Reforming American Schools
Trying to redesign the pre-college education system.
Created with DALL-E 3. Prompt: “fantasy school.”
01. What’s wrong
You should read at least the intro and TL;DR of this piece to understand this post.
The American education system frustrates me a lot. So many people genuinely care, but Goodhart’s law has taken over everything and outdated norms have become ossified.
Over the past few months, I’ve been brainstorming how I would improve pre-college education if I had complete authority (and the education system was a perfectly functioning institution). For this post, I’ll focus on 3 of the issues I care about most.
The education system should focus more on building a resilient citizenry well-suited to democracy.
Most people leave school with little sparkliness (“roughly, the mix of intelligence, curiosity, and agency”).1
The education system is bad at helping students build long-term understanding and skills that transfer.
Disclaimers: (1) This post is a v1 based on a combination of empirical observations and some (not enough) research. I decided to send this out first and receive general feedback before diving deeper into the literature. (2) The beginning is somewhat rough around the edges but the post gets better as it goes along. (3) Here, I’m focusing on iterating on the existing education system instead of redesigning it from first principles. (4) I think you miss out on a lot if you don’t check out the footnotes, so please do.
02. Democracy
“The best argument against Democracy is a five-minute conversation with the average voter.” - Winston Churchill (but not really)
This is what Trump’s official campaign website looks like at the time of this writing. Here’s a promotional video made by Biden.
It’s hard to find anything even vaguely resembling real policy in either. The strategy is clearly to demonize the other side, encourage tribalism, and emphasize that the election is an existential struggle between incompatible cultures.
Regardless of your political opinions, I think the fact that this is normal should be incredibly scary, and point towards some of the failures of our democracy.
Our democracy is incredibly vulnerable to tribalism.
Our democracy consistently makes poor decisions.
Our democracy is myopic.
All of these issues are downstream of the voters. Changing a few words can help provide a nice perspective.
People are incredibly prone to tribalism.
People consistently make poor decisions.
People are myopic.
I’m bullish on the idea that thoughtful reform with curricula optimized for the transfer from classroom learning to “real world” thinking, can help make voters at least somewhat better on average. This is probably one of the most critical things for democracy as a whole, so even small improvements across the board would likely be worth the opportunity costs. I see 3 core objectives for any educational reform here.
Give students a strong foundational knowledge of important activities of states.
Economics
Foreign policy
Help students deeply internalize epistemic humility and how imperfect and contingent their values and beliefs are.
Encourage students to think for themselves.
Teach them to recognize and avoid groupthink and propaganda
Teach them to critically analyze convincing work
Implemented well, investment here has large rewards. Better voters choose better leaders and keep them more accountable. Good leaders kept accountable choose better policies.
03. Sparkliness
Education is the non-coercive rearrangement of desires. - Gayatri Spivak
People form substantial parts of their identity in the education system, so we might as well be more deliberate about the traits we choose to cultivate.2 The vague space of traits encapsulated by sparkliness are my best guess for the meta-level traits that grant the highest chances of leading fulfilling lives.3
More practically, if you evaluate education by returns on investment, sparkly people are the ones who become great researchers, writers, innovators, and professors. They’re the type of people to think hard about the world, care a whole lot when things are broken, and fix them.
What sucks is that most schools aren’t conducive environments for sparkliness. Some combination of social pressure (probably caused by high levels of insecurity), fake work, and boring curricula prevent curiosity and agency in every school I’ve been to. Even the “prestigious” schools I’ve gone to or have friends at are just more fake work that’s ideally counteracted by more passionate teachers, less boring curricula, and more sparkly friends.
When I try to brainstorm environments that encourage curiosity and agency, there are a lot of factors I can describe that seem ideal. Malleable schedules that adapt to intellectual whims. Most people doing and learning interesting things by default. An enabling environment that offers room for meaningful personal growth and action. That all sounds great in the abstract, but from my experience, this doesn’t work nearly as well as one would hope without having a seed population saturated with intelligent, curious, and agentic people. On the other hand, when you get a lot of sparkly people together, the environment engineering is just a bonus. They’re probably going to do awesome things anyway.
My only solution to this is to be really deliberate about engineering elementary and middle school environments to scaffold and nurture curiosity and agency from a young age. Ideally, 1.5x as many people are already well along the way by the time they reach high school.
To be incredibly explicit for the sections ahead, the primary goal here is to create an environment where the median student is a fair deal (wow how quantitative) more intelligent, curious, and agentic and the secondary goal is to create an environment where the students predisposed to sparkliness thrive even more.4 You should evaluate all of the “proposals” I will make based on that.
04. On learning
Students forget substantial amounts of what they learn. The feel-good fallback is that schools “teach you how to think” through general problem-solving skills. After some surface-level research, this seems to be easier said than done.5
Spaced repetition systems offer a fairly simple fix for the learning issue (I won’t explain more; see the footnote).6 When used well, spaced repetition systems help students remember what they learn better, often enable greater understanding, and potentially lower stress in exchange for 20-30 minutes daily. For a while, it genuinely puzzled me why more schools (and people) haven’t adopted them, but Andy Matuschak has some good answers (see “Barriers to adoption”).
To increase the chances of long-term memory, there’s also an easy way to ensure that students don’t just delete their cards as soon as the final is over. We can have low-stakes tests for subjects years later that factor into the grade for the original class (like a statistics test 2 years after you took the statistics class).
It’s helpful to think of the “teach you how to think” issue as a symptom of low transfer of classroom learning to the desired environment. Here, I think there’s a lot of low-hanging fruit in teachers and curricula just being more deliberate about the types of transfer they want (building up for a future class, getting better at identifying X in the real world, etc) and optimizing for them.7 The stereotypical “but what’s the point of this” question that irritates teachers to no end is valid here.
Neel Nanda has a good post with more on teaching, but there are some traits I want to emphasize about the best teachers I’ve had. They all happened to be STEM teachers, but I hope something useful should still generalize.
Broadly they teach like some mix of Andrej Karpathy in his course, Jeremy Howard in fast.ai, and 3Blue1Brown.
Their teaching style often matches fast.ai’s top-down philosophy.
They make learning whole.
They focus on intuition → For most STEM classes, I want to have the feeling that I could’ve invented the concepts I’m learning about if I had thought about the right things in the right way at the right time.
Specifically, they build intuition like 3Blue1Brown. It’s hard to pin down exactly what makes 3Blue1Brown good at intuition-building8, but I can point to some examples of great pedagogy (neural networks, linear algebra, calculus, bitcoin).
Importantly, the focus on intuition isn’t at the expense of actually diving deep.
04. Ideals for elementary and early middle school
I believe the key goals of elementary school should be to 1) build the core foundational knowledge that students need and 2) nurture sparkliness.
We can get good mileage here by changing a lot of established school norms.
We don’t need to fill extra time.
On a daily basis: New subjects and lessons within subjects should only be added insofar as they directly serve one of the two key goals. I think we should look at each additional subject/lesson as “taking” from free time instead of the other way around.
In elementary school, I think this extra time should be divided 60/40. Kids should get the majority of the free time to do whatever they want. During the remaining 40% they should be (gently) mandated to do something “productive”. They can choose between options like reading9 , learning more about a subject that interests them10, or writing a story. Students doing especially cool things should be shouted-out (in an incredibly thoughtful way). Teachers should also gently encourage children to continue at home, and keep in touch with their students’ mini-projects. Some of this extra time could also be open for teachers to teach something exciting outside the curriculum. The important part is for the kids to feel like they have agency. This time is where I hope to sow the seeds of sparkliness.
On a long-term basis: A lot of classes feel like they take a year just because they have to take a year. This probably wastes hundreds of hours worth of time per student.The principle above also applies to homework. Homework should only be assigned if it helps cement what was learned, aid in the transfer process, and/or push at the boundaries of their knowledge.
Teachers should try their hardest not to make school feel routine and formulaic. We can do simple things like not having a fixed time slot for each subject (with the teacher choosing a random order instead) and switching away from the grids of desks that are standard in most classrooms.
Ability grouping should be the norm, starting from the 4th grade.11 10-year-olds who are particularly good at math but struggle with reading might be in a “seventh-grade” math class and a “fourth-grade” reading class. I don’t think class names should even be associated with grades and “seventh-grade math” should be called something like “Level 7” math instead. If students show exceptional interest in something, they should be guided towards accelerating through the prerequisites for that field.
For some reason, most elementary schools I know of mostly only have visits by authors. That’s great, but it should be just as common for there to be visits by accomplished software engineers, physicists, and economists that students can sign up to attend. Being surrounded by accomplished and agentic people can lead to big unlocks in what they think is possible.
Importantly, you need really good teachers. They should be visibly and infectiously intellectually curious, encourage high engagement in their students, and teach in an effective way. They should encourage asking questions, and give their students credit by giving them answers that encourage more questions. As a friend related, when he asked about colors as a child, his parents “would properly go into absorption/emission spectra, electron orbital confinement, etc” and let him ask follow up questions about the parts he didn’t know.
For me, this means that K-3 education looks a lot like it does now. Kids learn basic reading, writing, and math and take classes like art and music. Starting from 4th grade, ability grouping starts kicking in, and those “extra” subjects (as well as an increasing number of others) become classes that students can choose to trade their spare time to take.
We can use the idea of spaced repetition to inspire techniques like frequent low-stakes pop quizzes during elementary school. As students enter middle school they can be encouraged (but not forced!) and taught to use spaced repetition systems. Between ability grouping, SR-inspired techniques, and an emphasis on curricula being structured more deliberately, I’m fairly optimistic that we can make the overall progression of learning from K-8 at least a year faster on average for each required subject without making them less manageable (and I feel like this is lowballing it), which makes room for some of my ideals for high school.
05. Ideals for late middle and high school
The hope is that by this point you have a somewhat more curious student body. Now, the emphasis is on enabling them to learn and do interesting things.
Continuing with the trend, I think the average student’s schedule should become gradually more flexible as they progress, and they should be mostly open by ~sophomore year (I think you can make this happen by moving many required classes and prerequisites to middle school12). And again, I don’t think that students should necessarily be forced to fill in those open time slots with other classes. If there’s a class they want to take, they should be able to take that class (and if they don’t have the required prerequisites, the onus is on them to figure that out by themselves). If they want time to catch up on work or do cool projects (learning in an authentic context is underrated), they should be allowed to do that. If they feel like there isn’t anything interesting they want to fill a particular time block with, they should be allowed to just go outside or to the gym. I can only think of a few restrictions I’d want to be in place.13
School should also become more flexible and adaptive to intellectual whims. If I’ve become obsessed with computer science, I would want to be able to double up on computer science classes for a semester to explore that interest. If I have a strong side interest in microeconomics, but I don’t want it to take up too much of my time, I would like to be able to take a 2 hour/week class for a semester that would give me a broad overview.
How does this work?
I think schedules can be divided into 40-minute blocks as the primary unit of time, where different classes can take up different numbers of blocks.
I suspect that most students will still end up taking similar classes, so common courses will look much like the standard model. Courses that are uncommon enough could be turned into high-quality structured online courses made by a central organization (basically Khan Academy++).
The school can identify groups of students who are taking the same/similar uncommon courses and put them into classrooms with teachers knowledgeable enough about the topics that they can help facilitate and supervise the learning.14
Most software education systems right now are “chocolate-covered broccoli”, but if thoughtfully designed for each course by people with domain knowledge, and integrated with spaced repetition systems, I’m pretty optimistic.
06. Civics++
"[Decision makers] will make better choices when they trust their critics to be sophisticated and fair, and when they expect their decision to be judged by how it was made, not only by how it turned out." - Daniel Kahneman, Thinking, Fast and Slow (emphasis my own)
At least in my experience, civics courses aren’t broad enough in scope. They end at teaching how the American government works when that should really just be the start. We need more voters to have solid levels of critical thinking in the areas that are generally important to evaluating candidates and their policies.
Especially in this course, the goal shouldn’t be to train students to recognize and parrot the right thing to say on some standardized exam but to internalize helpful models of thinking about the world. This course requires practical exercises and examples, where each topic isn’t something you move through before going to the next, but something you continuously revisit. Most of the course should simulate “real world” critical analysis and decision-making.
I think this would work best as a non-time-consuming required class (~2 hrs/wk) that gradually builds and reinforces these skills starting from a young age.
Basic Curriculum (suggestions on how topics should be taught in italics):
Good understanding of the fundamental activities of modern states
Foreign policy
Success criteria: voters should be able to understand why countries make specific decisions and have a surface-level understanding of the implications of different policies.
A deep internalizing of epistemic humility
Beliefs, morality, and values are honestly pretty stochastic and are often the product of circumstances and social incentives.
I think this understanding crystalized for me when I started analyzing people’s opinions in the context of their lives, and hopefully that should help other people as well. For a fun example of what I mean, Leopold Aschebrenner’s views on geopolitics make a lot of sense in context with his broader life story.
A special part of the curriculum should just be going through something like Thinking, Fast and Slow (although with care). Getting up close and personal with the overwhelming evidence about your irrationality and how stochastic a lot of your decisions and beliefs are is one of the best ways to internalize humility.
Another special part of the curriculum should be going through something like Conspicuous Cognition. He has a lot of great work on beliefs, and I think general awareness would go a long way in terms of epistemic humility.
How to update your beliefs.
Understanding when to update your beliefs and when to hold your ground.
Learning to argue for your beliefs while still being able to change them.
Practices like automatically searching for and engaging with other perspectives whenever you form opinions about something.
Encouraging a search for truth as a general life goal over adopting a “soldier mindset” for any individual belief.
Critically analyzing and engaging in debates on complex issues should help with this. I don’t mean debates in the usual sense (and think that they’re overrated). I generally prefer slow debates to fast ones, and think that the idea of “winning” a debate is usually pretty messed up.
Doing adversarial collaborations would be really cool.
Being comfortable with uncertainty
Superforecasting + learning how to make decisions under uncertainty
Merely practicing forecasting should help with this. Quantified Intuitions and the introductory exercises in Metaculus are also great for humility.
Thinking for yourself
Basic critical thinking skills around statistics, ads, media, propaganda
How to interpret and use research to form opinions
Critically analyze convincing work. Identify points where you agree, points where you disagree, points that are internally consistent, and gaping holes. Identify the crux instead of having a knee-jerk general “how could you even think like that” reaction.
Understand regression to the mean. Thinking, Fast and Slow has a great chapter on it that is nicely summarized here.
Going through lots of examples is best here. Play out things like the Isreali Air Force scene. Regression to the mean is incredibly counter-intuitive (because our brains are designed to be good at creating stories), so it’s important to illustrate well.
Other basic skills
Understand memetics
Being cautious about updating after dramatic events
Recognizing and avoiding groupthink, mob behavior, extremist ideology
”But there is a step beyond thinking of yourself as x but tolerating y: not even to consider yourself an x. The more labels you have for yourself, the dumber they make you.”
07. More practically
So far, I’ve focused on ideals, but there are very real bottlenecks.
Meta-level: These kinds of things are usually too optimistic. Most of this sounds ideal in the abstract but can easily go sideways unless implemented well. There will be tons of tradeoffs and concessions that have to be made.
I’m probably too bullish on the civics course. Many students probably won’t even care enough to pay attention in class.
To make all of this work, teachers (especially elementary school ones) and curriculum designers need to genuinely care.
I am actually pretty optimistic about this. Many people talk about intelligent and curious children being beaten down by the education system. Fewer people talk about passionate teachers being beaten down by students who sleep through class, copy all the homework, and cheat on tests. Seeing more students genuinely curious about what they teach outside of grade optimization might be enough to reverse that.
Evaluating teachers and schools becomes harder when the objective is to make students more curious and agentic.
Making younger children more sparkly is probably harder than “give them free time + just let them ask questions and give them answers that lead to more questions.”
Students will game the “do whatever you want” high school environment to not take any stimulating classes at all or take even more classes they aren’t interested in because of external pressures.
For the first: if they didn’t want to take those classes anyways and those classes likely won’t affect their future much, is that really so bad?
For the second: I really don’t know what the solution is.
Despite all of those bottlenecks (and the fact that any large-scale education reform looks infeasible in today’s political climate), I think there are still tons of small takeaways that can be applied to improve the education system.
The education system should take the job of making better voters more seriously.
Infectiously curious and intelligent elementary and middle school teachers are underrated and we need to figure out how to get more of them into our schools. The trajectory change that a teacher who encourages curiosity and agency can give a child is massive. I can attest to this from experience. The trajectory I’m on currently (while not insanely sparkly by any means) probably would not have existed if a 5th-grade teacher I looked up to hadn’t made an offhand comment encouraging me to start programming. Imagine how many lives could be changed in an environment that deliberately tried to set up “Road to Damascus” moments like that.
The default is for most classes not to be deliberate about intuition, understanding, tacit knowledge, and transfer.
With practically no investment, a few thoughtful changes, and a ~equal amount of time studying on the student’s behalf, schools can make it much easier for students to remember the content they learn for longer.
The reforms I describe might not work on a large scale, but I think this piece could be seen as a starting point for any new schools that want to be better at these issues.
Since this section is titled “More Practically”, it would probably be the best place to discuss something Ben Kuhn brought up in commentary on this post: “One hypothesis is that it's impossible to reliably get a government-run institution to execute at a high level of quality (see also: RMVs, vaccine distribution, most public transit systems, etc.), in which case the best concrete paths to reform would probably focus on school choice15 rather than rolling out specific program changes.” Although this wouldn’t help much with the democracy and sparkle issues by itself (which were mostly idealism on my part), more school choice should exist and would force schools to get better at teaching and ensuring they provide real value to their “users”.16 This would be concrete reform with potentially bipartisan support, relatively measurable outcomes, and a much clearer value add.
08. Questions
As I said earlier, this is a v1, and I’m looking for feedback. If you have answers to any of these questions, reach out at vihaansondhi07@gmail.com.
How realistic is the idea that a long-running, well-implemented civics course would have a nontrivial positive impact on voter ability and reasoning?
How do you make ability grouping work?
How do you ensure that parents don’t force children to speed run through classes against their wishes?
How do you ensure kids don’t compare themselves to each other and feel pressure to catch up if they’re behind?
What are other ways we can make elementary and early middle school better environments for developing sparkliness?
How do we get more infectiously intelligent, curious, and agentic people to willingly be elementary and middle school teachers?
How do you make sure that teachers optimize for what the “point” (the transfer they want) is, and focus on understanding over pattern-matching and teaching to the test?
What metrics could you use for testing how good a school is at encouraging curiosity and agency? How do you make those metrics resilient to Goodhart’s law?
What would you add to the Civics++ course?
Do you have any directions towards interesting research and general literature on education and learning?
In what ways could students game this system to make things worse for everyone involved?
Are there any other ways that my ideas seem naive or excessively idealistic?
What were your best learning experiences (in or out of a traditional classroom setting)? What were your worst?
If you have extensive experience with spaced repetition, and you use it in interesting ways, reach out! I would love to share decks and exchange tips. Memory system culture is great.
I know my posting frequency has lowered significantly, but that’s mainly because I’ve recently been doing most of my writing here instead. I have a ton of things I want to write about, and my frequency will probably go up to ~once a week over the summer.
(Note from future me: this was a dominant part of my vocabulary until I read this. Now I try to be much more mindful about using it.)
To feed your brain’s associative machinery, these are the kind of people who ‘[learn] category theory together” at 5 am, spend weeks going through math textbooks because of sheer curiosity, take online courses and work on cool projects in their fun time. It’s an ideal, and it’s one that I wish more people moved towards.
If this seems too paternalistic or “Big Brother”-y, my argument is basically that most 3rd graders aren’t actively choosing their self-identities. It’s a messy stochastic process determined by things like the friends they happen to meet first and the activity they first do during choice time. I think that as long as you’re encouraging the values that are your genuine best guess for what would make those children happier and more fulfilled, it’s probably for the better.
I have no real evidence for this beyond the fact that finding some of these people and becoming more like them has completely changed my life trajectory, the way I think about the world, and the way I view myself. I feel much happier now than I did a year ago.
In general, I would find it a beautiful thing if we could reduce the number of “People who feel trapped in school despite wanting to learn, or who are interested in lots of subjects but don't know what to do with themselves. People who are really smart, but have been beaten down and don't have the sense of self-efficacy to advocate for themselves. People who don't have the space in their lives to build their own projects because they go to school in the day and work to support their family at night. People who feel like they don't have friends to share their interests with.”
If you’re wondering why this even needs to be stated, see section 4 of this great essay.
I’m very interested in pointers to good meta-analyses and general research on this.
If you don’t know what a spaced repetition system is, I would recommend checking out this great post by Michael Nielsen (or this if you want to rabbit hole). If you already know what spaced repetition systems are, but you’re only familiar with using them for pure memorization and not necessarily understanding, see the “Using Anki to thoroughly read a research paper…” section in the Michael Nielsen post I linked above, Using SRS to see through a piece of mathematics, Andy Matuschak’s post on using spaced repetition to create understanding, and/or their combined rejection of some of the standard critiques of memorization.
If you don’t feel like looking at any of these right now and just want to get on with this post, I’ll give a quick rundown. If you’ve ever had the feeling of learning something and not knowing whether it was going to stick, this is a (thoroughly research-backed) way to mostly fix that. When used like a virtuoso, it can be brilliant for going beyond “just memorization” and moving towards deep understanding.
If you become sufficiently memory system-pilled and want to talk more about them, reach out to either me (vihaansondhi07@gmail.com) or Jacob (who’s probably thought about it way more than I have)!
If the class is meant to serve as a prerequisite for a future course, the class should clearly be structured so that the most attention is given to the topics that are most important for that course. Those topics should be reinforced several times throughout the year. If the class is primarily meant to enhance critical thinking skills, you want to emphasize intuition and tacit knowledge through consistent application in new problems.
Some ideas:
All of the above (making learning whole + intuition)
A good understanding of the experience of being a beginner in the subject. He always seems to know what will immediately click and what will need more explanation.
He’s clearly thought a lot about what makes people engage with math.
It’s easy to tell how beautiful he finds math, and some of that rubs off.
Sidenote: I highly agree with a quote from this post: “I do think literary classics are classics for a reason, and ‘the canon’ is important. But you won’t get people to genuinely appreciate it by forcing them to read the Great Gatsby; if you want to inspire a love of reading, you don’t do that by shoving incomprehensible and boring books down students’ throats. Let them start reading for fun — for the plot, for the characters, for the thrill of finding out what happens next. After they start appreciating books, start teaching them more complex books (and for the love of God, stop being snobs about what constitutes ‘high literature’).)”
I also think that this is true the other way. From painful personal experience, I strongly encourage not stopping children from reading books that are slightly beyond them (unless grossly inappropriate) and ideally even including these books in classroom libraries.
Some form of genuinely effective tutoring by students in the grades above would be great. With some time spent on learning how to teach and potentially some incentives to teaching well, upper middle and high schoolers could probably teach small groups of elementary and lower middle schoolers well enough.
While that ideally sounds great, there are negative effects that should be considered.
We should also give serious reconsideration to which subjects are required and for how long. Why aren’t economics or computer science traditionally multi-year required subjects while others are? Of course, you can make a serious argument for any subject, and I wish we could introduce students in-depth to all of them, but we seriously have to prioritize.
Restrictions on phones during “free” blocks.
There should probably be a low minimum number of classes you must take per year.
One required “project” study hall where students are constantly learning new things or working on projects* with no real guidelines/restrictions/expectations, and keeping a teacher updated. My rationale for this being required is that many students might feel like they have to take classes all the time at the expense of being able to work on projects. Even people who feel forced into this class probably have a higher chance of ending up becoming passionate about something than another class they’re taking just for the sake of taking more classes.
*I.e. exploring a research question PhD-style, building a robot, working on a programming project, writing a blog post, filming a short movie, making a podcast, organizing an event, application of knowledge from a class.
Unrelated to restrictions: Riffing off the last bullet point, it would be cool to have school-sponsored clubs (that actually work well) that are basically forums for researching and discussing questions, where the goal is a collaborative search for truth. Basically like a debate club where you don’t pick one side in advance and aren’t solely rewarded for convincing other people that you’re right. Some form of extra credit could be a good incentive for attendees with genuine contributions.
An interesting consequence is that this system is probably somewhat more fun for teachers as well. There would be less standing in front of a whiteboard giving the same lecture 5 times in a row and more room to learn and grow naturally alongside the students whose learning they’re facilitating.
School choice would seem to provide benefits a priori (competition tends to lead to higher quality), and at a first glance at the research this seems to be true, at least for things like standardized test scores (counter-example).
Not endorsing any political views; I just think the principle on education is something that I agree with quite strongly.