Love reading your threads on AI and math education. Not sure what kind of feedback or visibility you are getting from what you send out, but I always find your ideas insightful and on point. Just wanted to give you that feedback for your great thinking and work here. Thank you.
This was a really interesting perspective. And for me personally, the timing was amazing, as I have been listening to a podcast this past week on the teaching of reading comprehension, and utilizing an approach which emphasizes the importance of an asset-based platform, as opposed to a skill-set approach (which is the norm in America), though they used a different term for it.
Yeah, the more I learn about AI, the more I think we need humans teaching.
I feel that teaching in the 3-minute video isn't meant to be an open-ended exploration or following student's lead wherever that takes us. It's a solid example of direct teaching done well. As such, it's programmable.
Here's where my imagination took me:
- Imagine an object-oriented microworld with visuals and a very readable code, like Scratch with restricted commands, a Geogebra app with only a few tools enabled, or any number of Logo toys like LightBot.
- Now imagine the program given scripts (recurring good prompts) from experienced teachers, such as, "Please talk or text to me, explaining what to do, like you would a child." What Katrine Bryan did in the video isn't a one-off invention. I've heard similar scripts before and used them myself whenever students got stuck on a visual task because switching representations from visuals to words is quite helpful. Moreover, we know the major representations for each given school math task, so we can invite people to try and switch between words, graphs, code, diagrams, sketches, and so on. I am not diminishing the pedagogical content knowledge required for this in the classroom, yet it's also reusable.
- The program hears/reads the word "square" and asks, "What size, in particular?" It draws a square of that size in the middle of the screen, with the line of code for that displayed. What if it doesn't hear the word "square", or the wait time is over some reasonable variable before it hears anything? Maybe the bot can invite the student to give it the name of a shape to start with something.
- The student can then tweak the code the bot made for size, angle, and position (the three variables that define squares on screens). The bot could prompt that, too.
- The key here is that we are in a MICROworld, so we only have a few objects and behaviors: squares, positions in space, and whether the square is blank or split into units.
- This scenario is like 1980s microworlds integrated with a current chat AI. I don't see anything a competent programmer couldn't implement in weeks (famous last words!). If we only had an hour, we could write a few scripts on index cards and prototype this scenario "unplugged" with some volunteer students.
Thanks for sharing that clip. The teacher was doing a good job bringing forth the student’s latent knowledge. Out of curiosity, how did the video camera filming the class work? Was the camera being controlled remotely over the internet or by someone in person?
Agree, Dan. Hard for adults too. Certainly nothing AI can help with. But coaches, some teachers and parents teach students their often metacognitive assets. Most mental health treatment improvements are based on teaching how to use assets to self regulate. My question is, if a student learns from a teacher, coach what some of their assets are, what happens when they inform LMMs of their assets and then asks a cognitive content question about a problem.
I’ve been thinking about what to teach students about how they can use AI . Teaching students to write prompts that generate responses that actually answer questions they have.
Do some experiments loading a student assets into the prompt framing the question about the problem. Express the assets the way a student might.
This led me to wonder if students could learn metacognitive skills for finding and expressing what they know. For now, I avoid visual problems.
What are your thoughts on equipping students with AI as a tool? Parallel to how Desmos puts the student…not the tool… in charge.
Hey Phil! My thought here is that one of the core gifts of a skilled teacher is their ability to see the assets that students don't even recognize in themselves, assets which the student often reveals non-verbally. It's a very different scenario where the student has to be conscious of their assets, confident enough to claim them, and then eloquent enough to express them into a chatbot.
I quote it too often, but I think of Angelou's adage about people will forget what you say and what you do but not how you made them feel. If feedback and assistance come from a genuine place, a place of recognition, our students will see that and be able to move forward with that support. While AI can give many a helpful tip and streamline processes, it can not provide the authentic care that goes behind it. You're right: we can build on student assets to tackle problems, and it is not an insult to AI or visionaries to suggest there are some things it can't do well. Thanks again for the perspective!
“What ideas do these students have?” I would like to hang onto the ambiguity: is this a question of idea inventory or is the asset the interest/curiosity they have?
Love reading your threads on AI and math education. Not sure what kind of feedback or visibility you are getting from what you send out, but I always find your ideas insightful and on point. Just wanted to give you that feedback for your great thinking and work here. Thank you.
This was a really interesting perspective. And for me personally, the timing was amazing, as I have been listening to a podcast this past week on the teaching of reading comprehension, and utilizing an approach which emphasizes the importance of an asset-based platform, as opposed to a skill-set approach (which is the norm in America), though they used a different term for it.
Yeah, the more I learn about AI, the more I think we need humans teaching.
I feel that teaching in the 3-minute video isn't meant to be an open-ended exploration or following student's lead wherever that takes us. It's a solid example of direct teaching done well. As such, it's programmable.
Here's where my imagination took me:
- Imagine an object-oriented microworld with visuals and a very readable code, like Scratch with restricted commands, a Geogebra app with only a few tools enabled, or any number of Logo toys like LightBot.
- Now imagine the program given scripts (recurring good prompts) from experienced teachers, such as, "Please talk or text to me, explaining what to do, like you would a child." What Katrine Bryan did in the video isn't a one-off invention. I've heard similar scripts before and used them myself whenever students got stuck on a visual task because switching representations from visuals to words is quite helpful. Moreover, we know the major representations for each given school math task, so we can invite people to try and switch between words, graphs, code, diagrams, sketches, and so on. I am not diminishing the pedagogical content knowledge required for this in the classroom, yet it's also reusable.
- The program hears/reads the word "square" and asks, "What size, in particular?" It draws a square of that size in the middle of the screen, with the line of code for that displayed. What if it doesn't hear the word "square", or the wait time is over some reasonable variable before it hears anything? Maybe the bot can invite the student to give it the name of a shape to start with something.
- The student can then tweak the code the bot made for size, angle, and position (the three variables that define squares on screens). The bot could prompt that, too.
- The key here is that we are in a MICROworld, so we only have a few objects and behaviors: squares, positions in space, and whether the square is blank or split into units.
- This scenario is like 1980s microworlds integrated with a current chat AI. I don't see anything a competent programmer couldn't implement in weeks (famous last words!). If we only had an hour, we could write a few scripts on index cards and prototype this scenario "unplugged" with some volunteer students.
Thanks for sharing that clip. The teacher was doing a good job bringing forth the student’s latent knowledge. Out of curiosity, how did the video camera filming the class work? Was the camera being controlled remotely over the internet or by someone in person?
It's called a "Swivl." Cool device.
Cool, thanks
You have the right to be insufferable, you are doing gods work, insuffer away
Agree, Dan. Hard for adults too. Certainly nothing AI can help with. But coaches, some teachers and parents teach students their often metacognitive assets. Most mental health treatment improvements are based on teaching how to use assets to self regulate. My question is, if a student learns from a teacher, coach what some of their assets are, what happens when they inform LMMs of their assets and then asks a cognitive content question about a problem.
I’ve been thinking about what to teach students about how they can use AI . Teaching students to write prompts that generate responses that actually answer questions they have.
Do some experiments loading a student assets into the prompt framing the question about the problem. Express the assets the way a student might.
This led me to wonder if students could learn metacognitive skills for finding and expressing what they know. For now, I avoid visual problems.
What are your thoughts on equipping students with AI as a tool? Parallel to how Desmos puts the student…not the tool… in charge.
Hey Phil! My thought here is that one of the core gifts of a skilled teacher is their ability to see the assets that students don't even recognize in themselves, assets which the student often reveals non-verbally. It's a very different scenario where the student has to be conscious of their assets, confident enough to claim them, and then eloquent enough to express them into a chatbot.
I quote it too often, but I think of Angelou's adage about people will forget what you say and what you do but not how you made them feel. If feedback and assistance come from a genuine place, a place of recognition, our students will see that and be able to move forward with that support. While AI can give many a helpful tip and streamline processes, it can not provide the authentic care that goes behind it. You're right: we can build on student assets to tackle problems, and it is not an insult to AI or visionaries to suggest there are some things it can't do well. Thanks again for the perspective!
Oh so proud of my cousin! =)
“What ideas do these students have?” I would like to hang onto the ambiguity: is this a question of idea inventory or is the asset the interest/curiosity they have?
Great commentary. Good linkage too.