The Resistance to AI in Education isn’t Really about Learning
When future historians write about the education industry in the early 21st century, they may debate which event caused more lasting disruption: the COVID-19 pandemic or the emergence of generative AI. The one point upon which a consensus might emerge is that both caught the educational establishment entirely unprepared — which is surprising.
For while the COVID outbreak may be seen as a “black swan” occurrence, the arrival of artificial intelligence (AI) had been anticipated for decades. Indeed, thirteen years before the release of ChatGPT 3.5, Apple had introduced “Siri,” its voice activated personal assistant which relies on machine learning technology. Given the ubiquity and popularity of Apple products, this impressive but milder form of AI should have been the perfect herald for what was to follow. Despite this, by the fall of 2023, educators were astonished and deeply worried by the capabilities of generative AI, viewing it as a threat to academic integrity and the teaching profession.
This reaction underscores a deeper issue: the resistance to AI in education is not truly about learning. It reflects a reluctance to re-evaluate the traditional roles of educators and to embrace the opportunities AI offers to enhance the learning experience.
The AI in Education Counter-Reformation
Articles and essays raising the alarm have appeared, such as this calling for active resistance to AI in education, while another piece by noted writing instructor John Warner included his insistence that he cannot use as AI as a tool because “I cannot outsource giving writing feedback to something that cannot read.” (The link to this article puts it in even more panicked language — “embracing AI means abandoning learning”.) Warner elaborates on his point:
“In fact, reading their work is not sufficient by itself. I must also have a conversation with them about my impressions of what they’ve written. This is the work.”
This is the work. One can almost hear the gravitas in the phrase. Imagine a Latin intonation! (Hoc est opus.)
I find such arguments unpersuasive. Warner presents AI for learning as a zero-sum situation where the instructor has no true role. What, one wonders, prevents Warner from reading and commenting on student work and using AI to augment his instructional commentary? And why doesn’t Warner make reference to the perennial dilemma of writing instruction— the inability of human teachers to generate sufficiently detailed responses to student writing in a timely manner?
As a college writing instructor of more than 30 years, I know from personal experience that evaluating student writing is an enormously time-consuming enterprise, which invariably leads to students receiving feedback several days after they have finished composing. Research on instructional feedback has consistently highlighted the critical importance of promptly returning instructors’ comments to students to maximize the impact on their learning.
For decades, I have tried to resolve this problem by using a variety of technologies (including macros with pre-written feedback and voice-based recorded comments). In 2018, I began exploring artificial intelligence as a teaching tool because I realized that the dilemma of providing instructional feedback could not be resolved by human effort alone.
When ChatGPT emerged (specifically 4.0), I quickly discovered that generative AI had advanced sufficiently to provide sophisticated analysis of human text (as long as it was provided with a clear rubric and specific prompts). I began using AI to provide rapid feedback to my writing students. And I informed them about my use of AI and gave them the option to opt out.
The use of AI immediately increased both the speed and the amount of quality feedback to each student. This is no small issue given that the most impactful service a writing instructor can provide is commentary on an evolving draft. For the first time, I received emails from students thanking me for my feedback.
And yet, when I first shared my experience with other college instructors, the response was not entirely encouraging. There were objections to allowing AI access to student writing because the drafts would be permanently stored in the AI database, potentially exploiting the students’ intellectual property. One colleague even told me what I had done was unethical.
I had two issues with these criticisms. First, writing instructors have been uploading student work into plagiarism detection databases like Turnitin for decades, and this has never been a major point of contention for most faculty or students. Second, classifying student assignments as intellectual property is a bit of a stretch, as intellectual property refers to work with inherent value. In contrast, student work is intended as a means to an end — acquiring knowledge — rather than an end in itself. (Which is why students dispose of assignments after receiving a grade.)
The more I heard such objections repeated, the more I was reminded of the psychotherapist who, after listening to a patient talk incessantly about a seemingly unimportant subject, feels compelled to say, “I don’t think what we are talking about is what we are really talking about.”
While there is much thoughtful concern about AI being expressed by critics in education, such as potential privacy data issues, these objections are expressed openly. It is my contention that much opposition to AI for learning is rooted in unspoken fears and attachment to flawed pedagogical models.
In order to thrive in the learning ecosystem that will evolve in the Age of AI, the teaching profession needs to do some difficult but essential re-evaluation of their role, in order to better understand where they can provide the best value to learners. This requires confronting some comforting myths and uncomfortable truths.
Problem #1: The Belief in Instructor Essentialism
The main argument against AI in education is that meaningful learning requires a skilled and caring human intermediary between the learner and the subject. I call this “instructor essentialism.”
Instructor Essentialism is the belief that almost all learners require the assistance of a human expert to guide them in order to achieve deep learning and mastery of any subject. It follows then that students would be fundamentally disadvantaged if instructional support tasks were taken out of the hands of human instructors and given over to AI.
A recent edition of the respected educational periodical, The Hechinger Report, explored the question of how the work of supporting student learning should be properly divided between humans and AI. Not surprisingly, the Report came down decisively in favor of human-dominated instruction, repeatedly emphasizing the current limitations of AI. It cited approvingly comments by Satya Nitta, a former IBM technologist turned edutech entrepreneur who founded a company called Merlyn Mind, which marketed an AI assistant tool for classroom instruction.
“Nitta said there’s something ‘deeply profound’ about human communication that allows flesh-and-blood teachers to quickly spot and address things like confusion and flagging interest in real time.”
Given that Nitta is selling a product which requires teacher adoption, it is not terribly surprising that he endorses the essentialist view (and is rewarded with a nice plug for his software in the article).
The doctrine of instructor essentialism is a core belief (really the core belief) of teachers everywhere. Culturally, it is reinforced by a considerable number of influential stories and films which emphasize the solitary, heroic instructor (Goodbye Mr. Chips, To Sir with Love, Dead Poets Society, etc) who serves as a pedagogical savior to students.
This romanticized view of teacher/student relationship belies the reality. The typical teacher/student relationship is mutually respectful but fundamentally transactional. After the primary school period, students work under the supervision of a variety of ever-changing instructors, each of whom has to attend to the needs of at least 100 students. The sort of meaningful mentor/mentee relationship celebrated in the movies is the rare exception, not the rule.
More importantly, the true essential ingredient of all effective learning experiences are the learner’s own actions, not the instructors. As Herb Simon, cognitive science pioneer and Nobel Prize winner, once observed,
“Learning results from what the student does and thinks and only from what the student does and thinks. The teacher can advance learning only by influencing what the student does to learn.”
In other words, at best teachers can add value to the learning experience, but are not an absolute necessity. Like AI, teachers may augment and support what the learner does, but in the end, learning depends upon the learner’s own efforts and innate ability.
Learning Versus Schooling
Instructor essentialism emerged in part due to the error of conflating schooling with learning. Attending school is a near universal experience. One of the fundamental functions of schools is to supervise young children and adolescents while their parents are at work. For this reason, teachers in schools are indispensable, but it has little to do with learning. (One of my graduate school professors pointed that the two primary functions of schools are education and childcare — and that they are consistently successful at only one of those things.)
By the time young people are ready for college, they have spent most of their lives with teachers and find it difficult to imagine education without an adult standing at the front of the classroom. This is similar to how our ancestors struggled to envision land-based travel without a horse leading the vehicle.
The fallacy that an activity requires an intermediary between the subject and the object is not restricted to education. A parallel can be drawn between the current developments in AI and education and the investment industry’s experience in the 1970s with the introduction of index fund investing. This financial innovation largely eliminated the need for ordinary investors to rely on professional money managers, (who received large fees regardless of whether investors made or lost money).
At the time, the money management industry mocked the value of index funds, and insisted that serious investing required the expertise of seasoned professionals. Yet, in just a few short decades, index funds revolutionized finance, bringing benefits to millions of ordinary people who had never previously experienced them. Index and other mutual funds are now the dominant model for personal investment.
Comparing teachers to money managers might be dismissed as unreasonable, even outrageous, by those who regard educators as occupying a special moral category where their actions are seen as more innately altruistic than those of finance professionals. However, education, like money, is a valued commodity. Teachers are society’s primary learning managers. Any innovation which reduces or even, in some situations, eliminates the need for them would be perceived as a threat to their job security rather than be celebrated as a boon for their clients, the students.
As someone who has spent his entire career in education, I view teachers not as cardboard saints, but as people just as complex in their personal motivations as every other group of workers. While teachers are genuinely attracted to the public service aspect of education, they are often equally drawn to the perceived job security and the reassuring, unchanging routine it offers.
Problem #2: The Closed World of Academic Culture
In addition, many teachers have spent little time working in non-academic professions. This is especially true for college instructors, who must devote five to seven years to graduate education before obtaining their first full-time position, and thus have little time to explore careers outside academia. This common lack of non-academic work experience heightens the anxiety that educators feel when contemplating the potential impact of generative AI on their work lives.
Moreover, teachers have far less experience than other professions in having to adjust to significant work changes. The special protections afforded by academia (tenure, academic freedom) have often allowed teachers to successfully resist pressures to transform their practices and thus have, to a large degree, insulated them from the frequent changes that have been required of other professions. This is reflected in classroom design, which remains essentially the same as it was three centuries ago. An instructor from the 18th century could walk into a 21st century classroom and quickly grasp where he or she is supposed to stand and how to use the whiteboard.
The future of teaching in the age of AI is not as bleak as many educators imagine. To return to my psychotherapy allusion, once the real issues can be openly acknowledged and fears addressed, healing and growth can commence.
The AI revolution will disrupt every industry it encounters. That the education sector was struck first is both its misfortune — and its opportunity. AI may prove to be the catalyst for the sort of evolution that the teaching profession has needed for a long time.
A Way Forward
Faculty need to critically reexamine many of their cherished assumptions about their roles and begin to recognize the possibilities inherent for education in the world of AI. A few things to consider.
- Generative AI as a force multiplier for formative assessment: As I observed earlier, AIs guided by clear rubrics can solve a major issue that has plagued formal learning — the problem of feedback. Frequent AI-augmented feedback, along with written and oral feedback from an experienced teacher, can help us achieve unprecedented learning outcomes.
- AI can create a wide variety of uniquely powerful learning tools: AI provides valuable feedback to writing students (example here). Before AI, creating effective learning simulations was a time-consuming and costly process. Now, AI reduces both the time and expense, enabling students to practice their skills frequently in a controlled environment (example here). Additionally, AI-powered retrieval tools can automatically generate emails to test and reinforce learners’ knowledge after a course has ended (example here). AI also assists teachers with assessments by applying institutional rubrics to program-level outcomes (example here).
- AI can accelerate the adoption of the science of learning: For sixty years, educational researchers have made great strides towards increasing our knowledge of how learning works. Yet, much of what has been learned has never made it into common classroom practices. Using AI to design curriculum aligned with the science of learning can accelerate the adoption of evidence-based learning and thus increase the professionalism of the teaching field. (Phillipa Hardman has been a pioneer in this work.)
- More opportunities for social learning: The learning experiences that we can create with AI afford teachers and students more content to discuss in small group, class-wide, and one-to-one conversations. The social element of learning, which is the uniquely human aspect of teaching, can be greatly increased.
AI doesn’t eliminate teaching but offers teachers powerful instructional strategies, reducing reliance on the long-dominant lecture-based model. AI won’t eliminate the need for teachers — but it can bring about the end of a model of teaching which is antiquated and unsupported by what we have learned over the past 50 years from research into the science of learning. AI does not represent a dark age for teaching — but potentially a golden age of learning.
