AI and Education
AI could impact education as profoundly as the printing press or the internet, but first we need to embrace the Popper–Deutsch view of the theory of mind.
“Seeing is not believing”
Learning = Discovery!
Every time we learn something, we have discovered it for ourselves. That “eureka” or “click” moment when things suddenly make sense.
That last sentence is a loaded one for philosopher Karl Popper. It encapsulates his theory of knowledge: conjecture and refutation. That click in our mind is a guess, a conjecture, a theory which is suddenly put to the test in our mind - a process of refutation. We examine its implications and how it fits with everything else we know. As soon as it passes that criticism, “it all makes sense.”
Therefore, education and teaching should primarily be about ideas and their implications, not just facts. It’s one thing to know the names of all the species, it's something entirely different to understand evolution.
During learning, different people form different theories in their minds. As a teacher, you're constantly trying to guess what those theories are, and offer suggestions that could challenge them.
Recently, I was teaching a friend how to swim. I could sense that he was scared, even though the water was only shoulder-deep. So I said, “Let me show you how to recover your standing position, no matter how you fall.” For a while, all we did was fall in random directions and practise standing up safely. That helped him a lot and it helped me focus on the rest of the lesson.
The first role of a teacher or an AI is to guess what the misconceptions are.
"Autocorrect gets me."
Even without AGIs, current AIs can play a crucial role in understanding misconceptions in our theories.
Think about spelling autocorrect on our phones. More often than not, it replaces a misspelled word with the right one. If we had to constantly correct the autocorrect, we wouldn’t use it. It has gotten so good. A friend once told me, “It’s as if it gets me.”
Autocorrect today is a sophisticated technology it’s not just the closest word match, it also takes into consideration the words before, etc. This is a predictive model. Without worrying about AGI, current AIs can be thought of as very, very highly sophisticated predictive models.
This complexity has led to some interesting properties. These days, when I have code that needs bug fixing, I just paste all of it and prompt in ChatGPT: “Fix it”! It “gets” what I meant, and more often than not, fixes the errors.
This is more likely complex pattern matching to similar errors than an AGI. But that is already very helpful.
Because we can use this technology to pattern-match and identify misconceptions.
“I liked the book better”
Writing has played a huge role in education allowing us to copy and transmit information reliably across time and space.
The advent of photography, motion pictures, and even virtual reality helped enhance learning, but none were as revolutionary as text. The internet is perhaps the other technology that had a similarly transformative impact on learning.
All of these technologies helped in copying information reliably and increasing its reach. But at the end of the day, learning happens in the mind.
Talk to fans of any fiction series, and most will say, “I liked the book better,” even when there are multi-million dollar movie adaptations of the same story.
Even today, if someone wants to master a topic, textbooks are still the best material available. Many great online courses exist but they also include "reading materials."
At the end of the day, individuals learned on their own, with little interaction or feedback (checking the answers at the back of the textbook was the best feedback most of us got). For most things, we sought out other humans like teachers for feedback.
AI has, for the first time has the huge potential to provide feedback on our understanding. While each of us have different misconceptions, there are common misconceptions in our ideas when we are learning, and this technology can be used even through pattern matching to help address most of them.
“You’re still here, you must be interested.”
While that’s great, there’s an even more important question we need a better understanding of:
Where does the interest in learning something come from in the first place? Why do some people find a topic fascinating while others don’t?
Once someone is deeply interested, they’ve reached escape velocity in their learning journey. They’ll find their own resources and people to help them understand.
As a teacher, I believe that’s my primary role: to spark that interest, while assisting in the learning journey. The best way I know to do that is by trying to show how beautiful these topics are. Learning them is like watching a great movie (or reading a great book).
Most importantly, this calls for a deeper shift in how we think about the mind and learning itself. Without the right theory, we risk falling into the “bucket theory” of mind the mistaken idea that learning is just about pouring more facts into someone's head.
David Deutsch’s idea of universal explainers means that everyone can understand and learn anything. There is nothing about physics that prevents anyone from being able to grasp it.