The Detriments of AI for University Education
Written by: Samuel MacQuarrie
Photo by Aerps.com on Unsplash
It has become increasingly clear that heavy reliance on artificial intelligence (i.e., Large Language Models) is not incompatible with the attainment of high grades. Indeed, in some cases, students who use AI extensively may outperform their conventionally minded peers in this regard. As a student of the arts, I cannot speak to whether this is a regrettable state of affairs among the sciences, but among the arts, where much depends upon a student’s ability to write, I believe censure is long overdue.
My principal contention is that grades attained via written assessments produced through AI cease to indicate genuine academic standing. Essays (among other written assessments) are exercises through which students develop the ability to think clearly, argue for a position, and communicate effectively. When these intellectual processes are passed off to AI, the resulting grades are indicative of neither academic ability nor intellectual development. If this contention is justified, then there are a number of serious consequences that must be accounted for should Saint Mary’s (among other universities) maintain its current state of enthusiasm for the integration of AI into university education.
Let us consider the first consequence. An A+ achieved honestly, as a result of one’s own intellectual labour, is quite distinct from one obtained by presenting a large language model’s output as one’s own work (if the reasons for this are not obvious already, they will be made so in due course). Transcripts, however, do nothing to reflect this difference and evaluators of them are left to wonder whether the grades before them are actually representative of a student’s abilities. This not only devalues the achievements of students who complete their work independently but further it undermines the significance of attaining a university degree. Second, at least some students who manage to attain high grades relying heavily on AI are unlikely to objectively assess what they have actually achieved. That is, they are likely to develop a false sense of accomplishment. Deficiencies they may otherwise have remedied are likely to go unnoticed as their intellectual development is substituted for the attainment of grades. Finally, and closely connected to the preceding point, students who rely heavily on AI throughout their undergraduate education are likely to find themselves unprepared for environments, such as professional degrees, where assessments are carefully invigilated and one is expected already to have developed the ability to reason independently.
Now of course it might be objected that my account distorts the role of AI use in education. After all, students have long relied upon technology to improve their work. Calculators and search engines, for instance, have not led students to flounder in their ability to perform mathematics or undertake research; rather they have allowed students to spend less time doing menial tasks and more time engaged with intellectually demanding work. Improvements to efficiency are not necessarily detriments to education. AI, on this view, merely removes menial labour and allows students to focus on more important forms of intellectual engagement. A student who develops their own ideas but uses AI to refine their expression has simply employed a more efficient method of communication. Furthermore, since AI is becoming increasingly accessible, grades achieved through its use remain legitimate because all students have the opportunity to benefit from the same technology.
I will now explain why these objections fail. To begin, let us dispose of the idea that equal access to AI renders grades attained through its use legitimate. Such an objection fails to address my concern. My umbrage is not with the notion that some students possess superior tools by which to complete their assignments; it is that the use of such tools defeats the purpose of the assignments. That some students may be better able to produce satisfactory material using AI does little to indicate the extent to which they comprehend the material they have submitted and less to indicate the extent to which they have developed their intellectual abilities.
Why do I insist that AI assisted assessments are unrepresentative of intellectual ability or development? Why am I unwilling to hold AI in the same regard as calculators and search engines?
The laziest kind of AI use hardly needs discussion here. When students simply submit the instructions they have been provided with to LLMs, presenting the subsequent output as their own work is blatant plagiarism. One hopes that this is not the kind of AI use that is increasingly tolerated. In any case, I am making the stronger claim that even when students rely on AI for developing their work rather than the outright production of it, the result is still unrepresentative of their intellectual ability.
When planning to write an essay, one virtually never begins with a fully articulated argument. While one might have a general idea of what they would like to establish and how they might go about it, the ways in which premises connect, the strengths and weaknesses of supporting arguments or even the cogency of one’s initial intuitions are not things that can be transcribed from conception. It is in the process of writing, line by line, word by word, that students must think deeply about the issue with which they are dealing. As one writes, they come upon unexpected difficulties, unanticipated connections and novel ideas. By carefully selecting one’s phrasing, refining and editing, a passive engagement with a topic transforms into a meaningful intellectual exercise.
It is imperative to recognize that vague prompts are not the same as developed ideas. Tempting as it may be to credit yourself with the ideas expressed in a guided response, if your assignment is to write an essay, writing an outline or a draft is merely brushing the surface. Organizing ideas into a coherent argument is essential to developing clear thinking and communication. If you allow AI to do the work of structuring, phrasing, or evaluating, you have missed central components of the assignment of essay writing. Thus, the notion that AI merely performs intellectually insignificant tasks is wholly mistaken.
In light of what I have established, there are two final concerns I wish to bring to my readers’ attention. First, students who fail to develop the ability to write and argue effectively will likewise be impaired in their ability to assess the quality of writing and arguments. Not only is this a major deficiency in intellectual development in and of itself, but further, it is a deficiency that is likely to compound. A student who is unable to recognize the superficiality of an LLM’s output is thereby more likely to rely upon it. Second, AI generates its output by drawing upon pre-established patterns in the available data; it is often sycophantic and is purposefully indecisive. As a result, using AI encourages students to reproduce conventional lines of thought in place of actively participating in inquiry. This not only robs students of their individuality but also gravely inhibits the ability of a university to cultivate free and original thought.
In summation, I have made the case that relying upon AI to complete written assessments is not simply dishonest; it is a disservice to the education of oneself and others. If students continue to rely increasingly upon AI for their written work, the university will have lost something far more valuable than reliable grades.