What does it take to be true?
A new school year is about to begin here in Canada, and ChatGPT is on the minds of many: teachers and students, of course, but also parents and administrators, all of whom are invested in the quality of education.
Even more of us should be concerned, however, about inappropriate student assistance from artificial intelligence (AI). Prospective employers trust high schools and universities to produce certain levels of skill in their graduates, including the skills of research, reflection, and writing—each of which is now subject to doubt because of AI. Governments likewise rely on diplomas as a certification—formal or informal—when an international student applies for permanent residency. A lot is at stake here.
ChatGPT is still in pretty rough shape, of course. It notoriously makes up both fake information and fictitious citations, so it cannot be relied upon as a research assistant or tutor. It also composes in a style that is so conventional—both bland and correct—that an attentive reader, and especially an attentive grader, should be able to spot it. (Few students, even good ones, have its command of punctuation. If a student paper uses a semicolon correctly, I immediately start sniffing for plagiarism or AI.)
Later versions of ChatGPT, however, are already in the pipeline. How should we regard that?
It occurs to me that assessing the morality of such assistance and then modifying our pedagogies accordingly might be aided by considering other forms of assistance we already take for granted—or don’t.
Ghostwriting, for example. What do you make of it? A recent essay in The New Yorker by Prince Harry’s ghostwriter, J. R. Moehringer, opens up this shadowy world in a way that shows it isn’t at all unethical. But why not, when Moehringer’s name doesn’t appear on the cover of Spare?
Ghostwriting isn’t an ethical trouble spot for the same reason that no one should be shocked to find out that politicians employ speechwriters and nightly talk-show hosts rely on writers’ rooms. These people need high-level help. Only a naïf thinks they utter only words they themselves compose.
The problem arises when such writing diverges so sharply from the actual ideas and style of the subject or speaker that the audience is now dealing with truly a different person. Now the prose isn’t true—true to its ostensible source.
Professional authors themselves get help, and often quite substantial help, from editors, whose names appear only in the acknowledgements of books or on the mastheads of periodicals. I am grateful for the ministrations of such shadowy angels—up to the point that they try to make me sound like someone else. Then I bristle, both on my own behalf and on behalf of the reader. It’s my name at the top, after all, and this writing should be true to me—even if it is in fact a better voice than I can produce on my own. The writing shouldn’t be wildly, or even just significantly, better, let alone different—whether in concept or in expression—than I could and would produce.
Again, adult readers know about editors and we all happily conspire together in the polite fiction that the final version of the article or book provides communion merely between author and reader. We acknowledge that the version in question is editorially mediated. But only if the mediation amounts to what truly is a co-authored product should the editor’s name be lifted out of the acknowledgements and put on the title page where it belongs. (The late James Sire, long-time senior editor of InterVarsity Press, made it clear late in life that he should have asked to be named co-author of Francis Schaeffer’s early books, since Jim claimed to have rewritten them from top to bottom.)
Stepping sideways to the musical recording studio, arrangers and producers usually affect the final version of a song quite substantially—and even more nowadays in the world of multiple tracks, sampling, AutoTune, and other manipulations. The artist still gets most of the attention, but at least the producer is always named as such, and industry aficionados recognize to what extent each recording is truly a joint effort. (Think of George Martin, Phil Spector, Quincy Jones, or Dr. Luke.)
At a minimum we expect the named artist to perform at least the melody for at least most of the song. And we expect the songwriting credits to reflect all creative input. (I do wonder, I confess, about a half-dozen composers listed for a three-minute pop song. But if they feel they have to share the credit and royalties six ways, who am I to judge?) Whatever else happens in the recording, we’re generally content to let the artist and producer do their respective things.
If the final product is simply too weirdly divergent from the artist’s own talent and persona, however, murmurs arise. Something false has happened.
Another possible parallel is the work of an eminent artist. In assessing the worth of a painting or sculpture, we want to know if the artist produced the entire piece, or most of it, or merely the inspiration and supervision of it. Millions of dollars are sometimes at stake in the differences between “from the hand of,” “from the studio of,” or “from the school of.” What are the pertinent differences in written work?
Indeed, in scholarly writing we are supposed to acknowledge the help of others, whether in acknowledgements or footnotes. But it remains a judgment call as to just how much the author is obliged to credit others for what she says when what is at stake isn’t a direct quotation, figure of speech, or a fact.
What warrants specific mention when it’s “just” a good idea one picked up along the way? One is reminded of the cynic’s rule of thumb: “Stealing from one author is plagiarism. Stealing from several is research.”
Back to the academy now, and we descend into the murky zone of unofficial aides to student writers: parents, friends, spouses—even children, if the adult learner is blessed with sufficiently skilled offspring. (I have such a colleague, a fine social scientist, who has run his manuscripts by his brilliant eldest child for more than a decade.)
Parents have been staying up late for a long time now assisting their children with speeches, science fair projects, and even just homework. Friends help friends find sources, compose outlines, and proofread drafts. Spouses get their beloved thesis writers over humps to get them finally graduated. I expect that none of these benevolent folk suffer a single pang of conscience. Should they?
What do we think of students writing papers with help from parents and friends? The questions come quickly: What kind of help, and how much of it? What about essays aimed at competitions, for university entrance, scholarships, and the like? What does a degree signify if a major part of it—a thesis or dissertation—was largely composed by someone else?
Surely the issue is that word “largely.” I expect most of us would raise a yellow ethical flag, if not a red one, in cases in which the writing no longer reflects the skill and effort of the writer but is now truly a team effort. That binary judgment—Is that his work, or is it his-and-hers?—seems reasonably clear and enforceable, requiring as it nonetheless surely does a judgment open to question by other experts in the field. How much is too much? A product substantially/mostly/importantly untrue to the named author.
What we will need to consider, therefore, is how much ChatGPT involvement renders an assignment truly co-authored, rather than assisted. And this question will apply not only to the originality of the final prose, as if a student could legitimately take an AI-generated piece and just re-word it. The research, yes, and the structuring of the argument are also key elements in such study. If ChatGPT is doing a substantial amount of any such work, we face again the issue of co-authoring.
As a first-order approximation of policy, therefore, I suggest that instructors make very plain to students what is expected from the student for each element of an assignment. The specter of ChatGPT can, in fact, prompt us to do what we ought to have done heretofore: place appropriate limits around cooperation with others in the production of written work, whether human or artificial intelligences.
To offer students such specificity for each stage of an assignment, furthermore, will require us to think through and articulate more carefully than we have before just what good question-asking, good research, and good outlining look like. What are the formulas, the heuristics, that make ChatGPT already a better writer than many of our students? These we ought to be articulating and teaching our students so thoroughly that they don’t need ChatGPT.
So far I have made the situation relatively easy by considering the issues binarily. If a piece of work is substantially true to the author—to the author’s research, reflection, and writing; to the author’s own talent, skill, effort, and voice—then all should be well. If instead that work is substantially dependent on the contribution of another, or several others, then it cannot count solely to the credit of that student.
A much more difficult situation for many students and instructors, however, is work that is graded on a basis other than pass/fail. Quality of research, quality of reflection, and quality of writing all can be better or worse, a B+ or a C-.
I had recurring arguments in my last academic post with well-meaning students and fellow teachers regarding the “editing” of assignments by native speakers of English on behalf of international students. I averred that quality of expression was ingredient in the final grade, so that “touching up” (let alone substantially revising) written work by more skilled writers was not ethical.
Instead, I suggested, students should take assignments that were already graded to our writing center for help from tutors toward the next assignment. Otherwise, student grades would vary with the level of outside help requested and received, which seemed unjust to their fellows. And students would not get the benefit of receiving grades that truly represented the quality of their own work, which seemed unjust to them. (Some of my colleagues came around to this way of thinking. Others did not.)
In another sideways jump, consider the literal and metaphorical fields of live performance. Coaches of athletes and teachers of musicians can get as involved as they like in the training of their charges. But when the crucial moment comes, they have to stand aside for the live performance, on which everything depends.
I find this parallel instructive. Many instructors concerned about ChatGPT have decided already to implement this model so that some or all of a student’s grade will come from live performances. This policy makes perfect sense on its own terms, of course.
We also have to recognize, however, that this policy imposes its peculiar burdens on both students and instructors. What sort of live performance, written or oral, can substitute for the long essay, with the expectation of careful revision and conscientious proofreading? Do we really want to institute a culture—already all too typical of the harried pace of most secondary and post-secondary schooling—of first draft-as-only-draft?
Do we want now to penalize students who have slower mental “clock speeds” than others (even as over time they produce equal or better work), or who suffer more from performance anxiety than do their peers (without qualifying for a true learning disability and therefore special treatment)? Yes, thinking quickly and coolly is a fine skill set. But making everything dependent on live performance will privilege unduly a skill set that is not necessary to many important occupations, to effective and creative thinking, and, indeed, to the properly examined life.
The advent of ChatGPT thus comes as a warning to us: both to examine pedagogies that already deserve re-thinking and to prepare pedagogies for the steady rise of AI. I haven’t yet asked ChatGPT to generate an essay on “appropriate uses of ChatGPT in written assignments,” and I won’t. But I wish I could write a better one now myself.
Maybe soon I will, though—with your help?
John Stackhouse holds the Samuel J. Mikolaski Chair of Religious Studies at Crandall University in eastern Canada. His most recent books are Can I Believe? Christianity for the Hesitant and Evangelicalism: A Very Short Introduction, both from the Oxford University Press.