A fast moving target

August 18, 2024

There is a need to rethink and pre-empt the broad impact of AI on education

A fast moving target


O

n November 30, 2022, the world changed when OpenAI launched ChatGPT – a chatting app built around a class of artificial intelligence models called Generative Pre-Trained Transformers (or GPTs for short). Since the launch of ChatGPT – and the subsequent emergence of a plethora of other LLMs/ GPTs – we are seeing a re-birth of interest in AI and a revolution unfolding before our eyes. This is a revolution like no other, at least in the 21st Century, and is moving at a breathtaking pace, affecting a variety of fields, and entire lives as a result.

One of the areas where AI is likely to have the most significant and far reaching impact is education. The second most significant challenge (after the threat of creating intelligent machines that may bring about the end of humanity as we know it) is the challenge of mass and widespread unemployment and disruption to work. OpenAI estimates that LLMs/ GPTs and this class of AI models are likely to redefine and automate 10 percent of the work of 80 percent of the people and 50 percent of the work of another 19 percent (leaving just 1 percent most creative people, who may remain untouched by AI).

While these numbers may reflect an approximation for the developed world, I believe these significantly underestimate the impact on countries in the developing world, such as Pakistan. Here, the impact on a majority of low-skill knowledge workers – such as developers, designers, content writers and average professionals – will be much more severe. Will those losing work be able to transition to higher order work? Or will they fall into the abyss to become losers of the AI age? That remains to be seen. But it will largely depend upon how our educational system responds to AI.

The impact of AI on education can be understood in multiple ways and at multiple levels. At the most fundamental level, the use of AI by teachers and educators can improve productivity and free up their time from mundane tasks so that they can focus on more creative and substantive ones. Teachers can use AI to make lesson plans, generate problem-sets, check exams, write reports and communicate with parents and school administrators. Teachers can also use AI in the classroom as a tool and aid to substantiate learning, illustrate, demonstrate and simulate complex points, encourage classroom creativity and practice, etc. Teachers may also encourage students to use AI as a personal tutor. No teacher has the time to answer every question by every student. Used in a personalised tutor mode, the AI can be a great and patient teacher. Students may use the AI for conversations about problems and difficult topics to prepare prior to coming to a flipped class or go back after a class to clarify and build upon what they’ve learnt. All these have been hypothesised to be useful and recommended modes of teaching and learning with the AI.

Challenges arise when the AI becomes, not an aid but a wholesale replacement of teaching and learning activities. Should students be allowed to use the AI, solely, to solve homework problems or turn-in essays written by AI? How should one assess the learning and contribution of a student to a co-written essay or solution? Will this result in the kind of learning achieved without the use of AI? If not, will it be better or worse? A recent study, provocatively titled AI can harm learning, by professors at the University of Pennsylvania looked at math learning in 1,000 high school students in a Turkish school system found that students using AI performed better on an (AI-assisted) practice sessions but worse on the (unassisted) final test than those who did not use AI during practice and, actually, “struggled” to practice and learn.

As AI takes on the role of a teacher and tutor, it is important to rethink what we need to learn and how to learn with AI in schools. 

Large language models are now quite good at summarising texts, producing decent quality letters, emails and essays; even reasoning and solving simple (already solved) problems. Should students be allowed to use these language capabilities to turn in writing assignments, essays, and problem sets? If this were to become a norm in our education system, what would this do to our ability to write in the longer run? It is possible, like with the example of the pocket calculator, this skill may just vanish without anybody noticing. But writing is also linked to our ability to think and learning to write helps clarify our ability to think. In a world with ubiquitous AI where we might never have to write anything, will we also never require thinking? Will we have to develop other mechanisms, besides writing, to teach our children to think?

Questions like these are at the heart of the debate at the intersection of education and the AI. There are still no clear-cut answers.

The AI is here to stay. It will continue to make inroads in our professional lives. Our education system needs to incorporate and build around the emerging realities of the AI age. Today’s students (and tomorrow’s professionals) need to learn what will help them advance their careers in the not-so-distant a future. As AI takes the role of a teacher and tutor, it is becoming important to rethink what we need to learn and how to learn with AI in schools.

For almost over a year now, there has been a massive movement around the world aimed at rethinking education and the future of work in the era of AI. Scientists, scholars, educationists, policymakers and practitioners are thinking hard and running experiments to understand the impact of AI on the learning process and how education must evolve to account for easy and wide access to these powerful AI tools. These questions must be answered not after-the-fact but through prior experimentation and rigorous scientific evidence.

Many countries, school districts and schools – including in Pakistan – have announced changes in policies and educational practices and have provided tools and trained their teachers to teach AI effectively. The approach to this need not be outright rejection or extreme caution. A carefully planned roadmap will be needed at the school as well as the national level to properly understand AI’s impact on the learning process and devise strategies that maximise benefit and minimise harm.

Schools, school leaders and teachers don’t have to be followers. They can be leaders and partners in this important exercise as they create the localised evidence that helps adjust their own adoption and AI strategies.

The learning and future of our children are too important to leave to chance or be left without access to the best learning and professional tools available. This includes AI.


The writer is a former member of the Planning Commission. Currently, he is the managing partner of INNOVentures Global (Pvt) Ltd and co-founder of PakGPT, a platform to promote AI skills and education. He can be reached at athar.osama@gmail.com

A fast moving target