One day this May, while grading papers, a ritual at the end of the semester, I came across a character description that had absolutely no basis in the novel I was reading. Two years ago, I would have stopped and thought that this was unfortunate: Either the student hadn't really read the novel carefully, or they were simply relying on an inaccurate memory of the scene rather than going back to check the veracity of their interpretation. big transaction.
But now, a year and a half after ChatGPT was made public and large-scale language model (LLM)-based chatbots entered everyday use, things are different. (To resist the common assumption that these programs have agency or anthropomorphic qualities, I will refer to them as “machine learning applications” and “LLM-based chatbots” rather than “artificial intelligence” or “AI.”) Instead of seeing this inaccuracy as something I could flag so that the student could come back and discuss it with me, I had the uneasy feeling that what I had been reading for a page and a half was synthetic text pushed out of a chatbot rather than the textual expression of the ideas of a young woman I knew, with whom I shared a university seminar room twice a week.
While there may be reasons why faculty require students to use chatbots for their schoolwork, the use of a chatbot did not align with my learning outcomes for this assignment. We have also discussed at length in class the issues with the functionality of LLM-based chatbots in relation to the type of work we do in class (and in general), and even conducted class exercises that demonstrated how ChatGPT would return incorrect responses to very basic questions about literary works.
I also modified some of the assessments to significantly reduce the temptation to use chatbots. Still, I couldn't help but suspect that this particular erroneous sentence was more likely to be chatbot text, not stated as such and claimed to be the student's own ideas and words. Reading the rest of the essay and encountering no other red flags, I decided to reconsider my suspicions. As in Kurt Vonnegut's immortal words, and as is often the case in the life of a college professor at this moment, such is the case:
In addition to being a lecturer in humanities, I am also the director of academic integrity at my university. It has been a year and a half since the mass release of “generative AI” programs to the public, and in that time, conversations have been taking place about these applications and their relationship to teaching and writing. I have read far more books about algorithmic techniques than I ever expected, and I have heard and spoken about them in faculty meetings, administrative meetings, student-led initiatives, and regular conversations with professors who find their students trying to appropriate LLM-based chat bot texts as their own, not to mention in hallway and water cooler conversations.
In all these forums, I heard and read a variety of metaphors about these machine learning applications. There was the saying, “When the calculator came along, people were shocked.” There was even a performance where the first speaker at a conference read out an overly thoughtful and abstract opening statement about “AI” and then announced that what he had just read was generated by ChatGPT.
But what annoys me the most and what I see the most is that the instructors which one Embrace the new frontiers that LLM-based chatbot technology is opening up or They fear them, as if they have no other (or much) of a choice.
As you may have guessed by now, I am not a fan of these technologies (my overall view on this whole issue is pretty well summarised in a recent book by Ulises A. Mejias and Nick Couldry, and you can read my own view here). And here I should admit (in another way) that “these machine learning applications have all the benefits, even if they can do real harm”. But in the realm of cultural analysis, critique and interpretation, no one has yet shown me that chatbots are really useful. And that doesn’t even touch on all the environmental, bias and copyright issues. I have seen that chatbots can be useful in some computing contexts, and they may eventually be good for business analysis and so on, but there is no reason to concede to OpenAI, Microsoft or Alphabet the future of humanities education.
Either way, my point is that the false dichotomy I hear over and over again between “acceptance” and “fear” is just that, a false one, and it prevents us from recognizing more nuanced and possible responses to this age of machine learning. I am not embracing the miraculous future of this technology, nor am I afraid of it. But I am troubled by how chatbots are unfortunately impacting my ability to trust my students.
Let me be clear here: if students use chatbots and cite their use, this has absolutely no bearing on what I am arguing here. Some instructors ask their students to use chatbots based on their LLMs and allow them as long as they cite them. That's fine. I don't endorse that approach, but I understand that other instructors give their students other learning outcomes and goals, and I'm not here to impugn those outcomes, goals, or practices. Even if my students, who I ask not to use chatbots, did use chatbots and cite their use, I would not accept it as an assignment well done, but it doesn't affect my credibility.
I don't think students think that their unauthorized use affects their personal relationships. But those of us who actually still believe in the enlightening power of higher education cannot see the instructor-student relationship as an instrumental exchange of goods (written assignments) and payment (grades), or as a mechanical input-output relationship. In classrooms, office hours, and meetings, if we make the effort, if we foster dialogue and the sharing of perspectives in our joint examination of reality and pursuit of truth, a true mutual sharing between people can be created. And the creation and grading of assignments is an extension of that mutual sharing in the relationship. But that examination and pursuit together requires honesty in the instructor-student relationship, which means that both parties need to be honest with each other in their communications.
The fundamental problem with behaviors traditionally understood as cheating, plagiarism, and fraud is not a failure to put in effort, but a failure to represent oneself honestly to others. We are social beings, and therefore must be able to trust one another. In an academic environment, part of trusting one another is being able to trust that what a student tells me in an academic piece accurately reflects how the piece was created. If a student has someone else write an essay for them and then submits that essay with their name at the top, that submission is a lie. The convention of writing one’s name at the top of a paper or file makes it hypocritical and does not accurately reflect how an essay was created. Writing one’s name at the top is a claim that “I made this.” If the words or ideas in the piece are not ours, we cite where we found those words or ideas. This convention of citation allows readers and viewers to be confident that even when we put words and ideas into a piece that are not ours, it still reflects the reality in which the piece was created. Aligning representation with reality creates honesty and builds trust.
The essay I read in May may have represented the reality of how it was created, or it may not, but the hype and constant debate surrounding chat bots in law programs, and frankly the fact that I have seen so many students try to pass off synthetic texts as their own without proper attribution, has left me in a state of somewhat distrustful of students right now. And I don't like that.
I went to college because I found Old English poetry in a library book in my early twenties (I had to go to college to know how Old English worked). And I went to graduate school because I found the richness of a liberal education far more valuable and stimulating than any other possible path for my life. I became a (non-tenured) professor because I wanted to share all I learned with those yet to come and learn from them, too. I stayed because I love probing reality every day with my students and colleagues. We both seek knowledge, and dare I say wisdom. And to continue that dialogue, we need to be able to trust each other. I look forward to when this moment calms down and we can once again focus on other, more important issues: celebrating the arts. questioning systems of power and thought. compelling arguments. ways to imagine a better future for ourselves and each other.
In the meantime, I will likely continue the conversations I had with my students earlier this semester:
Student: Why can't I use a chatbot to write this essay?
Me: I don't care what OpenAI's products can do. I care about what you think.
