3 ms·

Well, it is a fact that Zuckerberg (and Gates') personalized education projects didn't work. Replicating the effects of tutoring with technology was exactly wha

Well, it is a fact that Zuckerberg (and Gates') personalized education projects didn't work. Replicating the effects of tutoring with technology was exactly what they tried to do. If there were such massive gains to be had, why didn't they at least manage to capture some gains? You don't have a theory to answer that. The article writer does. In their view, learning is a social activity, and kids don't like to sit in a room with dozens of others each with their own personalized algorithmic "teacher". That's not really all that surprising. If it was only about personalization, it would be easy enough to do that with far simpler technology than modern computers. The article writer's theory also neatly and immediately suggests an explanation for why personal tutoring still works, when automated custom tutoring manifestly hasn't so far. That's because it is social. Even if it's just you and your overworked homeschooling parent, you care because they care. Personalization isn't key. Personalization from a tutor works because it's hard and costly, and convinces the student that this matters. That's just a theory. It's not necessarily right, but it does present a convincing explanation for what we observe, and in particular the observation worth repeating, that no one can seem to get algorithmic personalized education to work, no matter how rich and clever they are. And from my personal experience with teachers (including my parents) I have often heard from them that the challenge isn't explaining, it's to get the students to care.

dalbasal 3y ago
Well said, and it rhymes with many of the mistakes we have made in the past. We discount the social, the emotional, the storytelling... Often they're what matters. It also rhymes with common parenting folk wisdom that having "the wrong friends" is detrimental to grades. >the challenge isn't explaining, it's to get the students to care. This. The reason not one student learned French in my four years of mandatory French classes is that we didn't want to know french. We didn't know any french people. We didn't care about french songs or movies. OTOH, when telenovelas became a local fad, every girl had a clique of Spanish speaking 13 year olds. The actual problem in that case is: How do you teach french/philosophy/whatever to a class full of kids who don't want to learn that stuff? Maybe they want to pass exams and get grades. Maybe you can rejigger discipline mechanisms to motivate them... but that is never going to get you to good.
- ghaff 3y ago
  I'd reword your French example a bit. I had the proverbial four years of high school French which I cared about and performed at probably about a median level (which is to say fairly well) of my high school subjects in general. Especially in a pre-Web world though, absent pestering my parents to arrange some sort of exchange student situation--which would have gone well beyond my level of caring for any other subject and may or may not have been practical--I was probably never going to get to a high level of French proficiency without some sort of immersion.
- nradov 3y ago
  After about age 10, children are more influenced by their peers than by parents or teachers. If friends aren't interested in learning then they won't care either. https://www.psychologytoday.com/intl/blog/a-matter-of-personality/202301/who-influences-teens-more-peers-or-family https://www.psychologytoday.com/intl/blog/a-matter-of-person...
TheOtherHobbes 3y ago
Not yet, anyway. It might be possible with an AI that modelled human psychology. I don't think that's an impossible goal. In fact I wouldn't be surprised to see it happen within 10 years (climate change and war permitting.) The problem is that people like Gates and Zuckerberg are fully paid up members of the "You can't let people use technology without exploiting them" group. So AI-assisted personalised teaching would come with a cost - a corporate learning and skills profile of the pupils. Which would inevitably be sold to other corporates. From a market point of view that kind of information would be immensely valuable for hiring and HR. But it would be it would be apocalyptically disastrous from the point of view of personal freedom and political accountability. And no doubt the service would be tiered, so a minimal unsophisticated model would be available to almost everyone, while more sophisticated and effective models would require a paid subscription. Teaching is one of those professions where you can see what cultures really value. In the US it's clearly not open access to high quality education. Automated tutoring won't change that unless cultural values change too. Without that, it will just recreate the current situation with added social costs.
thaumasiotes 3y ago
> Personalization isn't key. Personalization from a tutor works because it's hard and costly, and convinces the student that this matters. > That's just a theory. It's not necessarily right, but it does present a convincing explanation for what we observe Far from it. We observe that people who already believe that they need to learn something do better with individual tutoring. > and in particular the observation worth repeating, that no one can seem to get algorithmic personalized education to work, no matter how rich and clever they are. People got algorithmic personalized education to work quite a long time ago. It's called a book. The algorithm is "when the student asks about something, tell them". And it works very well, just like it always has. What people aren't getting to work is forcing students who don't want to know things to know those things anyway.

dalbasal 3y ago

Well said, and it rhymes with many of the mistakes we have made in the past. We discount the social, the emotional, the storytelling... Often they're what matters. It also rhymes with common parenting folk wisdom that having "the wrong friends" is detrimental to grades. >the challenge isn't explaining, it's to get the students to care. This. The reason not one student learned French in my four years of mandatory French classes is that we didn't want to know french. We didn't know any french people. We didn't care about french songs or movies. OTOH, when telenovelas became a local fad, every girl had a clique of Spanish speaking 13 year olds. The actual problem in that case is: How do you teach french/philosophy/whatever to a class full of kids who don't want to learn that stuff? Maybe they want to pass exams and get grades. Maybe you can rejigger discipline mechanisms to motivate them... but that is never going to get you to good.

ghaff 3y ago
I'd reword your French example a bit. I had the proverbial four years of high school French which I cared about and performed at probably about a median level (which is to say fairly well) of my high school subjects in general. Especially in a pre-Web world though, absent pestering my parents to arrange some sort of exchange student situation--which would have gone well beyond my level of caring for any other subject and may or may not have been practical--I was probably never going to get to a high level of French proficiency without some sort of immersion.
nradov 3y ago
After about age 10, children are more influenced by their peers than by parents or teachers. If friends aren't interested in learning then they won't care either. https://www.psychologytoday.com/intl/blog/a-matter-of-personality/202301/who-influences-teens-more-peers-or-family https://www.psychologytoday.com/intl/blog/a-matter-of-person...

TheOtherHobbes 3y ago

Not yet, anyway. It might be possible with an AI that modelled human psychology. I don't think that's an impossible goal. In fact I wouldn't be surprised to see it happen within 10 years (climate change and war permitting.) The problem is that people like Gates and Zuckerberg are fully paid up members of the "You can't let people use technology without exploiting them" group. So AI-assisted personalised teaching would come with a cost - a corporate learning and skills profile of the pupils. Which would inevitably be sold to other corporates. From a market point of view that kind of information would be immensely valuable for hiring and HR. But it would be it would be apocalyptically disastrous from the point of view of personal freedom and political accountability. And no doubt the service would be tiered, so a minimal unsophisticated model would be available to almost everyone, while more sophisticated and effective models would require a paid subscription. Teaching is one of those professions where you can see what cultures really value. In the US it's clearly not open access to high quality education. Automated tutoring won't change that unless cultural values change too. Without that, it will just recreate the current situation with added social costs.

thaumasiotes 3y ago

> Personalization isn't key. Personalization from a tutor works because it's hard and costly, and convinces the student that this matters. > That's just a theory. It's not necessarily right, but it does present a convincing explanation for what we observe Far from it. We observe that people who already believe that they need to learn something do better with individual tutoring. > and in particular the observation worth repeating, that no one can seem to get algorithmic personalized education to work, no matter how rich and clever they are. People got algorithmic personalized education to work quite a long time ago. It's called a book. The algorithm is "when the student asks about something, tell them". And it works very well, just like it always has. What people aren't getting to work is forcing students who don't want to know things to know those things anyway.