The Impact of LLMs on Learning and Education

22 min readJan 1, 2024

Like a lot of my peers, I’m neck-deep in AI these days and interact quite a bit with educational institutions.

I get asked quite a bit in my interactions with industry professionals, educators, educational institution administrators, policy makers and students about my opinion on the impact of AI on Learning and Education, specifically, Generative AI and even more specifically LLMs.

Oftentimes, queries by industry professionals come with unarticulated, preset positions on the topic influenced (probably) by professional affiliations. The intent in those cases seems to be to find commonality and confirmation of individual positions thus establishing a foundation for possible subsequent business engagement..

Some others are either mildly curious or just making conversation while looking for more important or interesting bits of information to emerge.

Then there are those that have the need to comprehend the impact of AI but still are scattered in their comprehension, and looking to be educated (usually by some typical work situation). Most of them employ the battle-tested technique of making a few self-aggrandizing statements on professional accomplishments and importance followed by a battery of questions (with a challenging edge to them) designed to extract information and educate themselves while, simultaneously, maintaining control of the conversation.

Finally, there is that very small percentage who (I feel) are sufficiently unbiased, transparent, knowledgeable and (surprisingly) interested in, my opinion, the rationale behind my opinion and the contextual elements of my rationale. Some of them even go as far as to discuss weaknesses in their own positions in the same way as mine and express empathy towards opposing viewpoints. A rare, near-extinct breed.

In the majority of these conversations, the analogy of the calculator and its impact on education crops up. Mostly to shore up arguments. Sometimes as a contrast agent.

Arguments run the gamut, consisting of assertions that range from the irrational to rational but flawed conjectures, to elaborate conjectures dressed up as proofs (mine typically fall into this last category — you will see soon :-) ).

Vis a vis calculators, time has shown us that our past fears (mostly in educational circles) about calculators threatening math education was, for the most part, unfounded. Calculators are ubiquitous today yet seem to have caused no perceptible harm to math aptitude and when students use them to augment their learning, they actually help. Papers have been published to this effect without causing anywhere near as much of a stir as the original scare (viz., A meta-analysis of the effects of calculators on students’ achievement, J. Math Education).

In retrospect, the calculator, like the early computer, was a technological bellwether. In some ways not too different from the current AI technology phenomenon i.e., Large Language Models (LLMs) currently making the same rounds in academic circles and raising similar outcries.

But there is a key difference this time though.

The scope of the threat posed to learning and public education by LLMs, is much broader, fundamental and too real for casual dismissal by analogy.

LLMs and Generative AI have a direct impact on any form of educational activity that involves reading, comprehending, writing and communicating (all universal learning primitives). They possess this uncanny ability to generate extremely human-like sequences of words based on its “comprehension” (transformation) of human-supplied prompts and its ability to write (generate) full responses. These models. uncomfortably expose to us a behaviorally similar, poorly understood, seemingly un-human, computational machine lurking within our heads. This uncanny-valley alienness has added to the already sizable pile of panic twigs. The upshot? The collective “surrender/fight/flight” toned responses.

More importantly though, the disruptive entry of LLM and Generative AI technology from the online consumer-technology domain into academic contexts has been heavily funded by corporate efforts with poorly concealed market domination ambitions. Consequently, this uninvited but seemingly erudite guest has triggered fear, panic, curiosity, optimism, excitement and every other emotion amongst the academic community and the students. Yet again, academicians and institutions are forced to deal with a disruptive technological phenomenon that academic research community germinated but didn’t plan for to be unleashed this way; that too at the worst possible time i.e., post pandemic amidst soft job markets, record enrollment declines in teacher certification and a global environment of chronic uncertainty.

Of the many questions posed to me, the recurring ones have been (usually of the form “In your opinion….” followed by):

Will these LLMs/Gen AI models have no effect on Education?
Will these LLMs/Gen AI models have just minimal but positive impact?
Are these LLMs/Gen AI models a leading indicator of the much needed transformation of education itself?
Will these LLMs/Gen AI models free teachers from having to teach content as a whole?
Are these LLMs/Gen AI models a threat to the modern public education system and cause devastating damages to student learning?
Do these LLMs/Gen AI models need to be constrained/controlled/subordinated and only then be used sparingly in pedagogical contexts?

Deliberately avoiding the convenient “it depends” response, my response (yet another conjecture) has been:

“As institutional learning has now been reframed and consequently perceived by industrialized society today and, the way LLMs & Gen AI (in the opaque, commercial and media attention grabbing manner) are being unleashed on society today by giant corporations, LLMs, in my opinion, are a net negative to students and to teachers as they stand and could accelerate the demise of the current public education system if educational institutions don’t comprehend the enormity of the situation and adapt quickly”.

This eye crossing, mouthful of an answer has received a variety of responses. Some agree, Many disagree. Others don’t understand. Most (students) don’t care!

Seeing that the answer consistently confused and raised so many questions, I realized that there was a need for definition, context setting, communication and hypothesizing to help comprehend the response.

And so here goes (for those interested!):

First, let’s consider LLMs.

Essentially, LLMs are models and, like all models, are a particular, transformation based representation (one of myriad possible representations) of the writings of the world that the model construction engines have been presented with. This applies to Gen AI models too.

This LLM view of the world takes the form of a supermassive corpus of structured, semi-structured and unstructured text. The complex functional form and associated coefficients of these LLM models then form the information basis for the AI algorithms which use them to act essentially, as probabilistic token prediction machines.

These machines predict most likely sequences of symbols given an input of transformed, prior sequences of symbols.

Within these algorithmic machines, likelihood is a computed statistic (to determine “most likely”). These computations are based on (yet again) mathematical transforms performed on the text documents/content posted publicly on the internet.

At this juncture, it is absolutely crucial for us to understand what is and is not included in the content the LLMs are presented with as they shape the model’s comprehension and representation of the world.

LLM input content includes all of the widely known web-sites on the internet — Wikipedia, Reddit, Newspapers, Blog sites — as well as obscure ones.

But it

does not include corporate intranets, private discord channels, slack conversations, private discussion forums, private mailing lists, the “dark” web and most of the world’s books.
does not include the vast amount of web content that’s written in languages that LLM companies chose to ignore.
does not include non-digital collections (e.g., the Library of Congress).
for the most part, does not include technologically inconvenient content like most of the audio content on the web, viz. podcasts and video as transcribing that volume of content, and doing it correctly, isn’t cost effective yet for the industrial providers.
does not include the “recent” web, as it takes enormous amounts of time and compute resources to process and train models on the entire public internet.

Therefore when statements like “LLMs ingest likely sequences of symbols” is made, the word “likely” in the LLM context can be expanded as “the collective publications of mostly English speaking adults writing publicly online in the past decade”. It does not mean “a fair, recent, sufficiently broad representative sampling of human knowledge, thought and dialogue around the world over time across the majority of cultures and languages”.

Second, consider “Learning”.

It is well understood (certainly in academia) that the term “learning” does not refer exclusively to standardized test outcomes, graduation rates, grades, college placement, or other measurements of academic learning.

But in conversations with some of the educators, educational institution leaders and administrators, it struck me as that was the assumption being made, sometimes explicitly but most times implicitly (often catching me by surprise).

Now, “learning” in general terms refers collectively to the set of activities that endow us with the skills to effectively comprehend, communicate, engage, act, navigate, appreciate and contribute to the world around us.

Certainly, one aspect of this learning journey includes acquiring the skills needed to perform well in the assessments that are conducted with the intention of measuring our comprehension of specific areas to compete in the job market. These measurements also play a role in defining subsequent steps in our learning journey. But they do not define learning.

The many dimensions of Learning

Every one instinctively knows and understands that, learning ultimately is much much more than those measured metrics. But in this metricized, monetized world we live in, we forget.

Learning is nuanced, multidimensional and simultaneously is about

Systematically improving the physical and mental skills that we innately possess as human beings to better comprehend us, others, other things around us, their abstractions, their similarities and their differences.
Better communicating ideas (ours and others) preferably in our own unique voices and styles to motivate, exhort, influence, change and help attain lasting outcomes
Diligently reasoning through our hypotheses by providing clearly stated assumptions, structurally sound and well grounded logic (as opposed to pattern recognition) to convince, educate, exhort and drive decision making.
Empathizing, comprehending and therefore “learning” that the keys to consistent success lie in mastering the comprehension of the diverse range of thoughts and emotions swirling within us and in others
Recognizing, comprehending, analyzing and making informed decisions despite all the uncertainty (aleatory and epistemic) innate to the world around us
Gaining the skills needed to be able to solve real world challenges and to achieve common, mutually beneficial goals
Gaining the skills needed to recognize, face, de-escalate and resolve conflict in its many forms
Recognizing and learning to hold steadfast belief in knowledge and in the power of humanity’s abilities to overcome adversity
Constantly striving, discovering, communicating and improving those, most accommodating and accepting, forward solution paths (despite inconvenient origination points) that have the capacity to reconcile conflicting human values and desires in any challenging situation
Understanding who we all truly are within ourselves, where our true emotions, intentions and aspirations lie, what we truly value, and how inevitably, like the universe, we change over our lifetime and therefore realizing that commonality we share with others and accepting the reality that others and things around us change too
Understanding the need and gaining the skills to form trusting relationships and maintain them despite differences, conflicts and the vagaries of life
Comprehending and gaining the skills to unflinchingly handle the one, most difficult constant in our world i.e., change.

This list goes on. Learning is human, complex, continuous and is fundamental to life itself.

Third, consider Schools

Now, this requires emphasis: Schools are not corporations. They may (unfortunately) resemble them today. But that was not the original intent.

In an increasingly knowledge dominated economy, companies have realized their increasing need for highly educated employees and have moved swiftly to incorporate their reframed version of schooling into their organizations and sought to shape schools in their mold.

But schools from time immemorial, were and still are places designed to be mini-representations (a safe microcosm) of an ideal society with built-in educative, administrative, regulative, social constructs all aimed at maximizing the extent, intensity and impact of the learning experience via immersion.

Learning in schools happens, to this day, primarily through trusted teacher-student, student-peer and student administration relationships with the curriculum of study used to set context and boundaries to guide pedagogy.

By definition, schooling includes the teaching of those topics and skills that are measurable via tests viz., reading, writing, mathematics, history, science, art etc. But these measurement tests were and are not the sole purpose of schooling.

Schools are also intended to awaken and develop in us an appreciation of the beauty (and flaws) of and, the fractal, ever growing, awe inspiring, daunting nature of knowledge. Schools also teach us that knowledge goes well beyond tests, books and the internet. True knowledge ultimately lies in comprehension.

Most importantly, schools, by design, teach us about the unavoidable struggle required to gain knowledge, and that struggle is an essential component of preparing ourselves to succeed in the real world. That struggle should tackled squarely, should not be feared or shirked. Students need to understand that failure, despite effort, does happen and is in no way a reflection or a judgment of an individual’s quality or ability to learn and to succeed. Quite the contrary.

Schools are, by design, set up to educate students (through structured as well as spontaneous interaction experiences) about the fact that people are very different and that the struggle to understand each other and to work together to achieve common goals is messy, complex and completely normal.

Schools are designed to teach us that, while learning self-protection skills are essential, real safety isn’t in subversion, weapons, surveillance, toxins, punitive measures and deterrents. Those elements exist (in unfortunate abundance in the real world) and may even be justifiable in extreme circumstances but true safety lies in relationships based on shared experiences, trust, transparency, tolerance, respect, affection and allowing ourselves and the people around us the freedom (and the consequent responsibility) to make choices and recognizing that mistakes are the real keys to learning.

Schools are set up to show us that egalitarian values are critical to the collective acquisition of knowledge and that, (truly paraphrasing Euclid) “there is no royal road to knowledge”. While wealth can certainly be deployed to enable education, wealth by itself cannot infuse knowledge or guarantee learning let alone unbiased learning.

Schools are set up to show us that education strongly influences art and artistic taste. Schools shape values by simultaneously teaching and allowing the student body to question the philosophical underpinnings of social values.

Schools have long recognized that the act of questioning the status quo is a long standing tradition, is healthy, is normal and needs to be encouraged despite being a messy process. It is a critical and required skill for real comprehension and therefore growth. Paradoxically questioning authority is a sign of trust and learning. So is allowing the questioning of authority,

Unfortunately, all of these are not simple concepts that can easily be standardized, packaged and measured in exams or in lesson plans. Nevertheless, schools, as institutions and environments, continue to this day to foster and shape student learning in all these areas with varying degrees of success under tough economic conditions.

About Teachers

Teachers are, to this day, expected to teach portions of that whole body of knowledge, except they are far more constrained in their choices of methods today than in the past.

Today, topics such as

institutional economics and corporate funding,
high levels of regulatory scrutiny
requirements to ensure student success in the high stakes tests whose results define monetary success for students in a mega-corporation dominated world

attempt to monopolize teachers’ attention and consequently attenuate their focus on teaching.

Thus today, schools as they have been reframed and therefore perceived in most countries, have been forced to be much less about the constructs described earlier and much more about the measurable elements like the tests, the grades, the credentials, the fundings and the donations received and the sizes of the salary offers made to graduating students.

It needs to be emphasized here that this is not because the educators want it this way but because of the pressures placed by

The general tax paying public who view higher education and most colleges (at least in America) as a bad investment despite more and more job requirements indicating the contrary.
Corporates and economic institutions (banks and lending institutions) who have reframed today’s schools mission in utilitarian terms to suit their goals and have been able to drive the educational agenda via economics

In response to all this, even those teachers passionate about educating and learning, have settled down to the grim reality that teaching today is first and foremost about ensuring that

Their students achieve high scores leading to high income offers
They maximize their own performance measurement metrics

All other aspects are “nice to haves”.

To complicate matters, unfortunately, the kind of learning that actually matters in the long run, is too complex a construct to easily metricize and define in monetary or utilitarian terms. Given the highly personalized nature of that kind of learning, it is also a very difficult model to scale.

Finally, teachers today are also acutely aware that they are not trusted to adhere to these reframed metrics of pedagogy for various reasons (philosophical and practical). Therefore their evaluations, compensations, resource access priority and extent of access to school controlled resources have been explicitly tied directly (and in many cases contractually specified) to ensure compliance to, and maximization of these metrics of learning.

Therefore, logically, teachers have responded by shifting their emphasis to these quantized indicators of learning instead of focusing on the process of learning itself.

This has paved the way to their own substitution with AI technology automatons, therefore hastening the crisis in the public educational system.

The Students

Student reactions too to this reframing and AI onslaught are similar to the teachers but a bit more varied.

The rare breed of students that are truly interested in learning and not too concerned about failure and the metrics/indicators of learning, are definitely not rewarded or valued for it by the current system, our educational institutions or their peers. These students are, in fact, quickly made to realize by academic administrative responses that they need to curb their extraneous learning & tinkering instincts (i.e., learning outside of the indicators of learning) to gain academic support, gain acceptance, to survive, to thrive in the educational system, the institutions that define the bulk of their student life and subsequently their professional life.

Students that aren’t interested in the kind of learning described earlier, realize that they too have to do whatever it takes to gain the freedom they crave to do the things they are interested in. This results in them pursuing any means required to traverse the shortest path to winning the exam game and to attain the freedom to do the things they value.

The students for whom this system, as it stands today, actually works are the ones who unquestioningly accept learning in its narrow, highly constrained form, or are just interested in meeting/exceeding others’ expectations of them or both. For these students, school is a system they exist in where they strive every day to regain control over their time and their economic situation by gaming assessments to maximize their test scores using methods that minimizes risk.

True Learning is the one legitimate strategy for “gaming” the system ( an inefficient, but safe and valuable one) but many alternate shorter paths exist that offer less resistance, less struggle, and more freedom but with varying levels of risk.

This is the overall situation into which LLMs have been casually lobbed in. And the student impact has been rapid, widespread, devastating and predictable.

Any student whose interests weren’t aligned with what was being measured already had the motivation to find the shortest path through assessments. Prior to the arrival of LLMs, that typically involved copying or finding the answer online or avoiding struggling through the challenge of learning by delegating or simply not doing the work.

After LLMs, the path is now clearer and simpler i.e., find and copy the prompt from someone, copy the answer. The only risk is that the teacher/institution could attempt to detect LLM-generated content. Unsurprisingly, AI and market demand have triggered a rush to fill that “white-space” and to present solutions to that problem for a price (https://www.stealthgpt.ai/). And so avoiding learning is what students do.

Sidestepping messy questions about ethics, integrity, and honesty as most schools today are no longer places that teach those concepts anymore (students are expected to know them), the reaction of students is logical if not justifiable.

The vast majority of disinterested students now have even greater incentives and ability to avoid learning in a society already exhorting them to avoid extraneous learning, and just to score well, graduate and qualify to step in line to claim the glittering economic lotteries awaiting the top tier performers (per metrics).

The upshot of all this is that, at best, using LLMs, students might learn some skills in transcription, editing. tool usage, coding (aka hacking) and the ability to mask their reliance on generated content.

But then there’s more.

Even for those students whose interests are aligned with school and learning, LLMs change the equation subtly and problematically. When faced with a seemingly intractable problem that the “aligned” student is unable to solve, he/she turns to ChatGPT just for guidance (say, like they would with a TA). But the student, unlike with a human TA, is presented with the solution and the explanation. The student now proceeds to comprehend the solution and the associated rationale and (say) solves the problem again on his/her own thus staying true to the task assigned. Rinse. Repeat. But the rub here is that comprehending an explanation of a solution is not the same skill as finding a solution. The student, unknowingly (and unannounced), has shifted the task set by the teacher from one of learning to solve a problem to one of reading, comprehending and recreating the solution to a problem. Good enough, right? Well, not really.

In the real world, if students only ever encountered problems that LLMs could answer, they would get the right answers. But that position essentially posits that because LLMs exist, humans knowing how to solve problems does not matter anymore! The reality is that these future professionals will encounter problems they haven’t encountered before and they may not even be able to describe them to an LLM, due to lack of exposure. Akin to the way that driver ability to navigate visually has declined as GPS has become ubiquitous.

But then there are even more subtle problems aside from not learning to solve problems. One of them is that LLMs have a skewed view of knowledge and are wrong quite often. Especially when prompted on uncommon problems or in an uncommon natural language. Students that aren’t fluent in English, or have creative teachers that design uncommon assessments, are more likely to get wrong answers, and yet LLMs will still happily generate a “likely” answer. Students, not knowing the subject, will have few skills to even detect these flaws, introducing noise and randomness into their comprehension, shifting learning from a process of responding to targeted feedback from a trained teacher, to a process of searching a space of probable answers that they are not yet capable of judging. Thus LLMs throw an inequitable monkey wrench into the already complex orchestration of self-regulated learning skills that so many students already struggle with. LLMs amplify these inequities by rewarding “normal” and punishing “different” while we all know that advanced education, innovation and progress is essentially “different”.

LLMs and Teachers

A key issue is that the teachers and institutional leaders and administrators themselves often do not know the subtle and often profound limitations of LLMs.

Teachers and education administrators and policy makers are, today, excitedly talking about the ability to have a free teaching assistant, the ability to quickly generate lesson plans and detect cheating with the right tools. But none of these things are factually correct and in-fact reveal more about those teachers and education administrators than, I suspect, intended.

LLMs don’t teach; they generate lesson plans that people have shared online, and not particularly good ones; and there is no reliable method to detect what was generated. While the industrial marketing machine has convinced educators and education leaders into thinking that LLMs are an intelligent education aid, factually, LLMs are just generating likely things that other educators have posted online, whether they are good or not.

This lack of LLM literacy, coupled with the hype, is leading educators to give up on the authenticity and the insight required to structure meaningful learning for their students such that it maximizes the learning experience.

This broader self-sabotaging trend amongst students, teachers, and policy makers adopting a tool that supplants the broad scope of human learning and ability with a narrow tool that generates a pastiche of whatever was posted online cannot be explained away as a flaw of the tool. It is an indicator of the paucity of a clear vision of education and consequently learning. This vision announces that “students and educators don’t need to authentically know things, or create things, or write, or think; we can comfortably delegate all of that to english-fluent adults who came before us, and then edit their work if it seems wrong or bad.”

This vision for human experience is not a comforting or a viable one.

Losing the ability to do manual arithmetic to calculators was one thing. It did not supplant our ability to do mathematics or the need to reason. What calculators did was narrow in scope. LLMs, however, have a much broader impact. They are supplanting the students’ ability to learn the skill of “learning to think”. They are short circuiting students’ opportunities to authentically write, synthesize, reason, and struggle by reinforcing a vision that equates learning with producing correct answers irrespective of the path. They, simultaneously, are short circuiting teachers’ attempts to craft authentic learning experiences for their students and to prepare them for what we know is an unpredictable future filled with as yet unseen challenges.

At its dystopian worst, what we will slowly and incrementally converging to, is a system of teachers (heavily aided by automated agents) who cannot do instructional design on topics never dealt with before (they don’t have to), students (heavily aided by automated agents) who cannot learn new things (they don’t have to), and a chronic dependency agents serving up questions and answers on an ever increasing body of remixed content combined with an ever shrinking body of authentic content published on the internet.

The future (imo) will be the search for those slivers of authentic content reminiscent of tiny evolutionary mutations.

A fair and skeptical question to ask here is “Won’t students and teachers ultimately reject these tools, because they just aren’t good at supporting learning and teaching?” That might be true if schools operated in alignment with the definition of learning and schools described before. But in countries globally, particularly with ones where income inequalities are stark, public schools are even more chronically underfunded than ever before, teachers and students are under even more pressure to maximize indicators of learning in order to retain access to basic resources like housing, health care, and safety via job offers.Educators in the current system of education, with its complex economic ties to corporations, have been forced to paint themselves into a corner. They just cannot afford to care too much about the innate learning and problem solving abilities of future generations. They necessarily have to acquiesce to the demand to focus on the metrics of learning in order to maximize return on educational investment pronto. This in turn means lower taxes and in turn hiring educators from locations where higher learning is still viewed as a good investment. This hiring will supplant the need to solve the really difficult and innovative problems that we have not faced before but will face us in the future given the unprecedented changes we see occurring around us.

In many of my interactions with teachers and education leaders, I see members of the staff are eagerly trying to integrate LLMs into learning, to help them overcome the severe resource scarcity issues they face and out of genuine desire to address educational inequities by giving every learner access to a personal teacher. Those are very valuable contributions and they need to be encouraged.

But there still is insufficient critical thinking on setting boundaries. It requires us to face the real “elephant in the room” questions i.e. what is our definition and understanding of learning? Does the current pervasive definition of learning as meeting or exceeding measurable metrics of learning for economic advantage need to be reworked in a manner that better accommodates the many required objectives for human learning to continue to be relevant in the future?

We get that the reality is that LLMs are here and are being deployed as a tool for learning and that they will get better and better. But learning to learn requires struggle. As a grizzled and failure familiar student, it is clear to me that learning is hard, takes focus, needs acceptance of difficult to receive feedback, is confusing, involves repetition, needs someone caring that you continue to do all of these things despite difficulties and realizing that despite all of that it is fun and rewarding experience. There’s nothing about a tool that promises to replace all of these things with answers that we can lightly edit that acknowledges this essential need. This is because the whole point of an LLM, when applied to learning, is to remove struggle. So, no matter how good LLMs are or become, they will always prioritize a future that replaces human learning with convenience (by design). Now there’s a conflict of interest.

If we care about people having knowledge and abilities — and we should, because abilities are the only things that keep us alive, not to mention the LLMs coming — then education must stay focused on them, and not let LLM building minorities in companies inadvertently subvert the purpose of educational institutions.

So What Now?

The questions I get asked at this stage are,

Enough with the Scaremongering! What should we do?
Should we ban LLMs from schools?
Should we ban them in general?
Should we teach teachers and students who to use them more responsibly?
Should we put guardrails on LLMs themselves, holding model providers accountable?

All of these different recommendations (posed as questions I might add) miss the point in my opinion: Schools, as they work today, have been working this way for too long. It’s been a while since most schools have actually incentivized learning and no use or disuse of LLMs can change this. LLMs are yet again harbingers.They are not the real issue. They have brought the real issues into sharp focus.

In my opinion, Schools need to go back to first principles definitions of teaching and learning. Schools need to reeducate themselves about the massive reframing of perceptions and expectations of learning that has happened and why that matters so much today. Schools (and humanity) need to realign with the original, much richer and complex definition of learning so that LLMs and all of the other technological distractions that have come to dominate attention become relevant only in ways we want them to be.

If schools were already in collapse because of their broken incentive mechanisms, and LLMs are now accelerating it, then it’s about time we took school reforms that have better incentive proposals seriously, before that collapse comes and not blame/fear LLMs. And reform is about clarity, vision, conviction, collective action and will.

Educational institutions certainly need to be much more cautious about uncritically accepting any world view where freely available tools comprehensively devalue the human ability to learn.

Educational institutions need to critically re-examine their role in ensuring humanity is retained in education, learning, schools, teaching and their relationship to automatons is humanly determined.

Technologies threatening to subvert the human capacity to learn will no doubt continue to emerge and require institutions to cope but ruthlessly centering on the truer, more expansive, inclusive and nuanced definition of learning and encouraging the building of relationships, trust, and individual student identities and interests can keep us on the right course.

Students will then want to build with each other, and with the adults who support them. This kind of liberatory vision of learning is the only one that prevents youth from passively accepting the world and its latest innovations as immutable fact and the inevitable future. But, this is not the kind of educational vision that is being encouraged sufficiently in colleges today. Students today are not interested in truly questioning the status quo or looking for truly meaningful alternatives as they are too concerned about economic rewards and reducing student loans. That’s what has been drilled into them as what learning is all about. This encourages an approach similar to the underlying philosophy of an LLM: the only things that can be said are that which has been said before by our collective definition of “normal,” and at the expense of anyone on or beyond the margins defined as normal.

So what, then, should a teacher do about LLMs?

Use them with extreme caution.

Educate students and administrators extensively about their flaws.

Demonstrate that they aren’t the intelligent form of life their creators (humans!) claim; that they are frequently wrong; that they will not save humanity, or bring forth eternal human prosperity.

Celebrate the humans involved in and human design elements intrinsic in LLMs.

Celebrate human learning and achievement by highlighting and validating that learning requires struggle but that can be playful and rewarding.

Reinforce the fact that LLMs, like most modern technologies, are great at making things easier but are deeply flawed by being incapable of comprehending and educating us about the rich capacity of humanity to create, surprise, commune, and love, and that we should not give up these things so willingly, lest we give up our humanity.

In my interactions with students and young professionals, I’ve found that while most ignore me in favor of their devices, some do respond at the mention of a radically new way of seeing the world with AI in it but where they are at the center of it. That gives me hope. Life will find a way.

The Impact of LLMs on Learning and Education

Written by keshava rangarajan