Interesting to think about training AI to teach when I think we are still figuring out how to better train teachers to teach - or at least to teach better (it's probably one of the most complex jobs you could conceive of)
Completely agree with your perspective in this post and the previous. I am an AI optimist and avid user in general but there is a reason why this report lists 'educational' as one of the 'least exposed' categories :)
I suggest spending some time chatting with ChatCPT about ideas about description versus explanation in the context of teaching and learning. It can describe, but cannot explain well. It has no ability to ask the next question in a conversation about a concept. It can only respond in the frame presented by the user, and so cannot lead the user beyond where they are. So it is had to see how new learning can occur. And that is in an ideal one on one situation, let alone the complex classroom environment Dan described.
You conjured such a great vignette of the quantum mechanics of learning in situ - there is so much dark matter! When I visit classrooms as a school board member, I am always so impressed with the advanced cognitive load that educators are balancing as they manage both the micro learning and the macro management of the classroom!!!
I appreciate your recent series on AI, Dan. I'll speak for myself here, but I always felt there was something spiritual in holding space for and with students in a learning environment. So many of the memories students and I shared were simply because we invested in seeing each other. This connection seemed to open the door for the math (or whatever else I was trying to teach). I can see genAI helping with the proliferation of _knowledge_ but I just don't see (yet?) how it could do the work of connection.
Another insightful post - thinking too how AI can't "see" the brilliance in wrong answers, insight from body language and non-verbal cues. Side note: I am currently listening to "The Right Kind of Wrong" by Amy Edmondson and feel too many people are looking to AI in ed as supporting the 'right kind of rIght'.
I appreciate your posts and perspectives, especially regarding AI lately. I agree, I'm not sure I can imagine a machine doing this, but I also admit that I don't see enough teachers attending to students and their thinking and nuance the way you describe here. And I'm not blaming them - the systems doesn't put enough value there. Perhaps AI could help with SOME of that information, but I think teachers will need to still use the information an act. The question for me is how can we support teaches more to be able to interact with students and their thinking the way you describe? And can AI help with that? I'm not too hopeful, given that's not usually where we invest much of collective energies (prior to AI).
I would also add that in addition to the petabytes of information teachers process during a typical class period, they also need to process information about **themselves** as well. There's two metacognitive elements/questions teachers are constantly asking: 1) "What about this student's response/reaction/energy/etc. do I need to understand to clarify my own questions/pedagogy about the content?" and 2) "What am I feeling today? How will that affect the classroom community? How can I best serve my students given my own energy/experiences/etc.?"
It's probably tempting to say that AI doesn't have to worry about question 2, but I think that would ignore the importance of community- and relationship-building that are essential for any classroom. As a teacher, there were days when my own life events/moods/etc. (be they good or bad) helped motivate class.
https://youtu.be/bEJ0_TVXh-I. I'm involved as Alumni Representative (advisor) for the math club at a local college, being a '14 alum w a math AA, and a professor shared this video. My reply comment, is, Exactly! 100%
I for one am glad your mammalian ancestors hid from predators in the savannah so that you could write a Substack post about noticing a student struggle to calculate tax on a sandwich.
But on the failure of imagination bit, Justin Reich at MIT has started a very interesting project to collect a massive amount of data on tutoring interactions (he's focusing there on the presumption it'll be marginally easier to capture than whole-class teaching moves). Although there's a danger of reductionism, I am intrigued by the possibility of capturing "worked examples" of effective teaching taking place. Might be worth giving him a shout if you haven't lately.
" My mammalian ancestors hid from predators in the savannah so I could notice a student struggle to calculate tax on a sandwich." all time quote
Interesting to think about training AI to teach when I think we are still figuring out how to better train teachers to teach - or at least to teach better (it's probably one of the most complex jobs you could conceive of)
Completely agree with your perspective in this post and the previous. I am an AI optimist and avid user in general but there is a reason why this report lists 'educational' as one of the 'least exposed' categories :)
https://www3.weforum.org/docs/WEF_Jobs_of_Tomorrow_Generative_AI_2023.pdf
I suggest spending some time chatting with ChatCPT about ideas about description versus explanation in the context of teaching and learning. It can describe, but cannot explain well. It has no ability to ask the next question in a conversation about a concept. It can only respond in the frame presented by the user, and so cannot lead the user beyond where they are. So it is had to see how new learning can occur. And that is in an ideal one on one situation, let alone the complex classroom environment Dan described.
You conjured such a great vignette of the quantum mechanics of learning in situ - there is so much dark matter! When I visit classrooms as a school board member, I am always so impressed with the advanced cognitive load that educators are balancing as they manage both the micro learning and the macro management of the classroom!!!
I appreciate your recent series on AI, Dan. I'll speak for myself here, but I always felt there was something spiritual in holding space for and with students in a learning environment. So many of the memories students and I shared were simply because we invested in seeing each other. This connection seemed to open the door for the math (or whatever else I was trying to teach). I can see genAI helping with the proliferation of _knowledge_ but I just don't see (yet?) how it could do the work of connection.
Another insightful post - thinking too how AI can't "see" the brilliance in wrong answers, insight from body language and non-verbal cues. Side note: I am currently listening to "The Right Kind of Wrong" by Amy Edmondson and feel too many people are looking to AI in ed as supporting the 'right kind of rIght'.
But why not just train teachers to do that work?
I appreciate your posts and perspectives, especially regarding AI lately. I agree, I'm not sure I can imagine a machine doing this, but I also admit that I don't see enough teachers attending to students and their thinking and nuance the way you describe here. And I'm not blaming them - the systems doesn't put enough value there. Perhaps AI could help with SOME of that information, but I think teachers will need to still use the information an act. The question for me is how can we support teaches more to be able to interact with students and their thinking the way you describe? And can AI help with that? I'm not too hopeful, given that's not usually where we invest much of collective energies (prior to AI).
Is it your perspective that there isn’t an opportunity for a virtuous relationship between good teachers and personalized programs / AIs?
You are quickly becoming my favourite writer in substack!!
I am confused. Who/what calculates sales tax for individual items? Isn't this an academic problem, broken cash register or not?
I would also add that in addition to the petabytes of information teachers process during a typical class period, they also need to process information about **themselves** as well. There's two metacognitive elements/questions teachers are constantly asking: 1) "What about this student's response/reaction/energy/etc. do I need to understand to clarify my own questions/pedagogy about the content?" and 2) "What am I feeling today? How will that affect the classroom community? How can I best serve my students given my own energy/experiences/etc.?"
It's probably tempting to say that AI doesn't have to worry about question 2, but I think that would ignore the importance of community- and relationship-building that are essential for any classroom. As a teacher, there were days when my own life events/moods/etc. (be they good or bad) helped motivate class.
The Alpha private school in Austin, Texas is trialling replacing teachers with AI and "guides/coaches": https://www.youtube.com/watch?v=QFczkA88bcs
https://youtu.be/bEJ0_TVXh-I. I'm involved as Alumni Representative (advisor) for the math club at a local college, being a '14 alum w a math AA, and a professor shared this video. My reply comment, is, Exactly! 100%
I for one am glad your mammalian ancestors hid from predators in the savannah so that you could write a Substack post about noticing a student struggle to calculate tax on a sandwich.
But on the failure of imagination bit, Justin Reich at MIT has started a very interesting project to collect a massive amount of data on tutoring interactions (he's focusing there on the presumption it'll be marginally easier to capture than whole-class teaching moves). Although there's a danger of reductionism, I am intrigued by the possibility of capturing "worked examples" of effective teaching taking place. Might be worth giving him a shout if you haven't lately.