Hacker News Reader

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.

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.

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-person...

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.

> 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.