Great post! Thanks for sharing the paper on "warm demanders", looking forward to reading it.
I'm just a homeschooler, not a teacher, so very limited sample size. But I've lived that "I can't do it!" experience so many times. Sometimes it's true, and you need to adjust the level of difficulty. Other times you just need give a break. Or a snack. And every now and then, it's a stern "of course you can do it, get yourself together!" "warm demand". The trick is in the student/teacher relationship that enables the right choice.
I still believe that AI *can* be a contributor to improved learning outcomes. But it's not at the expense of this core relationship. Rather, it's the relationship that needs to be centered and empowered.
> Rather, it's the relationship that needs to be centered and empowered.
I think this is absolutely right, both relationally and in a technical sense. For most kids, the relationship is the shipping network for learning. It's the highways.
We have people right now trying to create an alternate shipping network, something using drones maybe, and it's a very interesting, big bet. But we're continually watching those drones reach their limits, touch down on the side of the road, and hitch a ride on the highway. That's why I'm interested in improving the highway.
Marcus Luther of The Broken Copier wrote a fantastic piece about this recently.
He stated that marking writing has taken up so much of his time over the years.
And that time builds trust, strengthens relationships and helps teachers get to know their students. It creates a bond. It can be a reason some children go to school at all.
The comment about what students do to avoid learning was an aha moment. That’s something I’ve rarely seen in education writing and is likely a significantly underrated component of student “underperformance”.
Thinking Classroom is an excellent book on the topic. And it's constructive, instead of lamenting that students won't engage, it explores lots of useful tweaks that get MORE students to engage. Which is all I want, I'm not looking for some General Theory of Disengagement, give me something that increases engagement by 1% and I'll say "thanks!" and use it and then go looking for the next 1%. I've been doing this for a long time, it all adds up.
In discussions about public schools, I’ve often made the point that we must recognize the difference between students who “can’t” and those who “won’t”. More as a thought experiment than genuine policy, I’d like to limit access to public schools to just the willing. Kids of any ability that want to learn are welcome and will be accommodated to the greatest extent possible. Those that won’t, that are unwilling to put forth effort, get to go to some sort of basic daycare instead. As we learned during the pandemic, parents were valuing the daycare aspect of school almost more highly than the educational component. The experiment also recognizes that “teaching”, as commonly use, doesn’t exist. What we actually do is facilitate learning. Learning is done for oneself - it can’t be done to you.
Since you’d have the “daycare” alternative, mandating school attendance wouldn’t be necessary. Beyond that, teachers can usually distinguish students who are putting forth effort from those who aren’t.
While I admit that students who REALLY WANT TO LEAVE should be given some other opportunity, most resistance to deeper learning is not a character flaw.
What you describe sounds like deeming only those who are skilled at complying with teachers to be worthy of staying in school. (Or, I can easily imagine folks turning it into that; I spent some years in public schools, too...)
Some students believe that they are just too stupid and so they work harder at pretending they understand because they're afraid you'll figure out how stupid they are and kick them out. Others have never experienced an "aha!" moment in math. It's very, very worth exploring *why* the students are avoiding the "struggle" part of learning.
Student materials, especially tech-oriented, often instill patterns of "just answer the problem, it's not about understanding.
I think it’s very important to structure a learning environment that meets each individual student where they are and facilitates them progressing at their own best pace.
"exploring *why* the students are avoiding the "struggle" part of learning"
Aren't they trained to? Here's a math problem. It can be identified as a certain type of math problem, and once we've made that identification, it is solved using the proper procedure for math problems of that type. If you don't remember what type of math problem this is, or you do remember that but don't remember the proper procedure, you need more practice.
Anyway, once you get past "identify problem type and correct procedure", all should go smoothly, just follow the steps til you get to a correct answer (don't forget to draw a circle around it.)
Another fantastic example of how relationships are at the core of teaching and learning. Tools can be helpful, innovations can help, but the heart of growing is understanding.
I know it's Mathworlds around here but let's remember that in April 2023, one Bill Gates predicted that within 18 months -- by October of this year -- we'd have AI tutors that were better than humans at teaching kids to read.
We do not have AI tutors that are better than humans at teaching kids to read.
I haven't gone super hard on Bill there because I don't think that's what he predicted. His actually line was: "If you just took the next 18 months, the AIs will come in as a teacher’s aide and give feedback on writing."
That said, the transcript of that talk that he previously posted at Gates Notes has been removed—straight 404'd—along with several of the predictions I dinged him for a few months back.
I'd love to think we helped drive the 404ing. Gates was quoted as saying this: "The AI’s will get to that ability, to be as good a tutor as any human ever could," which hasn't happened in 18 months, isn't going to happen in the next 18 months, and I think is highly unlikely to happen in the next 18 *years*.
Quick question for you Dan: Agreed that student's often need the "warm demander" you're referring to — but what's stopping students to go to ChatGPT the moment they hit a wall with an AI tool that actually makes them stop and think? Won't tools need to consider this trade-off (helpfulness vs strictness) if they want to compete for the engagement of students when tools like ChatGPT are so available?
> but what's stopping students to go to ChatGPT the moment they hit a wall with an AI tool that actually makes them stop and think?
I think the answer is "nothing." I mean—sure, web filters and what-not. But you're right.
Thankfully, this is not a problem I feel compelled to solve. 95% of students need human relationship to mediate their learning. I am focused on strengthening that relationship. With posts like this, I'm trying to nudge the bar we're measuring AI tutors against into a reasonable position.
"95% of students need human relationship to mediate their learning"
Count me in that group. Some day, I'm going to take piano lessons (I'm 63, so I should probably get started on that sometime soon, shouldn't I?) And when I do, I know I'll need a human teacher even though they are probably lots of "learn piano online" apps out there. I know I'll need a human teacher because it's the fear of disappointing that teacher in our weekly lesson that would get me to actually practice.
A zillion years ago I reviewd the Carnegie "new" math software (still gets lots of visits on my blog). After two mistakes it just told you to ask your teacher, or it gave you the answer. Yea, right. (OK, and they had word problems with women riding in the Tour de France at 12 mph...)
*Three* zillion years ago (1998?) I started wondering why computer math programs never identified the most common mistakes and addressed them instead of just saiying "Wrong!" When somebody gives the GCF instead of the LCD... show the difference!
"He wanted me to give him the answer. That definitely isn’t the move, but what is?"
IDK if this is rhetorical or not, but I, too, have a 7 year old son at home, and I would have had him get out my measuring tape and measure out 50 inches and then 1 inch. Then I would ask, "Now, what do you notice about your number line?"
What’s wrong with his strategy? It’s an efficient use of the number line to show his thinking, whether he actually used it to solve the problem or figured it out in his head and just used the number line to show his process. What were you hoping he’d do?
The reasoning is correct, but as a human tutor facing students who repeatedly say "I don't know," if there are better coping methods, AI might also be able to imitate them.
The reason most AI tutors are difficult to use is that the designers did not consider the relationship between educators and learners, but instead treated AI as a question-answering machine.
I think that instead of providing answers in segments, it might be more helpful to list some examples from life to help students make associations. AI should be good at this approach.
It seems to me the problem is that math has still not moved very far from three ideas: the idea that there is only one answer to to each problem. All the AI stuff seems predicated on that. The idea that there is such a thing as grade level mathematics. Finally, the idea the learning of math is somehow linear... that you can't add 2 digit numbers until you can add i digit numbers nor can you learn multiplication before 3rd grade and so on.
All of this is hogwash as far as I am concerned, and it is the basis for why we are so terrible at teaching math, especially in the early grades. We begin to identify children as deficient at math as early as Kindergarten and we wonder where math anxiety comes from. There is so much effort to remediate before children have had adequate time to even learn anything. We have lists and lists that say children need to know this or that and if they don't they are somehow deficient.
Once I had the joy of having my students for 3 yrs.(2nd-4th grade). Math time was problem based with lots of objects available to use to picture the children's thinking. We also had morning calendar which was a time to play with numbers and patterns and also a time figure out how to describe each child's mathematical thinking. Children learned from each other as much or even more in some cases than from me. I never said we can't talk about numbers below zero because we are not in 5th grade or that we couldn't attempt to use those numbers in equations because we were not in 6th grade. We formulated problems that were interesting to us and THEN went about how we might solve them. We verbally explained our thinking, and when we could, we wrote about our solutions. As we did that we exposed everyone in the class to new ideas and widened our understanding of how numbers worked. My 4th graders were ready for many concepts contained in 6th grade math at the end of the year. And they all loved math which wasn't limited to a "math time" but was rather simply part of their day.
We are so lost as a nation concerning what teaching and learning means. The international test showed that despite all the remediation at 4th grade we are losing the middle and the bottom is more bottom than ever. It also showed that adults 16-60 are getting dumber and dumber, not just in math but in literacy and problem solving too. This is why so many were willing to hand our nation over to DT. Let him solve the problems, figure things out, it is all too much.
I teach an online Stats class, and all my students have to do a semester project, with multiple checkpoints. Hardly anyone gets it right on the first try, so I suppose I could grade these according to some rubric and say, "You met 5 of the 6 criteria, 83%". But since the pandemic I haven't been doing that. Instead I say, "You're missing this one item, please add it in so I can give you a full-credit grade." Or "Your conclusion still needs some work, please rewrite it and try to work in the phrase "not sufficient evidence to conclude." As a consequence of this, nearly everyone gets a perfect score, but hardly anyone gets their project done by the due date (it helps that I locate all this in a discussion forum, if I had to do all this back-and-forth by email I'd go crazy.)
Anyway, I mention this because we often tend to equate "demanding" with "strict", as in, "This was due last week, I'm taking points off for late work." And my experience is that many students like this "strict" approach because it lets them off the hook, they take their C and are happy to get it. I wouldn't say I'm a "warm demander", I'd say I'm an extremely annoying demander, because I won't let you go til you get it right. But being demanding in this way has meant being less demanding in other ways, it's meant having a policy where the only due date that really matters is the day I'm required to hand in grades (a thing I don't tell my students explicitly, but eventually they get the idea.)
Great post! Thanks for sharing the paper on "warm demanders", looking forward to reading it.
I'm just a homeschooler, not a teacher, so very limited sample size. But I've lived that "I can't do it!" experience so many times. Sometimes it's true, and you need to adjust the level of difficulty. Other times you just need give a break. Or a snack. And every now and then, it's a stern "of course you can do it, get yourself together!" "warm demand". The trick is in the student/teacher relationship that enables the right choice.
I still believe that AI *can* be a contributor to improved learning outcomes. But it's not at the expense of this core relationship. Rather, it's the relationship that needs to be centered and empowered.
> Rather, it's the relationship that needs to be centered and empowered.
I think this is absolutely right, both relationally and in a technical sense. For most kids, the relationship is the shipping network for learning. It's the highways.
We have people right now trying to create an alternate shipping network, something using drones maybe, and it's a very interesting, big bet. But we're continually watching those drones reach their limits, touch down on the side of the road, and hitch a ride on the highway. That's why I'm interested in improving the highway.
Marcus Luther of The Broken Copier wrote a fantastic piece about this recently.
He stated that marking writing has taken up so much of his time over the years.
And that time builds trust, strengthens relationships and helps teachers get to know their students. It creates a bond. It can be a reason some children go to school at all.
AI will never, ever provide that.
The comment about what students do to avoid learning was an aha moment. That’s something I’ve rarely seen in education writing and is likely a significantly underrated component of student “underperformance”.
Thinking Classroom is an excellent book on the topic. And it's constructive, instead of lamenting that students won't engage, it explores lots of useful tweaks that get MORE students to engage. Which is all I want, I'm not looking for some General Theory of Disengagement, give me something that increases engagement by 1% and I'll say "thanks!" and use it and then go looking for the next 1%. I've been doing this for a long time, it all adds up.
Yes. This should be discussed. I had a nifty online discussion w/ adult ed folks about students who resist conceptual learning.
In discussions about public schools, I’ve often made the point that we must recognize the difference between students who “can’t” and those who “won’t”. More as a thought experiment than genuine policy, I’d like to limit access to public schools to just the willing. Kids of any ability that want to learn are welcome and will be accommodated to the greatest extent possible. Those that won’t, that are unwilling to put forth effort, get to go to some sort of basic daycare instead. As we learned during the pandemic, parents were valuing the daycare aspect of school almost more highly than the educational component. The experiment also recognizes that “teaching”, as commonly use, doesn’t exist. What we actually do is facilitate learning. Learning is done for oneself - it can’t be done to you.
How do you intend to determine who can't vs. who won't? Is there a poll? A blood test?
Since you’d have the “daycare” alternative, mandating school attendance wouldn’t be necessary. Beyond that, teachers can usually distinguish students who are putting forth effort from those who aren’t.
Well, I think we can all rest assured that race or gender or socioeconomic status will not find a way to creep into the decision-making process.
While I admit that students who REALLY WANT TO LEAVE should be given some other opportunity, most resistance to deeper learning is not a character flaw.
What you describe sounds like deeming only those who are skilled at complying with teachers to be worthy of staying in school. (Or, I can easily imagine folks turning it into that; I spent some years in public schools, too...)
Some students believe that they are just too stupid and so they work harder at pretending they understand because they're afraid you'll figure out how stupid they are and kick them out. Others have never experienced an "aha!" moment in math. It's very, very worth exploring *why* the students are avoiding the "struggle" part of learning.
Student materials, especially tech-oriented, often instill patterns of "just answer the problem, it's not about understanding.
(tho' also, there's a lot of compliance that is avoiding learning... it's complicated!
I think it’s very important to structure a learning environment that meets each individual student where they are and facilitates them progressing at their own best pace.
"exploring *why* the students are avoiding the "struggle" part of learning"
Aren't they trained to? Here's a math problem. It can be identified as a certain type of math problem, and once we've made that identification, it is solved using the proper procedure for math problems of that type. If you don't remember what type of math problem this is, or you do remember that but don't remember the proper procedure, you need more practice.
Anyway, once you get past "identify problem type and correct procedure", all should go smoothly, just follow the steps til you get to a correct answer (don't forget to draw a circle around it.)
Another fantastic example of how relationships are at the core of teaching and learning. Tools can be helpful, innovations can help, but the heart of growing is understanding.
I know it's Mathworlds around here but let's remember that in April 2023, one Bill Gates predicted that within 18 months -- by October of this year -- we'd have AI tutors that were better than humans at teaching kids to read.
We do not have AI tutors that are better than humans at teaching kids to read.
I haven't gone super hard on Bill there because I don't think that's what he predicted. His actually line was: "If you just took the next 18 months, the AIs will come in as a teacher’s aide and give feedback on writing."
That said, the transcript of that talk that he previously posted at Gates Notes has been removed—straight 404'd—along with several of the predictions I dinged him for a few months back.
https://www.gatesnotes.com/ASU-and-GSV
https://danmeyer.substack.com/p/bill-gates-tells-oprah-about-edtechs
What a time to be alive.
I'd love to think we helped drive the 404ing. Gates was quoted as saying this: "The AI’s will get to that ability, to be as good a tutor as any human ever could," which hasn't happened in 18 months, isn't going to happen in the next 18 months, and I think is highly unlikely to happen in the next 18 *years*.
https://www.cnbc.com/2023/04/22/bill-gates-ai-chatbots-will-teach-kids-how-to-read-within-18-months.html
Yeah, he just didn't say they'd get to that ability in 18 months. Please don't make me defend billionaires, Ben!
Quick question for you Dan: Agreed that student's often need the "warm demander" you're referring to — but what's stopping students to go to ChatGPT the moment they hit a wall with an AI tool that actually makes them stop and think? Won't tools need to consider this trade-off (helpfulness vs strictness) if they want to compete for the engagement of students when tools like ChatGPT are so available?
> but what's stopping students to go to ChatGPT the moment they hit a wall with an AI tool that actually makes them stop and think?
I think the answer is "nothing." I mean—sure, web filters and what-not. But you're right.
Thankfully, this is not a problem I feel compelled to solve. 95% of students need human relationship to mediate their learning. I am focused on strengthening that relationship. With posts like this, I'm trying to nudge the bar we're measuring AI tutors against into a reasonable position.
"95% of students need human relationship to mediate their learning"
Count me in that group. Some day, I'm going to take piano lessons (I'm 63, so I should probably get started on that sometime soon, shouldn't I?) And when I do, I know I'll need a human teacher even though they are probably lots of "learn piano online" apps out there. I know I'll need a human teacher because it's the fear of disappointing that teacher in our weekly lesson that would get me to actually practice.
My money is definitely on your predictions!!
I cannot get anyone to take my money. Everyone just loves to talk! The thrill of talking!
A zillion years ago I reviewd the Carnegie "new" math software (still gets lots of visits on my blog). After two mistakes it just told you to ask your teacher, or it gave you the answer. Yea, right. (OK, and they had word problems with women riding in the Tour de France at 12 mph...)
*Three* zillion years ago (1998?) I started wondering why computer math programs never identified the most common mistakes and addressed them instead of just saiying "Wrong!" When somebody gives the GCF instead of the LCD... show the difference!
"He wanted me to give him the answer. That definitely isn’t the move, but what is?"
IDK if this is rhetorical or not, but I, too, have a 7 year old son at home, and I would have had him get out my measuring tape and measure out 50 inches and then 1 inch. Then I would ask, "Now, what do you notice about your number line?"
Predictions aren't supposed to be checked on for Those Influential People!!!
Oops, Khanmigo forgot to add on the +C at the end.
Somewhere in the "bronze" tier of my issues here, but, yes, definitely noticeable.
What’s wrong with his strategy? It’s an efficient use of the number line to show his thinking, whether he actually used it to solve the problem or figured it out in his head and just used the number line to show his process. What were you hoping he’d do?
The reasoning is correct, but as a human tutor facing students who repeatedly say "I don't know," if there are better coping methods, AI might also be able to imitate them.
The reason most AI tutors are difficult to use is that the designers did not consider the relationship between educators and learners, but instead treated AI as a question-answering machine.
I think that instead of providing answers in segments, it might be more helpful to list some examples from life to help students make associations. AI should be good at this approach.
It seems to me the problem is that math has still not moved very far from three ideas: the idea that there is only one answer to to each problem. All the AI stuff seems predicated on that. The idea that there is such a thing as grade level mathematics. Finally, the idea the learning of math is somehow linear... that you can't add 2 digit numbers until you can add i digit numbers nor can you learn multiplication before 3rd grade and so on.
All of this is hogwash as far as I am concerned, and it is the basis for why we are so terrible at teaching math, especially in the early grades. We begin to identify children as deficient at math as early as Kindergarten and we wonder where math anxiety comes from. There is so much effort to remediate before children have had adequate time to even learn anything. We have lists and lists that say children need to know this or that and if they don't they are somehow deficient.
Once I had the joy of having my students for 3 yrs.(2nd-4th grade). Math time was problem based with lots of objects available to use to picture the children's thinking. We also had morning calendar which was a time to play with numbers and patterns and also a time figure out how to describe each child's mathematical thinking. Children learned from each other as much or even more in some cases than from me. I never said we can't talk about numbers below zero because we are not in 5th grade or that we couldn't attempt to use those numbers in equations because we were not in 6th grade. We formulated problems that were interesting to us and THEN went about how we might solve them. We verbally explained our thinking, and when we could, we wrote about our solutions. As we did that we exposed everyone in the class to new ideas and widened our understanding of how numbers worked. My 4th graders were ready for many concepts contained in 6th grade math at the end of the year. And they all loved math which wasn't limited to a "math time" but was rather simply part of their day.
We are so lost as a nation concerning what teaching and learning means. The international test showed that despite all the remediation at 4th grade we are losing the middle and the bottom is more bottom than ever. It also showed that adults 16-60 are getting dumber and dumber, not just in math but in literacy and problem solving too. This is why so many were willing to hand our nation over to DT. Let him solve the problems, figure things out, it is all too much.
I teach an online Stats class, and all my students have to do a semester project, with multiple checkpoints. Hardly anyone gets it right on the first try, so I suppose I could grade these according to some rubric and say, "You met 5 of the 6 criteria, 83%". But since the pandemic I haven't been doing that. Instead I say, "You're missing this one item, please add it in so I can give you a full-credit grade." Or "Your conclusion still needs some work, please rewrite it and try to work in the phrase "not sufficient evidence to conclude." As a consequence of this, nearly everyone gets a perfect score, but hardly anyone gets their project done by the due date (it helps that I locate all this in a discussion forum, if I had to do all this back-and-forth by email I'd go crazy.)
Anyway, I mention this because we often tend to equate "demanding" with "strict", as in, "This was due last week, I'm taking points off for late work." And my experience is that many students like this "strict" approach because it lets them off the hook, they take their C and are happy to get it. I wouldn't say I'm a "warm demander", I'd say I'm an extremely annoying demander, because I won't let you go til you get it right. But being demanding in this way has meant being less demanding in other ways, it's meant having a policy where the only due date that really matters is the day I'm required to hand in grades (a thing I don't tell my students explicitly, but eventually they get the idea.)