions of each one and how those impressions changed as I saw their particular passions appear in course projects—or as I came to rely on their help as supplemental instructors in my classes.  I recalled when they frustrated me, when they made me laugh, and when they made me think differently about concepts or my own teaching practices.   I imagined what each would bring to a graduate program, of course, but I also tried to articulate what they brought to our undergraduate program—and to me.  My teaching is better because of these students, and I wanted to articulate why as part of these letters. 

I suppose one could say that given all this information, ChatGPT could have produced an effective letter.  But truthfully, the letter wrote itself once I had worked through the process of remembering.  And I suppose others might suggest that it was the extraordinary qualities of these particular students that made the writing process meaningful; had the students not been quite so stellar, perhaps the process would have felt more routine—and perhaps I would have given it over to the AI in that case.

But I don’t think so.  My limited interactions with students who are not quite so remarkable still contain valuable stories and possibilities.  Reviewing those stories is not a waste of my time, however constrained that time may be. 

So, I am back to my original question: what qualifies a task as “routine” to the extent that it can be relegated to AI?  Who decides that?  And how will those decisions—whoever makes them—impact our work going forward?  These are just some of the questions that give me pause. 

For now, I need to proofread some recommendation letters.