One thing I never anticipated about being a parent was how many conversations would revolve around school. What are the schools like near our neighborhood? How much should I contribute to a 529 plan?

Right now the answer is a little unclear, but I’m leaning towards “I don’t care” and “$0.” We’re planning to homeschool.

This, but every day.

I don’t have incredibly strong views on the relative merits of different educational options, because the data’s really bad. Most of what people “know” about education is either

a) poorly-controlled studies (Los Alamos high does pretty well; I wonder if being located in the town with the most PhDs per capita has anything to do with it!)
b) studies with tiny sample sizes, or
c) stuff where the studies show minimal or zero effects but the idea sounds good. Small class sizes, for example, purportedly explain why Finland does so well, but Singapore has huge classes and is routinely number one in international rankings.

It’s incredibly hard to do meaningful randomized controlled trials in education, because you can’t control what happens at home and controlling for confounders is just depressing.

There’s a roughly annual tradition of 1) a big news story highlighting some school that has achieved really impressive results, and 2) the inevitable story about how a prior year’s big news story was actually a fraud. When I talk to people about the non-academic benefits of traditional schooling, they like to cite stuff like “socialization.” This is true. School socializes kids for an environment in which all goals are known at the outset, all their time is scheduled in advance, and the people in charge are selected mostly for conformity and conscientiousness. I would find the socializing argument very compelling if my big concern about my kids was that, as adults, they’d have trouble adjusting to prison.[1]

I do have some confidence that traditional schooling, public and private, is not a great system to participate in for practical and moral reasons. Practically, the educational system experiences negative productivity growth over time. Primary school educational outcomes in the US haven’t improved in over a generation, but real spending per pupil keeps rising. If we could somehow rediscover cutting-age 1970s-era teaching technology, we’d free up tens of billions of dollars per year. There are many arguments for why this is, from Baumol’s Cost Disease to the Evil Unions Theory to the Evil Textbook Cartel Theory. I blame a combination of all of them: nobody wants to economize on education, most people aren’t spending their own money, and the US is a rich country. Combine these factors, and the result is that prices automatically ratchet up over time, without commensurate increases in quality.

The gold standard for measuring the impact of policies is the randomized controlled trial, but an RCT just contributes to the ratcheting effect: you can test out a new policy that might lead to an improvement, but you can’t test out a cost cut and see if it doesn’t worsen impacts — nobody wants their kids to be in the getting-less-than-last-year cohort.

Basically, I have no confidence in anyone’s opinion about whether any classroom practice or government policy has any measurable effect whatsoever on educational outcomes. If you want to have an exceptional record of prognostication in education, here’s a secret: wait for someone else to express a strong opinion, and then say “I doubt it.”

If educational outcomes themselves aren’t broadly amenable to change, what is?

Well, if you’re not pre-committed to a model where one teacher lectures 15–40 kids, teaching them a particular curriculum in a prescribed order, you can do a couple things to make school suck less:

Eliminate age segregation: Unless you enlist in the military, you will never be in as age-stratified an environment as you’re in during school. School teaches you to assume that the people you interact with day-to-day have a very similar cultural background to you: same slang, same movies/TV shows, same clothes, etc. The first time you get a real job and work with people who have kids your age will be quite a culture shock. Getting rid of age segregation also helps with ego: if you’re the smartest ten-year-old and your peers are all ten, you feel like there’s a pretty good chance you’re the smartest person in history. But if being a smart ten-year-old just means you’re asked to do the age-twelve reading assignment, you get to be challenged again.
Don’t bother setting alarms: Some teenagers can wake up at 5am, but a lot of them seem to prefer 10. In theory high school teaches you to wake up on time, which is a valuable skill in the workplace (or at least a pretty humiliating reason to get fired). In practice, work teaches you to wake up on time; when I stopped working at Yahoo and started working at a hedge fund, I seamlessly switched from waking up after 9 to waking up before 5. Except on weekends, when I was back to teenager schedule. Somehow, since having kids I don’t think I’ve slept past eight.
Focus on open-ended projects: School assignments have to get graded, so they need to be circumscribed. This is especially perverse for anything uncertain. You can’t grade an essay based on whether a student found something true and original to say about The Taming of the Shrew, because over the last few centuries we’ve mined every iota of insight out of Shakespeare. And it’s hard to let students choose something new, lest you spend a month grading essays on The Hunger Games. So teachers are forced to grade based on measurable criteria: citing sources, spelling well, hitting the page count without doing anything too egregious to margins, adhering to the standard beginning-middle-end topic-supporting sentences-transition/conclusion Sierpinski Triangle of essay construction. You wouldn’t want to read someone else’s kid’s five pages on Straussian readings of Captain Underpants, but I would be motivated to read something like that written by my own kid.
Teach Internet research: One of the shocking things I learned in the workforce is that digital natives don’t necessarily understand Google. I think it’s because that’s a hard topic to grade. The way knowledge workers use Google is to save themselves three hours of effort by finding someone else who has solved a problem already, or at least solved a close-enough approximation. In elementary school the term for this kind of research is cheating.
Assuming coding knowledge the way you assume basic literacy and numeracy: history classes don’t have prerequisites like “Must be able to write in complete sentences” or “Must be able to subtract 1618 from 1648 to determine the length of the N Years War.” It’s just assumed. Similarly, a homsechooling unit on macroeconomics should assume that any reasonably-educated person can fire up the FRED API and download a time series on historical GDP growth.
Satisfice on all subjects, maximize on a few: Most intellectually-curious adults I know have a limited number of obsessions and a huge number of things they know but barely care about. Adults know enough about fractions to triple the amount of flour in a recipe calling for half a cup, but most of them don’t study math. The ones who do number theory for fun know enough civics to register to vote, if that’s their thing. Given how much information people forget after leaving school, it’s a safe bet that many semesters could be compressed into a couple hours, which frees up a lot of time for intrinsically motivating stuff. And some of it will stick! Diagramming sentences is the canonical “you’ll never use this in real life” skill, but I do use it in real life: my wife have a long-standing argument over whether or not it’s proper to say things like “There’s a sandwich in the fridge, if you’re hungry.” Diagramming sentences didn’t actually help me make my point, it was just a gambit in our who-can-overthink-this duel, to be countered, naturally, with references to Grice’s Maxims.

To summarize, you can make childhood education more like adult education. The adults I know who keep learning don’t subject themselves to equal doses of English Lit, Poli Sci, and Algebra. The adult self-designed curriculum looks a little more like this:

Scala 101 through Scala 400. Class materials: a couple tutorials, maybe an O’Reilly book.
Medieval French History. Course project: a historical novel set in medieval Provence.
Intro to Japanese Culture, culminating in a field trip to Japan. (Or culminating in the viewing of approximately 7,000 hours of un-dubbed anime.)
PE! Fitness goal: deadlift more than last week.

Adults who like learning tend to have one or two long-term obsessions, and a rotation of week- to month-long projects. Right now I’m brushing up on linear algebra and Python through Coding the Matrix. My wife is reading 100 books about death. We both have customized curricula, tailored to our interests, and we can drop classes at any time.

Who Are You To Talk?

Since nobody has good data on education, debates about education turn ad hominem pretty fast. The two effective ad hominem attacks in education are:

You’re telling people to do the weird thing you did, but for all you know you turned out okay despite your weird path, or;
You’re telling people to do something weird, even though you yourself didn’t do it.

In my case, our bets are hedged: I have no degree, although I did end up getting a nice post-college credential. My wife has half of a PhD in philosophy of education from Columbia, so she not only has paper credentials, but she knows where the ed policy bodies are buried.

(Everywhere. Ed policy is like a catacomb, basically.)

More importantly, we’re nerds, and we’re raising our kids to think that reading a lot, writing a lot, building stuff, and discussing academic topics is perfectly normal behavior. My only real concern is that they won’t get much exposure to non-software engineering and biology — but we will find a way. Maybe I’ll just disassemble all of their electronics periodically. For some of the noisier electronics, I’m already tempted to disassemble the with a hammer.

The Plan

I’ll start with the fallback plan: the fallback plan is to enroll my kids in public school. Our district is probably fine, although I’ve literally never checked. Assuming it’s not, we have some time; the oldest turns three in a couple months. If we’re being idiots here, we’ll admit it fast and just go back to normal; cutting losses is easy once you get in the right headspace. Even if we commit to homsechooling and turn out to be wrong, the losses are controlled: imagine what a sick college admissions essay it will be to tell the story of how your parents homsechooled you even though they’re not crazy hippies or survivalists!

I don’t view public school as a good option. A parent-friend of ours’ main gripe about local public schools is how often the teacher just has the class watch TV. Not educational TV, just whatever’s on. Public school is a sort of educational Freeganism, although in this version there’s a several hundred-billion dollar government program to stock dumpsters with the right goodies.

But assuming we don’t have to do that, the curriculum will look like this:

Suggest books to read.
Talk to them about the books.
Encourage them to write about the books.
Do some math worksheets. Watch some Khan Academy.
Do some more hard-core math; watch some OCW lectures.
Code, code code. (Little Claudia has already been introduced to Logo, and thinks it’s awesome. Next stop: TAOCP!) More seriously, I’ll probably do Logo, Smalltalk, Python, and Excel (Excel will come after Python so the kids don’t end up like me: good enough at Excel that I wait too long to implement something in a real language). If they’re interested, we’ll move on to Scheme and Haskell, but I think it’s okay for nontechnical people not to know those.

The cool thing about reading is that the verdict of history is pretty reliable; while most of the world’s good books are probably still expensive, most of the books we know are good — the ones that appealed to several generations in a row — are on Gutenberg.org.

Education has incredible compounding effects, at at least three levels:

In many fields, you learn to build abstractions on top of what you previously treated as concrete. That’s most obvious in math, of course: counting -> addition -> multiplication -> exponentiation; arithmetic -> algebra -> calculus; etc. But it works in fields like history and literature, too; it’s really fun to read quality mass-market fiction (I love Stephen King) with an eye to reverse-engineering what the author is trying to get you to think.
You find connections within and across fields. Good books merit rereading; the only way to understand history is to put it in the context of more history; and you can find tons and tons of connections between science and history (read The Prize and then reread any history book about the twentieth century if you don’t believe me).
Combine points one and two, and add some general scholarly phronesis and you start to pick up new topics a lot faster. This has some dangers — in programming, there’s the classic “Writing C++ in Python” problem, in finance there’s a tendency to look at every crisis as analogous to the coolest crisis you know a lot about — but it’s a force-multiplier.

We’re excited about this. School is stuck in a nineteenth-century model of pedagogy, funded by twentieth-century assumptions about the nature and magnitude of economic growth. But it’s the twenty-first century, and we’re teaching our kids accordingly.