Prior knowledge, skills and understanding reduce the constraints of working memory to focus on new learning, which when embedded is further released for more learning.
The more you know can do and understand, the more easily you will learn.
SO... WHY DO SO MANY SCHOOLS FOCUS THEIR MAJOR LEARNING EFFORTS ON THEIR FINAL YEAR?
When the evidence points to the greater effectiveness of prioritising the first and early years -
Another example of school data being more important than children?
It's also just short-sighted thinking. It looks like you're in the UK - I'm in the US where that is less of a phenomenon because high-stakes testing happens every year beginning around age 8. But it has the same impact of leading to short-sighted thinking, narrowing curriculum, and misunderstanding what students need over the long term to do well on tests, much less to be well educated overall. I think that short-sighted thinking is an inevitable consequence of the pressure schools are under to produce certain outcomes.
This is a thoughtful exploration of a complex topic, Dylan. But I have to chime in with some cautionary notes on the behavioral genetic side of things.
First, be really careful with Freddie deBoer. He's a powerful writer, but in his book The Cult of Smart he makes many misstatements around the current state of the science. Among other things, he suggests Eric Turkheimer's seminal paper (The Three Laws of Behavioral Genetics and What They Mean) indicates that "when it comes to certain quantitative outcomes in behavioral traits, children are shaped predominately by nature, not nurture." That's exactly wrong. Both in that paper and basically everything he's written for the past three decades, Turkheimer argues that genetic and environmental influences on complex behavioral traits are deeply entangled, and that trying to separate them out isn't all that fruitful. He eventually got so annoyed with deBoer he wrote this: https://ericturkheimer.substack.com/p/what-does-freddie-deboer-want-from
Similarly, while K. Paige Harden is much better on the science -- given she's an actual scientist (who did her PhD under Turkheimer) -- here too caution is warranted. One review, written by scientists sympathetic to her political message, contends that the central argument in The Genetic Lottery "mischaracterizes where the field of human genetics stands and what it promises." (https://onlinelibrary.wiley.com/doi/full/10.1111/evo.14449)
The review of her book in the New York Review of Books was even more withering:
Genome studies can illuminate things that genes cause, but genes don’t cause everything. Whatever scientific evidence emerges regarding genetic populations, it won’t explain why some students do well on tests any more than it will explain why some social scientists construct essentialist theories of intelligence. Educational success and biological essentialism are social and cultural phenomena, not genetic phenomena. True, genes help shape people, and people make up social and cultural situations. Likewise, grammar helps shape sentences and sentences make up Harden’s book. But we can’t reduce her contention that genetic differences cause social differences to the syntactical rules of an English sentence. Meanwhile, beneath Harden’s protestations that she’s an egalitarian hides a stealthy affirmation of the old, tenacious view that races and classes are natural kinds
Thanks for the comment, Ben. Those are fair criticisms. I'll dive into these links when I have a bit of time - thanks for sharing them! Both DeBoer and Harden are summarizing large bodies of scholarship, something I should do is try to better understand more of the underlying research.
One thing I agree with both writers about is that lots of people see intelligence as a third rail, and when people avoid talking about it we kindof concede that space to the eugenicists and race science folks who are happy to take up the mantle. I think there's a parallel story playing out for teachers. It's much less deliberate or sinister, teachers often avoid talking about intelligence publicly or engaging with that area of research. But lots of teachers I talk to end up with their own theories of what it means to be smart based on their everyday experience. I've met so many teachers who insist that some kids just can't learn certain things. While I'm far from an expert on this stuff, I was motivated to write about it because the more I learn, the more I feel like a lot of people's intuitions about intelligence and student potential are wrong.
None of that defends the biological determinism, that's not my goal, but hopefully it gives a sense of why I think this area is worth exploring and writing about.
I completely agree about the "third rail" aspect of intelligence. In fact, and as full disclosure, I was so interested that when I learned that Paige Harden lived in Austin, I sought her out to learn more from her, and (ahem) we were dating when she was writing her book. Suffice to say we had many conversations about this topic.
But a funny thing happened on the way to my own self-education on matters of intelligence, and in particular the role that genes play in influencing it. I won't try to summarize everything here in a Substack comment, but suffice to say, the more I learned, the more I came to believe that current science is not of much use in helping us to disentangle nature from nurture, they are too entwined. In follow-on commentary to their review of Harden's book, the authors of the the NY Review of Books review invoke a metaphor I've found helpful: The California coastlines exist, but how much of it is "caused" the Pacific Ocean versus the North American continent? The question is meaningless.
At risk of self promotion (as always), just this week I wrote about trying to look at things from a fresh perspective. A new book by Leslie Valient, a computer scientists at Harvard who helped develop artificial neural networks, argues that we should replace the idea of "intelligence" with the notion that what makes humans unique is our capacity to be "educable," to weave together experiences and beliefs to tackle novel problems. As I note, this definition of educable actually aligns with what some call "general intelligence," but I think there may be value in adopting "educability" as our framework, in part because it frees us of the sordid history of intelligence and eugenics, and in part because it captures something elegant about the role education plays in forming who we are.
The classroom environment is more complicated, and what students need in their post education lives is more than knowledge. If every student attribute is somehow similar, then perhaps learning rates can be similar. For example, most can learn fraction division in similar rates but transferring it to further situations and making connections to solve more complex problems is a totally different story. There are many things to consider (executive functioning skills for instance, reading comprehension level, etc) in a classroom.
I think that's a good point. There's a tension here where learning speed is more similar the more we zoom in on the micro level of individual skills, but school is about more than a collection of little skills. When we zoom out and think about bigger ideas and connections everything inevitably gets messier.
one problem with most classrooms is that they reward speed over depth. Deep learning most times is slow and students are discouraged from going deep. Anecdotally, when studying computing a few decades ago (no internet, no google then) I was the "fast" learner and I had a close friend that was slower but deeper. He will asked me questions when preparing for the exam on topics I thought I knew that made me go back to the notes and rethink the topic. I got the HD and he got a C or a D if he was lucky, which it seem unfair to me.
The Koedinger paper is at best very difficult to understand and believe.
First, EVERYTHING is filtered through a big, complex statistical model, and all the graphs that you see are outputs of this complex model. So it's not clear if all the parallel lines that you see are genuinely reflective of the data, or just artifacts of the regression model. Although the sample size sounds impressive, the data was actually a combination of many different smaller datasets across a wide range of studies, ranging from K-5 up to university level, so it's not clear how to think about the study population. Also, if you read the details of the paper, it's clear that some students really did learn faster than others, although the authors don't mention this in their summary.
Another problem is that the "learning" was very simplistic -- it's basically just simple computer multiple-choice questions that students can take and retake. It's not surprising that most students were able to learn the correct answers after 8 attempts, but that hardly justifies the sweeping conclusions. If you're genuinely interested in this, I recommend that you take a look at the supplementary materials, which have examples of the sorts of problems that students were answering. It's very underwhelming, and again this weak data does not support the conclusions of the paper.
Here's a simple analogy. Most people can learn how to operate a modern cash register with a little training (although even in this case there will be some who just don't get it). But that doesn't mean that everyone learns at the same speed -- instead, it means that operating a modern cash register is so cognitively undemanding that the majority of people can pick it up without difficulty. You don't even have to know any math, because the machine automatically calculates everything! But it would be absurd to suggest that that indicates that everyone learns at the same speed.
I agree that effective scaffolding can help students learn, and that we always want to improve our methods. But it doesn't follow that if just got the scaffolding right then everyone would learn the material.
Thanks for the comment! I won't pretend to understand all the technical details of the paper well enough to address your concerns, you're probably right.
Here's a slightly weaker version of my thesis, I'm curious if you would agree with it:
Teachers can narrow the gap between faster learners and slower learners by filling in gaps in prior knowledge and scaffolding complex ideas into manageable chunks.
That's not too far from what I argued in the blog, and I still find that to be a helpful guide for practical questions about teaching.
The other thing I'll say is that one type of learning that has never ended in my teaching career is learning that some skill I think of as a single skill is actually composed of a few smaller skills I didn't realize existed, and that once I've made some of those adjustments to my teaching kids have been more successful. I think it's much more helpful to respond that way to a teaching challenge than to attribute it to student-side characteristics of whether they can learn something.
Prior knowledge, skills and understanding reduce the constraints of working memory to focus on new learning, which when embedded is further released for more learning.
The more you know can do and understand, the more easily you will learn.
SO... WHY DO SO MANY SCHOOLS FOCUS THEIR MAJOR LEARNING EFFORTS ON THEIR FINAL YEAR?
When the evidence points to the greater effectiveness of prioritising the first and early years -
Another example of school data being more important than children?
It's also just short-sighted thinking. It looks like you're in the UK - I'm in the US where that is less of a phenomenon because high-stakes testing happens every year beginning around age 8. But it has the same impact of leading to short-sighted thinking, narrowing curriculum, and misunderstanding what students need over the long term to do well on tests, much less to be well educated overall. I think that short-sighted thinking is an inevitable consequence of the pressure schools are under to produce certain outcomes.
This is a thoughtful exploration of a complex topic, Dylan. But I have to chime in with some cautionary notes on the behavioral genetic side of things.
First, be really careful with Freddie deBoer. He's a powerful writer, but in his book The Cult of Smart he makes many misstatements around the current state of the science. Among other things, he suggests Eric Turkheimer's seminal paper (The Three Laws of Behavioral Genetics and What They Mean) indicates that "when it comes to certain quantitative outcomes in behavioral traits, children are shaped predominately by nature, not nurture." That's exactly wrong. Both in that paper and basically everything he's written for the past three decades, Turkheimer argues that genetic and environmental influences on complex behavioral traits are deeply entangled, and that trying to separate them out isn't all that fruitful. He eventually got so annoyed with deBoer he wrote this: https://ericturkheimer.substack.com/p/what-does-freddie-deboer-want-from
Similarly, while K. Paige Harden is much better on the science -- given she's an actual scientist (who did her PhD under Turkheimer) -- here too caution is warranted. One review, written by scientists sympathetic to her political message, contends that the central argument in The Genetic Lottery "mischaracterizes where the field of human genetics stands and what it promises." (https://onlinelibrary.wiley.com/doi/full/10.1111/evo.14449)
The review of her book in the New York Review of Books was even more withering:
Genome studies can illuminate things that genes cause, but genes don’t cause everything. Whatever scientific evidence emerges regarding genetic populations, it won’t explain why some students do well on tests any more than it will explain why some social scientists construct essentialist theories of intelligence. Educational success and biological essentialism are social and cultural phenomena, not genetic phenomena. True, genes help shape people, and people make up social and cultural situations. Likewise, grammar helps shape sentences and sentences make up Harden’s book. But we can’t reduce her contention that genetic differences cause social differences to the syntactical rules of an English sentence. Meanwhile, beneath Harden’s protestations that she’s an egalitarian hides a stealthy affirmation of the old, tenacious view that races and classes are natural kinds
(https://www.nybooks.com/articles/2022/04/21/why-biology-is-not-destiny-genetic-lottery-kathryn-harden/)
Thanks for the comment, Ben. Those are fair criticisms. I'll dive into these links when I have a bit of time - thanks for sharing them! Both DeBoer and Harden are summarizing large bodies of scholarship, something I should do is try to better understand more of the underlying research.
One thing I agree with both writers about is that lots of people see intelligence as a third rail, and when people avoid talking about it we kindof concede that space to the eugenicists and race science folks who are happy to take up the mantle. I think there's a parallel story playing out for teachers. It's much less deliberate or sinister, teachers often avoid talking about intelligence publicly or engaging with that area of research. But lots of teachers I talk to end up with their own theories of what it means to be smart based on their everyday experience. I've met so many teachers who insist that some kids just can't learn certain things. While I'm far from an expert on this stuff, I was motivated to write about it because the more I learn, the more I feel like a lot of people's intuitions about intelligence and student potential are wrong.
None of that defends the biological determinism, that's not my goal, but hopefully it gives a sense of why I think this area is worth exploring and writing about.
I completely agree about the "third rail" aspect of intelligence. In fact, and as full disclosure, I was so interested that when I learned that Paige Harden lived in Austin, I sought her out to learn more from her, and (ahem) we were dating when she was writing her book. Suffice to say we had many conversations about this topic.
But a funny thing happened on the way to my own self-education on matters of intelligence, and in particular the role that genes play in influencing it. I won't try to summarize everything here in a Substack comment, but suffice to say, the more I learned, the more I came to believe that current science is not of much use in helping us to disentangle nature from nurture, they are too entwined. In follow-on commentary to their review of Harden's book, the authors of the the NY Review of Books review invoke a metaphor I've found helpful: The California coastlines exist, but how much of it is "caused" the Pacific Ocean versus the North American continent? The question is meaningless.
At risk of self promotion (as always), just this week I wrote about trying to look at things from a fresh perspective. A new book by Leslie Valient, a computer scientists at Harvard who helped develop artificial neural networks, argues that we should replace the idea of "intelligence" with the notion that what makes humans unique is our capacity to be "educable," to weave together experiences and beliefs to tackle novel problems. As I note, this definition of educable actually aligns with what some call "general intelligence," but I think there may be value in adopting "educability" as our framework, in part because it frees us of the sordid history of intelligence and eugenics, and in part because it captures something elegant about the role education plays in forming who we are.
https://buildcognitiveresonance.substack.com/p/are-we-intelligent-or-are-we-educable
I'll be curious to see how your thoughts evolve around this topic!
The classroom environment is more complicated, and what students need in their post education lives is more than knowledge. If every student attribute is somehow similar, then perhaps learning rates can be similar. For example, most can learn fraction division in similar rates but transferring it to further situations and making connections to solve more complex problems is a totally different story. There are many things to consider (executive functioning skills for instance, reading comprehension level, etc) in a classroom.
I think that's a good point. There's a tension here where learning speed is more similar the more we zoom in on the micro level of individual skills, but school is about more than a collection of little skills. When we zoom out and think about bigger ideas and connections everything inevitably gets messier.
one problem with most classrooms is that they reward speed over depth. Deep learning most times is slow and students are discouraged from going deep. Anecdotally, when studying computing a few decades ago (no internet, no google then) I was the "fast" learner and I had a close friend that was slower but deeper. He will asked me questions when preparing for the exam on topics I thought I knew that made me go back to the notes and rethink the topic. I got the HD and he got a C or a D if he was lucky, which it seem unfair to me.
The Koedinger paper is at best very difficult to understand and believe.
First, EVERYTHING is filtered through a big, complex statistical model, and all the graphs that you see are outputs of this complex model. So it's not clear if all the parallel lines that you see are genuinely reflective of the data, or just artifacts of the regression model. Although the sample size sounds impressive, the data was actually a combination of many different smaller datasets across a wide range of studies, ranging from K-5 up to university level, so it's not clear how to think about the study population. Also, if you read the details of the paper, it's clear that some students really did learn faster than others, although the authors don't mention this in their summary.
Another problem is that the "learning" was very simplistic -- it's basically just simple computer multiple-choice questions that students can take and retake. It's not surprising that most students were able to learn the correct answers after 8 attempts, but that hardly justifies the sweeping conclusions. If you're genuinely interested in this, I recommend that you take a look at the supplementary materials, which have examples of the sorts of problems that students were answering. It's very underwhelming, and again this weak data does not support the conclusions of the paper.
Here's a simple analogy. Most people can learn how to operate a modern cash register with a little training (although even in this case there will be some who just don't get it). But that doesn't mean that everyone learns at the same speed -- instead, it means that operating a modern cash register is so cognitively undemanding that the majority of people can pick it up without difficulty. You don't even have to know any math, because the machine automatically calculates everything! But it would be absurd to suggest that that indicates that everyone learns at the same speed.
I agree that effective scaffolding can help students learn, and that we always want to improve our methods. But it doesn't follow that if just got the scaffolding right then everyone would learn the material.
Thanks for the comment! I won't pretend to understand all the technical details of the paper well enough to address your concerns, you're probably right.
Here's a slightly weaker version of my thesis, I'm curious if you would agree with it:
Teachers can narrow the gap between faster learners and slower learners by filling in gaps in prior knowledge and scaffolding complex ideas into manageable chunks.
That's not too far from what I argued in the blog, and I still find that to be a helpful guide for practical questions about teaching.
The other thing I'll say is that one type of learning that has never ended in my teaching career is learning that some skill I think of as a single skill is actually composed of a few smaller skills I didn't realize existed, and that once I've made some of those adjustments to my teaching kids have been more successful. I think it's much more helpful to respond that way to a teaching challenge than to attribute it to student-side characteristics of whether they can learn something.