This was an amazing read. Out of all the points, I found the idea of "near decomposability" quite relatable, both as a learner and a teacher. I can now clearly articulate why the "holistic" or "integrated curriculum" in the school level usually fail despite good intentions.
I think it points to the challenge of implementing curriculum well in general.
With programs that compartmentalize learning into bits but do not require teachers who know where this is all going (for example, elementary math teachers not understanding principles of algebra and geometry), key ideas in the building blocks stage may be lost in the service of the right answer or workaround to this or that question.
Thank you for making all of this logic so crystal clear. That image of the ant will stay with me for quite a while.
I think this quote warmed my heart the most.
“A well-designed learning environment rests on three pillars: what we teach (curriculum), how we teach it (instruction), and how we know students have learned it (assessment). These are not mysterious processes accessible only to initiates; they are engineering problems, amenable to analysis, iteration, and improvement. The extent to which these three elements are aligned largely determines whether learning occurs at all.”
To describe a curriculum designer as an engineer is more accurate than “someone with a PhD in curriculum”.
In my 24 years in the classroom, I rarely encountered curriculum that was expertly engineered. Most of it was crap. I ended up buying CMC math on eBay, being grateful for Read Well, and engineering my own way of teaching writing. Luckily, I had a supportive principal who cared more about results than obedience to the district.
You're right that the beach matters. You're right that decomposition aids learning. You're right that "natural" is not a pedagogical strategy.
But the ant metaphor has a strange implication you don't address. The ant crossing the beach isn't trying to learn-it's trying to get somewhere. It has a goal. The complex path is a byproduct of purposeful movement through a complex environment.
What's the educational equivalent? Children aren't just acquiring knowledge-they're trying to do something, understand something, become something. The "goal" shapes the path as much as the terrain does.
Your framework is missing a theory of purpose. Why is the ant crossing the beach at all?
Kieran Egan had an answer: the ant crosses the beach because the beach is full of stories, wonders, heroes, extremes-things that matter to a creature with imagination. The journey itself can be engaging if you design it right. Not "natural" engagement from abandoning children to complexity, but designed engagement through cognitive tools that speak to human meaning-making.
Your beach can be smooth AND significant. Your decomposition can be cognitively manageable AND imaginatively engaging. But only if you recognize that the ant isn't just an information processor. It's a storyteller. A wonder-seeker. A meaning-maker.
I certainly notice that the student who have a sense of purpose in what they are learning tend to show far better outcomes than the ones who do not. This isn't simply broken down by ability, either! I can see it in the same student depending on whether they are passively learning vs aiming toward a goal, even one as simple as understanding.
Perhaps one way to think about this is that a well-designed environment for learning would not only address cognitive architecture, but also the topic and the context in which the learned material will be understood.
So many golden sentences, but I'll begin with these: "Natural environments share several features that make them hostile to efficient learning . . . They provide inconsistent feedback; sometimes immediately, sometimes never, sometimes misleadingly." Coincidentally, I was re-listening today to your discussion about feedback loops on the Better Teaching: Only Stuff that Works podcast, which I refer to in Pathways to Information: Types of Knowledge (https://harriettjanetos.substack.com/p/pathways-to-information-accessing?r=5spuf):
"While struggling to make better day-to-day decisions, teachers often fall victim to insufficient information from feedback loops. In sports—and certainty in chess—[Hendrick] asserts, if you make a mistake, you get instant feedback. The loops are really short, and bad decisions are 'punished quickly, and you see what you did wrong'. But with teaching the feedback loops are long—so long, in fact that 'you might not find out that you did something wrong for six months, or even longer—or ever!'"
Thanks for this essay, professor. It brings to light the important—to me, at least—idea that teaching requires manipulating the instructional environment for the specific purpose of learning.
As I read your post, I was reminded of a commonplace notion in behavioral views of learning: Ontological change in behavior ("learning") are a function of environmental processes (e.g., reinforcement) just as phylogenetic changes in species are a function of environment processes (e.g., selection).
I also was reminded of an idea expressed by one of the elder behaviorists, Og Lindsley. I'll quote directly from the opening of a chapter which is one of the places he published the idea in ~1964:
"Children are not retarded. Only their behavior in average environments is sometimes retarded. In fact, it is modern science's ability to design suitable environments for these children that is retarded. We design environments to maintain life, but not to maintain dignified behavior."
Timely! I've a copy of Simon's book, guess this is a sign to read it.
The post made me think about how we learn and decide how much is enough to be useful. As far as learning or instruction goes, to my observation, it feels like learning in parts apply to learnings on-the-job. There's a tight constraint for time and resources, but a team needs to consume and comprehend a non trivial amount of information to be effective at a project. Iterating and course correcting along the way. Not easy as the world is complex and always on the move.
I wonder comparing organisational design with education systems design is an appropriate lens to address this ? The goal of the two systems is absolutely different!
Oh for sure. Major difference between schools and businesses is that businesses cannot avoid contact with the real environment at all. I felt I could draw some principles from former and recontextualise them to the latter
Thanks for the essay! I think it raises a lot of powerful questions. As someone who works in higher ed and does work on police training, I was wondering if you could share your insights on how this lines up with where some cognitive science and the 4 ways of knowing (propositional, procedural, perspectival, and participatory - for more info, see this paper - https://osf.io/preprints/osf/gw7r5_v2)? In much police training we seem to have problem where the tasks have been decomposed to the point they do not transfer to actual performance in the field.
Interesting essay and I was definitely pointing at work being done in a variety of areas suggesting that decomposition is problematic. I would argue that we are on the edge of a paradigm change where biology (specifically work on complexity and emergence) becomes the reference point for "science" instead of physics.
The best thing I’ve read all day! (It’s only 6 am here - okay, all week.)
I appreciate the revival of brilliant work on “teaching and learning” (sorry) that has been forgotten, rather than the new and trendy “strategies” compendium that seem to dominate contemporary pedagogical literature.
I can quibble with some of your assumptions about the implications of decomposition, but this framework rings true on every level. The devil is always in the details, and knowing students and their motivations as well as their misconceptions is as important as knowing the curriculum. (And more important, if the curriculum is poorly organized or lacking as you rightly imply many are.)
This post is a wonderful review of what we ALREADY KNOW!
I'm now re-reading some of those links back to Simon's work...
This quote, from near the end of your post:
"If you were designing a learning system from scratch, you would not copy human learning wholesale. You would correct for its limitations: our working memory constraints, our susceptibility to distraction, our tendency toward the path of least resistance."
This is ONE of my conclusions from your post = it seems that there is some potential there for AI to benefit students' learning, however, there is still the issue of "distraction" in the online world, and it doesn't seem that the AI models have managed to NOT be distracted? Maybe I'm wrong?
Also, regardless of AI, classrooms in schools are one environment that teachers have some potential to support learning through this lens - classrooms are the "educators' beaches" and THAT is within the control of classroom teachers who have some knowledge of evidence-based research. It's a pity that more teachers don't know about posts like this?
Thanks again for supporting MY Personal Learning, and I'll try to complete my "reading homework" so I can learn more! MUCH appreciated! :)
Re: "The best curricula, like the best watches, are designed so that component skills can be mastered to the point of automaticity before being combined." We use subassemblies and use them like the watchmaker, but they can be more varied than this article appears to concede. We mimicked other assembly processes. Sub-assemblies also can be situated in a select part of a larger, complex problem that is interesting to students, particularly the older ones we serve. I may have misinterpreted something, but we have found such absolutes do not work very well across all types of knowledge in a domain. Learners need much more practice with concepts than most curricula afford, making automaticity a further destination. In some cases, combinations can actually support some moves to automaticity of sub-assemblies. It becomes a step along the journey to fluency. This is more complex to "hold" in the design, but I'm glad we remained open to that possibility.
Simon's ant was also the inspiration for the behavior based robotics of Rodney Brooks, who recognized the role that environmental contingencies play in real time adaptive behavior. Although perhaps not recognized at the time by Brooks, his insight was a rediscovery of Skinner's behavior analysis. Skinner conceptualized moment-to- moment behavior change to be a variation-selection-retention process similar to biological evolution, but representing selection of behavior, not allele populations. This insight led to the development of programmed instruction, a form of explicit instruction that preceded today's AI based programs. Interestingly, today's AI technology often deploys a form of machine learning called Deep Reinforcement Learning. Good luck finding any references to Skinner and other early learning literature in the AI scholarship!
near decomposability is a fascinating concept that i will definitely try to keep in mind when designing anything grander than a single lesson. but i do want to add two things.
first, its good to be aware of naturalist tendencies and not confuse natural with good learning environment. but ofc we also should not forget, for what environment our brains, bodies and minds developed and expect learners to follow our nicely designed curriculum, just bc its nearly decomposable. just to say that motivation is another crucial factor and that often comes from seeing a big picture and understanding the importance of the learned through its relationships to other and concepts and abilities. if we hammer down the basics and only allow the students to see what comes next if they mastered it, we don’t teach children how to think, but how to follow menial learning tasks. we rob them of an intrinsic motivation of exploring and understand the world - which is huge and complex and cant be understood all at once. and i agree, if you tried that, you would drown in information. all im saying is that near decomposability is only part of how we should think about learning. yes, we need to make sure we get the foundation and the first floor before the second one can be build. but we also need to know how the building is supposed to look in the end and why we want to build it to begin with. otherwise the second floor wont match the first, its gonna be crooked or we will just stop and abandon the house unfinished because we lost interest.
Your articles are fantastic, Carl. Your distillation of educational concepts is very helpful.
Lately in my teaching, I have been feeling that lesson structure is an underrated component of teaching, and this article helps clarify that, so thank you!
It's ironic that educational researchers of the 80's and 90's took the wrong lesson from the ant metaphor. Simon's key point (and your point) is that if behaviour is highly determined by environmental conditions, then we have to be really careful when we design learning environments. Unfortunately, researchers-of-the-day used the metaphor to argue the precise opposite. They felt the ant story illustrated how cognition was primarily situated, which in turn meant that learners should be immersed in complex, real-life scenarios. (Of course, in making this argument, they had to conveniently ignore the very real cognitive limits of learners).
Thank you for writing a wonderful piece that calls that mistake out.
A well written and very instructive essay. Thank you.
This was an amazing read. Out of all the points, I found the idea of "near decomposability" quite relatable, both as a learner and a teacher. I can now clearly articulate why the "holistic" or "integrated curriculum" in the school level usually fail despite good intentions.
I think it points to the challenge of implementing curriculum well in general.
With programs that compartmentalize learning into bits but do not require teachers who know where this is all going (for example, elementary math teachers not understanding principles of algebra and geometry), key ideas in the building blocks stage may be lost in the service of the right answer or workaround to this or that question.
This! "elementary math teachers not understanding principles of algebra and geometry"
Yes, it can be confusing 🫤 I initially thought it’s Umes’s comment which wasn’t sitting with me well but thanks for another perspective @steven 🙂
Thank you for making all of this logic so crystal clear. That image of the ant will stay with me for quite a while.
I think this quote warmed my heart the most.
“A well-designed learning environment rests on three pillars: what we teach (curriculum), how we teach it (instruction), and how we know students have learned it (assessment). These are not mysterious processes accessible only to initiates; they are engineering problems, amenable to analysis, iteration, and improvement. The extent to which these three elements are aligned largely determines whether learning occurs at all.”
To describe a curriculum designer as an engineer is more accurate than “someone with a PhD in curriculum”.
In my 24 years in the classroom, I rarely encountered curriculum that was expertly engineered. Most of it was crap. I ended up buying CMC math on eBay, being grateful for Read Well, and engineering my own way of teaching writing. Luckily, I had a supportive principal who cared more about results than obedience to the district.
There is a stoic or Buddhist quality to the quote: educators should control what they can control.
You're right that the beach matters. You're right that decomposition aids learning. You're right that "natural" is not a pedagogical strategy.
But the ant metaphor has a strange implication you don't address. The ant crossing the beach isn't trying to learn-it's trying to get somewhere. It has a goal. The complex path is a byproduct of purposeful movement through a complex environment.
What's the educational equivalent? Children aren't just acquiring knowledge-they're trying to do something, understand something, become something. The "goal" shapes the path as much as the terrain does.
Your framework is missing a theory of purpose. Why is the ant crossing the beach at all?
Kieran Egan had an answer: the ant crosses the beach because the beach is full of stories, wonders, heroes, extremes-things that matter to a creature with imagination. The journey itself can be engaging if you design it right. Not "natural" engagement from abandoning children to complexity, but designed engagement through cognitive tools that speak to human meaning-making.
Your beach can be smooth AND significant. Your decomposition can be cognitively manageable AND imaginatively engaging. But only if you recognize that the ant isn't just an information processor. It's a storyteller. A wonder-seeker. A meaning-maker.
That's what you're missing.
I certainly notice that the student who have a sense of purpose in what they are learning tend to show far better outcomes than the ones who do not. This isn't simply broken down by ability, either! I can see it in the same student depending on whether they are passively learning vs aiming toward a goal, even one as simple as understanding.
Perhaps one way to think about this is that a well-designed environment for learning would not only address cognitive architecture, but also the topic and the context in which the learned material will be understood.
So many golden sentences, but I'll begin with these: "Natural environments share several features that make them hostile to efficient learning . . . They provide inconsistent feedback; sometimes immediately, sometimes never, sometimes misleadingly." Coincidentally, I was re-listening today to your discussion about feedback loops on the Better Teaching: Only Stuff that Works podcast, which I refer to in Pathways to Information: Types of Knowledge (https://harriettjanetos.substack.com/p/pathways-to-information-accessing?r=5spuf):
"While struggling to make better day-to-day decisions, teachers often fall victim to insufficient information from feedback loops. In sports—and certainty in chess—[Hendrick] asserts, if you make a mistake, you get instant feedback. The loops are really short, and bad decisions are 'punished quickly, and you see what you did wrong'. But with teaching the feedback loops are long—so long, in fact that 'you might not find out that you did something wrong for six months, or even longer—or ever!'"
Thanks for this essay, professor. It brings to light the important—to me, at least—idea that teaching requires manipulating the instructional environment for the specific purpose of learning.
As I read your post, I was reminded of a commonplace notion in behavioral views of learning: Ontological change in behavior ("learning") are a function of environmental processes (e.g., reinforcement) just as phylogenetic changes in species are a function of environment processes (e.g., selection).
I also was reminded of an idea expressed by one of the elder behaviorists, Og Lindsley. I'll quote directly from the opening of a chapter which is one of the places he published the idea in ~1964:
"Children are not retarded. Only their behavior in average environments is sometimes retarded. In fact, it is modern science's ability to design suitable environments for these children that is retarded. We design environments to maintain life, but not to maintain dignified behavior."
==> https://journals.sagepub.com/doi/pdf/10.1177/002205746414700107
Timely! I've a copy of Simon's book, guess this is a sign to read it.
The post made me think about how we learn and decide how much is enough to be useful. As far as learning or instruction goes, to my observation, it feels like learning in parts apply to learnings on-the-job. There's a tight constraint for time and resources, but a team needs to consume and comprehend a non trivial amount of information to be effective at a project. Iterating and course correcting along the way. Not easy as the world is complex and always on the move.
I wonder comparing organisational design with education systems design is an appropriate lens to address this ? The goal of the two systems is absolutely different!
Oh for sure. Major difference between schools and businesses is that businesses cannot avoid contact with the real environment at all. I felt I could draw some principles from former and recontextualise them to the latter
Thanks for the essay! I think it raises a lot of powerful questions. As someone who works in higher ed and does work on police training, I was wondering if you could share your insights on how this lines up with where some cognitive science and the 4 ways of knowing (propositional, procedural, perspectival, and participatory - for more info, see this paper - https://osf.io/preprints/osf/gw7r5_v2)? In much police training we seem to have problem where the tasks have been decomposed to the point they do not transfer to actual performance in the field.
Interesting essay and I was definitely pointing at work being done in a variety of areas suggesting that decomposition is problematic. I would argue that we are on the edge of a paradigm change where biology (specifically work on complexity and emergence) becomes the reference point for "science" instead of physics.
The best thing I’ve read all day! (It’s only 6 am here - okay, all week.)
I appreciate the revival of brilliant work on “teaching and learning” (sorry) that has been forgotten, rather than the new and trendy “strategies” compendium that seem to dominate contemporary pedagogical literature.
I can quibble with some of your assumptions about the implications of decomposition, but this framework rings true on every level. The devil is always in the details, and knowing students and their motivations as well as their misconceptions is as important as knowing the curriculum. (And more important, if the curriculum is poorly organized or lacking as you rightly imply many are.)
Bravo and thank you for sharing it.
This post is a wonderful review of what we ALREADY KNOW!
I'm now re-reading some of those links back to Simon's work...
This quote, from near the end of your post:
"If you were designing a learning system from scratch, you would not copy human learning wholesale. You would correct for its limitations: our working memory constraints, our susceptibility to distraction, our tendency toward the path of least resistance."
This is ONE of my conclusions from your post = it seems that there is some potential there for AI to benefit students' learning, however, there is still the issue of "distraction" in the online world, and it doesn't seem that the AI models have managed to NOT be distracted? Maybe I'm wrong?
Also, regardless of AI, classrooms in schools are one environment that teachers have some potential to support learning through this lens - classrooms are the "educators' beaches" and THAT is within the control of classroom teachers who have some knowledge of evidence-based research. It's a pity that more teachers don't know about posts like this?
Thanks again for supporting MY Personal Learning, and I'll try to complete my "reading homework" so I can learn more! MUCH appreciated! :)
So - you want me to message you on WhatsApp?
Re: "The best curricula, like the best watches, are designed so that component skills can be mastered to the point of automaticity before being combined." We use subassemblies and use them like the watchmaker, but they can be more varied than this article appears to concede. We mimicked other assembly processes. Sub-assemblies also can be situated in a select part of a larger, complex problem that is interesting to students, particularly the older ones we serve. I may have misinterpreted something, but we have found such absolutes do not work very well across all types of knowledge in a domain. Learners need much more practice with concepts than most curricula afford, making automaticity a further destination. In some cases, combinations can actually support some moves to automaticity of sub-assemblies. It becomes a step along the journey to fluency. This is more complex to "hold" in the design, but I'm glad we remained open to that possibility.
Simon's ant was also the inspiration for the behavior based robotics of Rodney Brooks, who recognized the role that environmental contingencies play in real time adaptive behavior. Although perhaps not recognized at the time by Brooks, his insight was a rediscovery of Skinner's behavior analysis. Skinner conceptualized moment-to- moment behavior change to be a variation-selection-retention process similar to biological evolution, but representing selection of behavior, not allele populations. This insight led to the development of programmed instruction, a form of explicit instruction that preceded today's AI based programs. Interestingly, today's AI technology often deploys a form of machine learning called Deep Reinforcement Learning. Good luck finding any references to Skinner and other early learning literature in the AI scholarship!
near decomposability is a fascinating concept that i will definitely try to keep in mind when designing anything grander than a single lesson. but i do want to add two things.
first, its good to be aware of naturalist tendencies and not confuse natural with good learning environment. but ofc we also should not forget, for what environment our brains, bodies and minds developed and expect learners to follow our nicely designed curriculum, just bc its nearly decomposable. just to say that motivation is another crucial factor and that often comes from seeing a big picture and understanding the importance of the learned through its relationships to other and concepts and abilities. if we hammer down the basics and only allow the students to see what comes next if they mastered it, we don’t teach children how to think, but how to follow menial learning tasks. we rob them of an intrinsic motivation of exploring and understand the world - which is huge and complex and cant be understood all at once. and i agree, if you tried that, you would drown in information. all im saying is that near decomposability is only part of how we should think about learning. yes, we need to make sure we get the foundation and the first floor before the second one can be build. but we also need to know how the building is supposed to look in the end and why we want to build it to begin with. otherwise the second floor wont match the first, its gonna be crooked or we will just stop and abandon the house unfinished because we lost interest.
Your articles are fantastic, Carl. Your distillation of educational concepts is very helpful.
Lately in my teaching, I have been feeling that lesson structure is an underrated component of teaching, and this article helps clarify that, so thank you!
Such an insightful read, thank you!!
Excellent read. Thanks Carl.
It's ironic that educational researchers of the 80's and 90's took the wrong lesson from the ant metaphor. Simon's key point (and your point) is that if behaviour is highly determined by environmental conditions, then we have to be really careful when we design learning environments. Unfortunately, researchers-of-the-day used the metaphor to argue the precise opposite. They felt the ant story illustrated how cognition was primarily situated, which in turn meant that learners should be immersed in complex, real-life scenarios. (Of course, in making this argument, they had to conveniently ignore the very real cognitive limits of learners).
Thank you for writing a wonderful piece that calls that mistake out.