ChatGPT, Hard Thought and the Advent of Flipped Learning

10.01.23. There was something pleasingly binary about the date on which the Alleyn’s Senior Leadership Team and I first engaged with the latest AI phenomenon, ChatGPT. (The GPT stands for the infinitely forgettable: Generative Pre-Trained Transformer which is a reference to the technology underpinning it.) The reason we, along with everyone else, were talking about this new chatbot, is because Dr.Greetham had posted a link to it on the Alleyn’s staff Team and advised us to try it. You won’t be disappointed, he promised. And he was right. It seemed to offer something completely different in AI support. Whilst AI help with learning has been available in education for a good number of years, the seismic, game-changing thing about this software is both its plausibility; (it responds, seamlessly, in natural language, if necessary mirroring the voice of the questioner in the answers it provides) and its scope; it seems able to answer pretty much any query you can throw at it. (When I invited it to write a start of term letter for parents, it really made quite a good fist of it, give or take a few Americanisms and the odd cliché. I promise that I did not use it, but how amazing that it was even a possibility!) 

So, 10.01.23 felt like something of a moment as we started to reflect in detail on the possibilities of this software and its implications for learning, for assessment, for the whole business of education in schools. In fact, it is being discussed everywhere and it turns out that we were relatively late to the party. ChatGPT was launched in November 2022 by OpenAI, an organisation founded in 2015 (by Elon Musk amongst others), in order to develop AI for the benefit of humanity. Five days after it had been launched, it had garnered more than a million users. Apparently, it took Meta ten months and Netflix three years, to do the same!¹

And no wonder. Since ChatGPT hit the airwaves (I suspect that metaphor is quaintly anachronistic), people have been posting in astonishment and disbelief, about the range and scale of its offerings. Whether it’s a Board report, a complete essay, an application letter, a lesson plan, the answers to your maths homework, some coding or a pastiche of your favourite Beyoncé tune, ChatGPT will confidently and courteously oblige. As one American College student who asked it to resolve his research paper difficulties pointed out in a December tweet: “we are witnessing the death of the college essay in realtime.”² 

And perhaps that is true. Our Head of English at Alleyn’s ran a blind marking test with the English Department. One of the essays submitted was provided by ChatGPT – it scored 34 marks (out of a possible 40). So, it can produce Grade 8 (or A* in old money) standard work, at least up to GCSE level. And because all you have to do is simply ask it, the interface is incredibly straightforward and no longer requires key words or pre-planning, its temptations, for those looking for swift, easy answers, are very clear to see.

But of course, great education, or even, moderately good education, is not, and has never been, about the speedy provision of easy solutions. Our job as educators is to teach our pupils how to think. How to use their own minds, their developing skills and talents, their growing knowledge, their ability to research, their judgement and capacity for discrimination, to arrive at possible solutions to the various problems presented to them. The fact that some of those answers are wrong is precisely the point. School is where we learn what to do and how to do it. It’s also where we learn what not to do. What doesn’t work. How to get things wrong and how to deal with that. We all know how important it is to learn to fail. To be comfortable with failing…  as Samuel Beckett so elegantly pointed out: “Ever tried? Ever failed? No matter. Try again. Fail again. Fail better!”. One of my biggest concerns about AI software, no matter how sophisticated, is that it might insulate our pupils from the realities of that. (I’ll never forget getting a D in my Maths homework when I was 11. I couldn’t quite believe that I had got such a grade. Or even, that such a grade existed, in a world where everyone around me was getting A and, at worst, A minus. How could I have got a D? I remember being so ashamed and anxious and upset about it. I didn’t tell my parents, or my friends or anyone… I sat on it for a good couple of days as I tried to process it. But that moment, tough though it seemed at the time in my first year of a highly selective, academically driven school, was, in fact, seminal. The world didn’t end. I worked out what I’d done wrong. More to the point, I worked out what to do next. I spoke to my teacher. I did, indeed, do better next time. And I found a way of recognising and being comfortable with what I was good at and what I would always need to work at…I learned to accept that. To love it, actually, as a reflection of what it is to be alive and to be young and to be a learning, imperfect, interesting human!) Thinking, learning, being educated, is not just about repetition and parroting things off and gathering information from massive data sets. It’s about emotion and character and social interaction and creativity and all the things that a chatbot has not (yet) found a way to replicate.  

That’s education – and ChatGPT doesn’t have any truck with it. So, it may have been “trained” through a machine learning technique called RLHF (Reinforcement Learning from Human Feedback), and through accessing a massive amount of data online, but it can’t actually think, nor does it know when, or if, it has got things wrong. (There was a lovely moment in a test a journalist ran in Berlin, where its answer to: “which mammal lays the biggest egg?” was: “the elephant”.) 

So, where does it leave us? The wonderful Alleyn’s students and staff, as we engage with this new technology at the start of a new year? Well, it’s exciting for sure. There may be limitations, but there are opportunities here too. We will be able to save time through this technology – in the administrative processes of our school, if not necessarily, in all the educational work. Equally, confronting something that feels seminal, that feels like a paradigm shift, is always thrilling. For us, it will involve careful reflection about what we should be asking our pupils to do in school and in class and what they can do at home. I suspect we will see more “flipped learning” in the coming months and years, as diagnostic assessment (what have you understood at what level), moves more frequently into the classroom and preparation for the lesson (read this, assimilate, come with questions) moves into homework. Flipped learning has been something of a holy grail for schools for a while, so this would be an exciting development. I also think that whilst we can let the chatbot do the heavy lifting on the low-level, information gathering, summarising work, there will always be a premium on higher order, reflective and hard thought. If the exam boards have to find ways round AI generated coursework responses, their solution will need to be more rigorous expectations around original thought and detailed, rich analysis of problems. That feels inspiring. 

And for all of us, ChatGPT has given us a sense of what is to come and some of the big questions that might need to be asked as a result of that. What is learning? What does it mean to be educated? Perhaps, above all, what does it mean to be human? ChatGPT’s answer to that (see below), doesn’t even come close! And there is great comfort in that! 

Chat GPT’s response to the question: “what does it mean to be human?” 

Being human refers to the characteristics and qualities that define the species Homo sapiens. These include the ability to think, feel, and communicate, as well as having a complex nervous system and the capacity for self-awareness and introspection. Additionally, humans have a unique capacity for cultural and societal development, and the ability to create and use tools. The concept of what it means to be human is also deeply rooted in philosophical, religious, and cultural beliefs, and can be interpreted in many different ways. 

Jane Lunnon

Listen again to Jane talk more about ChatGPT and flipped learning to Nick Ferrari on LBC Radio (30.01.23) here.

And read her interview with The Times here. A login or paywall may be in operation.