View from the classroom: Writer Saikat Majumdar on teaching narrative intelligence in the age of AI

The obsolescence of certain abilities due to new technology is as old as human civilisation. Socrates had famously warned people against the (then) new technology of writing, as it would make people forget. As indeed it has: documentary archives have, to a great extent, replaced the work of human memory. From there, it was but a short leap to calculators doing large but basic calculations and Grammarly or ProWriting Aid to check the grammatical mechanics so that the human mind can focus on the more advanced and imaginative challenges of mathematics and writing. We now have ChatGPT Atlas, the AI-driven internet browser with advanced memory and digital agent mode that can remember, customise and adapt to your needs and habits. Even something as easy as surfing the internet can now be done with the assistance of AI. Such outsourcing of existing human skills and decision-making to technology is the process known as cognitive offloading, and it has both pros and cons.

Finding ‘sophisticated challenges’

To make the best use of cognitive offloading, we must ensure that the human mind, now freed of certain tasks that can be outsourced to technology, can be accordingly directed to these more sophisticated challenges. But to do so, educational curricula and methods need to be updated periodically. Australian academics Vitomir Kovanovic and Rebecca Marrone have pointed out that back in the 1970s, when calculators became available, their impact was regulated by making mathematics examinations much harder. So, while students could complete basic calculations on the calculator, the time and energy freed up would be used to solve more sophisticated challenges. Not enough has been done yet to keep up with the new resources, such as ChatGPT or other AI tools, now freely available to any student with access to the internet, even though some educators (such as Ethan Mollick at Wharton) have personally started incorporating AI in teaching and assignments.

In the absence of new kinds of AI-incorporated assignments, students are likely to simply outsource critical engagement to AI, which, Kovanovic and Marrone point out, leads to what experts have called “metacognitive laziness”. The long-term implication of this metacognitive laziness is the devaluation of life itself. Looking ahead to a future where AI models grade their own textual output, the writer and University of Virginia professor Piers Gelly sees the end point of this value system: “that school is reducible to workplace preparedness; that work is reducible to the pursuit of maximum throughput at minimum cost; that there is no difference between school and work; that value is measurable in all contexts; that time is money; that anyone who says otherwise is a scammer or a sucker.”

In order to ensure that our education makes us unique and irreplaceable for as long as possible, we need a sobering look at the skills in which AI is already outperforming us or will do so in the near future. It is clear that when it comes to consuming and storing data, and conjuring them up quickly for use, human beings do not possess even a small fraction of what AI can already do. But creating narrative from data? That’s another matter altogether. Arguments, pitches, plans, proposals – these are all forms of narrative. Even one’s self-presentation is a narrative act. As information becomes freely and widely available, it is not the gathering of it, but the narrativisation of what truly matters that is going to be the defining human contribution for quite a while to come.

Mustafa Suleyman gives the example of Large Language Models (LLMs), which include ChatGPT. “Today’s LLMs,” he writes, “are trained on trillions of words. Imagine digesting Wikipedia wholesale, consuming all the subtitles and comments on YouTube, reading millions of legal contracts, tens of millions of emails, and hundreds and thousands of books.” To say nothing of blogs, artworks, music, Reddit posts, Flickr images and the entire gamut of the internet. Next to that, he points out, the average American reads a million words a year, probably retaining only a fraction of it, which is an infinitesimal part of what these LLMs consume in a single month-long training run.

Creating language

But as someone who teaches writing, I know deep inside that the consumption of language is only half the story of writing in a creative and imaginative way. The other part is about – you guessed it – creating … and creating something unexpected. And how does one create language that is imaginative, memorable and original? One piece of advice I always give my students is to create the simultaneous reality of the alien and the familiar. Good writing, as with any good art, shocks and comforts at the same time, and its language does the same, making familiar words utterly unfamiliar. The same word – used at home and in a pub banter and in a business transaction and in a poem – has multiple avatars, one shockingly different from the other, yet indistinguishable at the same time.

In the corporate world, while kinds of materials such as routine memos, presentations, brochures and correspondences are easily generated by AI, a more specialised material, such as a leader’s speech, or a company’s vision plan, requires the intervention of a human being with a unique subjectivity. We still need the weight of human experience to do this, and the idiosyncratic, unexpected workings of a human mind. No matter how many instances they swallow, LLMs, which lack subjective consciousness, still don’t know how to simulate this personal feeling, and hence their language use, while impressive in its range and weight, remains predictable. But for how much longer?

The important thing to remember is that the work of AI, for the foreseeable future, will only be as good as the prompts it receives. This is already evident in people’s use of ChatGPT today, as well as other chatbots such as Grok, DeepSeek or Amethyst – including students’ efforts to extract papers out of them. Still, at a rudimentary stage, the writing that comes out of ChatGPT is not striking enough, but importantly, its quality radically depends on the questions or prompts that are fed into it. Intelligent use of ChatGPT to produce writing is increasingly becoming essential, much the way the use of calculators became essential to the learning of mathematics. I incorporated the mandatory use of a generative AI Chatbot in one of my essay assignments last semester, with striking results that made me reflect more on the pros and cons of AI use in the classroom.

Prompt-engineering, therefore, is a wholly human function, and the more specific and innovative the prompts are, the better the results produced by AI. “One of the hottest jobs today,” writes Salman Khan, the founder of Khan Academy, “is being a prompt writer or prompt engineer.” The interchangeability of the terms “writer” and “engineer” in Khan’s statement, while worth a chuckle, indicates the way traditional disciplinary and professional functions are melting in this new reality. And the commercial applicability of these roles is as significant as their links to older professions. Khan goes on to say that open-minded and imaginative copywriters can easily transition into prompt-writers and create copy and campaign on a scale and quality unimaginable before, given the massive data available to chatbots, far beyond the capacity of any human being or even a team.

Coworking with AI

This is where an education that is as humane as it is human is vitally important, as only such an education can, in turn, shape requests and questions fed into AI chatbots that have intellectual, artistic and linguistic qualities deep enough for the AI to produce work of high enough quality. To get the most out of AI, human beings need to remain efficient, innovative and ethical co-workers, since AI will continue to respond to the behaviour it witnesses and receives from its human counterpart. A very simple example offered by the Wharton School professor Ethan Mollick is a study of a range of prompts where Google’s most advanced AI model responded best to a prompt that began with “Take a deep breath and work on this problem step by step!” Clearly, AI can’t breathe, relax or panic, and yet they responded to the humanity of this prompt. Our humanity at its best, the most sensitive, intelligent and kind, is our best investment in getting AI to perform at its best.

Co-working with AI is the inevitable future of our labour in most spheres of life and production, and co-working with AI in our best capacity is not only desirable but, as we will soon see, essential for our continuity and survival. The difficult news is that as we work more and more with AI, we will be slowly relinquishing elements of our mental muscles – intellectual, entrepreneurial, administrative, of all kinds – as AI gives us better and better results than what we could have achieved on our own. The crux of our aptitude will eventually depend on how well we can interact with AI, and how well we can feed it the best possible instructions to maximise results.

The important thing is to move out of the mindset that AI is something to be “used” or “deployed”, terms that have worked for our relationship with technology so far. The best way to relate to AI is to “collaborate” with it, as is already evident with the most successful examples of writing and editing with well-thought-out prompts fed into ChatGPT. The fact is that generative AI is much more than a technology, far more even than other general-purpose technologies such as the wheel, fire, electricity or the steam engine. It is capable of self-learning and continuous improvement, initially following human training and data-feeding, and then eventually on its own. As it learns, it will eventually take on more and more and more of human labour, of all kinds, from the massively physical to the most intricate of abstract thought, and as it does, it will gradually render all those aptitudes obsolete in us, like unused muscles and organs. Now we can even create our own digital agents and outsource many of our routine tasks and decisions to them. Even if we resist this in small ways, in the long run, and in the big picture, this will be inevitable. To stay relevant, human beings will simply have to imagine knowledge and aptitude in new ways, which will also enter into relationships with the perpetually expanding intelligence of the machines.

Saikat Majumdar’s writings on higher education include the books College: Pathways of Possibility (2018) and the forthcoming Open Intelligence: Education Between Art and Artificial (2026).