If you disagree with the ideas here, I’d be very happy to receive criticism and engage with other viewpoints.]
This essay starts off with the introduction.
Then there’s two big parts: one on competition, and one on metrics.
The two big parts are followed by a small part that ties them together.
Next there’s another small part on why some solutions suck.
This is followed by my own sucky solutions.
Then there’s the tldr; (You can also jump here if you’re pressed for time, although I really would prefer you read the whole thing…)
To conclude, we have the conclusion.
[Thanks to Rebecca Baron, Julian D’Costa, and Roberts Džeriņš for their help editing drafts of this essay.]
I’m a high school senior, and I attend Irvington High School. I guess giving that info means you could in theory come and track me down or something. Before that, though, you’d have to brave through the wails of hundreds of students suffering from academic stress.
Our stress situation was so bad, we even got a feature in the New York Times.
In general, there seems a pretty competitive culture here in many schools in the Bay Area (and elsewhere, I’m told), where a lot of students care about getting into a good college. This ends up translating into caring a bunch about test scores and taking hard classes. Which apparently ends up with people suffering more than medical school grads(!).
As the NYT article points out, Irvington has implemented a few policies with the goal of helping students out. The results have been lukewarm, at best *1. Here’s what our school tried:
- Bringing in a guest speaker (the one who had given us the stress survey mentioned in the NYT article)… who subtly hinted that less prestigious schools might be a better fit for some of us.
- Creating “stress-free” days with no homework…which many teachers ignore, or are simply bypassed when students spend time studying, as studying is not technically homework.
- Asking students to deeply reflect on the nature of their existence, the transient nature of time, and the seductive nature of AP classes…by allowing them to fill-in a schedule of how they’ll “fit everything into 24 hours” by themselves.
These all seem like the wrong approaches:
- Getting students to accept the reality that they might just not go to the best schools is good, I guess. But unless it also comes with the rallying call of engaging in a full-on socialist revolution, it doesn’t really deal with the whole issue.
- Stress-free days don’t really address why the stress is there in the first place.
- Letting students just write in their own time schedules is just stupid.
The point here is that there seem to be deeper problems that are causing these things. Most efforts so far seem to only treat the symptoms of the underlying issues.
I think I have several of the pieces behind why high school is broken, and this essay is an attempt to get those pieces across.
The God Who Rules High School:
It looks like competition in schools like mine are pretty bad; everyone seems to agree on this.
On one hand, we could stop dealing with a whole bunch of the tragedies and stress that come from competing so hard if we just, y’know, stopped competing so hard.
On the other hand, the above statement seems pretty surface-level and doesn’t accurately represent what the real situation is like.
Let’s say you’re a high school student. You want to go to a Famous University because that’ll raise your social status, which is good because then all the pretty boys and girls will like you. You know that GPA seems to be a big factor for a lot of schools, and it seems like all the other smart kids are taking AP classes to boost their GPA.
Last year, you tried that. And it was hell. You took a bunch of APs, and you just barely pulled through with the help of some stimulants and a messed-up sleep schedule. So maybe taking that many classes this year isn’t the best idea.
What do you do for this year?
Well, you could just take fewer AP classes this year; then you’d probably feel less like shit every morning. But then you’d also be losing an edge for college…*2
Still, you’re pretty sure that Rick, Morty, Heather, and all the other kids are just as tired as you are. So maybe you could convince them to also not take as many classes this year? Then you’d all still be on equal footing — and you’d all be more rested.
Except that everyone would do better if they were the only ones taking the AP classes, which screws the whole plan over. If you want to beat out all the other students, you want everyone else to be taking less hard classes, just so you can take those extra hard classes and stand out.
This ends up destroying trust — of course everyone would all outwardly agree to be against this whole dangerous competition situation. But on the inside, they might be scheming how to secretly take one more AP class than you to get an edge.
Then you’d likely all see each other in the same AP Game Theory game class you’d all promised to not take.
There’s a sort of vicious cycle happening here:
You want to gain an advantage over everyone else, so you’re trying really hard to do well. But so everyone else wants an advantage too. Maybe everyone would like to take a less extreme workload, but they can’t. If you tried to get the other students to “protest” by taking less hard classes, they’re better off refusing or lying to screw you over.
There’s a sense in which every person would like to slow down, but they can’t because they have no guarantee that other people will slow down with them. So the safest option for each person is to just keep running full speed, even it ends up wrecking themselves. As a result, we end up at this equilibrium where most of the competitive students are taking the literal maximum number of hard classes that they can without seriously endangering their health.
Except that sometimes they do endanger their health because they’ve really tried to take on way too much, and we say that this is sad because what else would you say? That this is an acceptable side effect of our competitive culture?
So everyone goes around talking about how bad this situation is and how they’d like it for things to change. Except that everyone is totally incentivized to just say this aloud and continue scheming internally how to betray the others, so nobody even knows if everyone else really means it when they say they things should change.
And we can’t really depend on universities to make the change because they’re competing for students, which in turn keeps the students going, and sometimes employers are also picky about Famous Universities, which keeps the cycle going in the first place.
The whole mess ends up looking like a giant faceless god on a meaningless rampage.
Let me back up.
There’s this guy, Scott Alexander, who’s a psychiatrist and a fantastic blogger, who wrote this experimental essay called Meditations on Moloch. In addition to being something like 14,000 words, Scott mixes poetry with game theory with granite cocks to talk about this ancient Canaanite deity, Moloch. He uses Moloch as a metaphor to represent this entire category of lose-lose situations where nobody really wants things to go wrong, but they do anyway.
Also, there’s a bit in there about dead whales.
Anyway, Scott actually uses education as an example of Moloch in his essay, and I’m using his essay as a starting point here. As for what exactly Moloch is… once he brings in the economics and evolution, it gets a little complex.
For explanation, here’s the two-person version of the Prisoner’s Dilemma:
Imagine you are one of two gentlemen looking to engage in a feisty round of fisticuffs:
The convention is for both of you to show up with only your fists (C-C). Still, you could easily win if you bring a revolver with you; then you’d easily win the duel (C-D). That’d be cheating, of course, but the other gentleman is likely thinking the same thing… And if you have some uncertainty, it’s probably better to bring some insurance. Then you show up to the with revolvers, and you both end up with bullet holes instead of bruises (D-D).
This situation is, I think, fairly analogous to what’s happening in the competitive school environment — there’s just more than 2 people involved. Without guarantees about everyone else, it can feel like the only thing to do is to keep pushing ourselves.
Moloch is trying to point at this general situation where many individuals are trying to compete, and there’s an opportunity to take on more risk for an advantage. In the end, everyone plays it safe and ends up taking the risk (e.g. bringing a revolver to the fisticuffs, overloading on AP classes) just to survive.
The tragedy of the commons is basically the above situation taken to its logical conclusion: When everyone tries to gain an advantage, it can actually be worse for everyone.
The canonical example is that of a “commons” in the olden times, a piece of land that no one owned. Every herder would like to bring their animals to graze, as this was a free resource, but if every herder brought their animals to graze, then the commons would eventually disappear, and everyone would lose out.
In the same way, when everyone in the school system is pushing themselves to the hardest, we all lose out.
Everything goes to hell and then suddenly we’ve unleashed a giant faceless god on a meaningless rampage, as we keep trying harder and harder to compete, at great cost to ourselves and those around us (e.g. more dead gentlemen, more stressed students).
So I think Moloch can help explain, in part, why we don’t just see people trying less hard. It’s not so much that they don’t want to, but without guarantees that other people won’t betray them, they actually can’t.
Gaming the System:
Competition aside, I think that another major problem here is the imperfection of metrics used in the whole education process.
Consider students applying to college: There’s often a frantic frenzy of studying before taking the SAT, for example, because students know that SAT scores are a thing that colleges look for.
(I know most colleges are notoriously opaque about their actual selection criteria, but it does seem like your GPA and test scores tend to be a major consideration. Unless you happen to be the sort of person who achieves nuclear fusion at 14. Then all bets are off.)
Why the SAT, though?
Prestigious universities generally look for well-rounded, intelligent, with future potential. Bonus points if they can later become rich and famous, and give regular blood sacrifices to their school mascot.
Say, though, that one thing that universities might want in their students is intelligence.
However, “intelligence” is a pretty vague concept.
Intelligence maps onto a bunch of characteristics, from problem solving, to critical thinking, to knowing how to hide a body. The schools need a way to go through thousands of applicants; that takes a lot of time! Thankfully, smart kids tend to do well on things like tests and academics, so these metrics are a proxy for measuring smartness.
Even then, it’s not especially easy for the schools.
(Not that I’m still bitter about that rejection or anything, Stanford.)
The metrics we use aren’t obviously perfect, meaning that you don’t necessarily have to be a genius to do well on the SAT *3.
Because, as many students have found out, it’s very possible to raise your SAT score with practice. And as many tutoring centers and test prep companies have found out, you can make a killing by selling these sorts of score boost guarantees to advantage-hungry students and their well-meaning parents.
Leaving aside the question of resource inequality and how we’ve basically started selling the real-life equivalent of RPG stat boosts (SAT +100!), I think this points at a deeper problem of what can go wrong with metrics.
Say you go through a test prep program and score higher. What actually happens? Are you now a more “intelligent” person, in the broadest sense of the term?
To be clear, I don’t doubt that test prep can boost your score. But I really doubt that it actually boosts your intelligence. I claim that the skills you train studying for the SAT — working under a 15 minute time limit, process of elimination, abstruse vocabulary, etc. — aren’t transferrable to other contexts and are largely useless.
“It might be useless, but the colleges care about it!” you claim, “So it’s still important to work on maximizing my score, even if the metric is largely useless.”
Which means we have a bunch of students trying hard to send a signal about how they’re smart…without really working on, y’know, actually getting smarter.
Sure, you could just do well in school and hope that what you’re learning transfers well over to a great SAT score…or you could just directly prep for the SAT.
The problem here, of course, is that the SAT is, like many other metrics, gameable.
It turns out this whole situation is perfectly described by this thing called Goodhart’s Law which states, “When a measure becomes a metric, it ceases to be a good measure”.
In short, whenever we start using a metric as a proxy for something we care about, it often gets disconnected from our original thing.
To give another example, here’s a “creative” way to ground your (hypothetical) kids:
As a parent, you’re hoping that your kid will look at the list and go through most of the items, so they can rack up the 500 points needed to get ungrounded. That’s the intent.
As a parent, what’ll actually happen is that your kid will quintuple your water bill as she does 5 loads of laundry in quick succession to get 500 points. That’s what actually happens when you set up poorly laid out incentives, your intent be damned.
I think this is fairly analogous to the situation with SAT scores. There’s the intended way of doing things (e.g. studying normally for the SAT or trying to go through all the chores) and there’s the gameable method that gets the same results with way less work (e.g. test prep or doing laundry 5 times).
This also happens with cheating in school. Sure, you could just study the hard way and hope that it helps you with whatever’s on your test in 4th period. Or, you could “persuade” your buddy from 1st period into telling you what’s on the test.
In each case, Goodhart’s Law shows up when a metric (that was initially used because it had an association with something we cared about) gets rendered meaningless when people realize it’s often easier to directly maximize the metric itself (rather than what the metric was actually trying to measure).
In the case of testing, we’ve ended up with a metric that’s been compromised. I see Goodhart’s Law as an aspect of Moloch itself. Often, our competition is towards a misplaced direction, as a focus on maximizing metrics actually leaves the original meaning behind.
So in our attempt to compete, we actually make the competition that much less meaningful. We end up mistaking the shadow for the figure, the robes for the wizard, the chemtrails for the jet plane.
These are the sacrifices we make to Moloch in Goodhart’s name.
Putting It Together:
Now I’m obviously not suggesting that metrics are terrible and we shouldn’t use them at all. Not all metrics are under such heavy optimization pressures as the ones I discussed above, and many metrics are useful. It’d be hard to drive across the desert in a car without either a speedometer or a fuel gauge.
But at the same time, flicking the needle on the fuel gauge won’t help your car go farther.
We need metrics because our intuitions often fail us when it comes to making good decisions.
Back to the college example, this is why we might want standardized tests when it comes to understanding applicants. Grading across multiple schools can be pretty subjective, so if we wanted to compare a student at Hogwarts with a student from Durmstrang, it makes sense to give them both the Spellcasting Aptitude Test. This way, we have a common scale to compare them by.
And I’m not trying to imply that everyone tries to game the system. Most people don’t. This is why we can sometimes have nice things. We’ve got things like laws, honor, and morals to keep us honest.
Still, though, I think it’s important to look at this Moloch/Goodhart’s Law combo, because it does show up in real life:
The rise of clickbait can be seen as competition gone wrong. As marketers try harder and harder to solicit attention, everybody loses out. Once one advertiser starts talking about “One Weird Trick To Get 5 Inches”, the others follow suit to stay competitive, until we’re bombarded with a bunch of ads that focus on attention over accuracy. We end up sacrificing the truth for sensationalism.
Global warming is a similar problem. It seems to be a responsibility countries seem to want to hand off to everyone else. After all, each country could save some money if they just freeloaded off the efforts of everyone else trying to reduce emissions. But if every country tries to be a little better off by refusing to take part, then nothing ever gets done in the first place.
And then we all die.
So even though we do have some safeguards in place, I still think it’s very much worth our time to examine how we can make things better.
Running in Place:
If we want to find solutions to the problem, though, there’s a trap we need to be wary of: situations where nothing really changes.
Imagine if someone suggested that we could fix problems with cheating on the SAT by just banning the things people could cheat with. That sounds pretty good. But there’s a sense in which trying to just directly block the gameable action when it comes to metrics just doesn’t really work.
Attempts to prevent cheating may just start up an arms race:
Say people decide to cheat on the SAT with smartwatches. So the testing agencies ban smartwatches. Then the cheaters switch to LED rings, which are then also banned. So the cheaters switch to special pens, and then the agency requires you to use their cheat-free pens. This back and forth just keeps escalating until we’re all, y’know, taking tests naked.
(And then someone figures out how to get subdermal implants and it starts all over again…)
What ends up happening here is both sides spend a whole bunch of effort, but they just end up right where they started. Even though the situation escalated, the “battle lines” where the problem was being fought haven’t really shifted.
More example of this can be found in the animal kingdom. Consider the gazelle and cheetah. Fast gazelles outrun cheetahs. But fast cheetahs can hunt down gazelles. Over time, this selection for speed means you have really fast gazelles…but also really fast cheetahs which chase them down.
And somehow, even though both groups put in all this effort, it’s as if all of it was wasted because their fundamental situation hasn’t really changed.
These situations are called Red Queens’ Races, where both sides are trying to beat one another, but nothing really changes, relative to either side.
I see Red Queens’ Races as another facet of Moloch. There’s a similar sense where everyone is trying to act their in own best interests, but this very action traps everyone in a negative loop. In the end, both sides sacrifice time and energy just to stay in the same place. I think it also captures an aspect of how competition can keep everyone trying, while nothing really changes.
When we’re looking for solutions, then, I claim that we don’t want to keep things static; we want to find new ways of solving the problem.
For example, it seems like the gazelles could have won if they, instead of merely running faster, had built traps on the savannah. Obviously that’s pretty unrealistic — evolution doesn’t work that way — but this absurdity does demonstrate what it takes to break out of a Red Queen’s Race.
This helps us see the right direction we’d want our solutions to go. We want our solutions to disrupt and push the envelope across new boundaries, not merely expend energy to maintain the status quo.
Owen’s Probably Terrible Solutions:
Here I’ll try to list out some solution ideas of my own.
“Wait!” you may say, “Owen thinks he can solve this? I’m pretty dubious that any solution he proposes will be any good at all.”
You got me there.
I have faith that you will probably think of holes in my plans and ways to criticize my ideas.
But, actually, that’s fine. That’s great. That’s fantastic.
In fact, I want you to viciously tell me why my plans won’t work. Be that asshole who games the system by doing the laundry 5 times in a row. Explain how my suggestions won’t work and why they’re unrealistic.
Otherwise, we won’t be able to come up with actually good solutions. As the saying goes, “If you can’t criticize, you can’t optimize.”
So here are some of my ideas for solutions. They get progressively more outlandish.
(Also, I’ve taken the liberty of attacking my own ideas to get you all started.)
1. Get Real:
Humans are really bad at numbers. And I think part of this leads to enforcing unrealistic expectations about getting into prestigious colleges.
I claim that being brutally honest and being more upfront about acceptance rates is a good idea.
I think it’d be refreshing for schools, counselors, and tutoring centers to just come right out and state the reality of the situation: most people don’t get into the ultra-elite colleges. For example, in a packed room of 100 students, only 4 people and a dismembered corpse get accepted to Stanford.
The vast majority of people (like me) (definitely not still bitter about that Stanford rejection) simply aren’t going to get in.
But there seems to be this narrative that “everyone” can get in. Stories about the 1% extraordinary individuals in special situations don’t help especially when the other 99% of us aren’t extraordinary.
I also think that tutoring centers in particularly guilty of this when they advertise their students heading off to prestigious schools — they don’t show the other 50 people that didn’t get in, but the implication is, “Hey you could join the ranks of these students as well!”
Steps could also be taken by groups who manage college admissions, like the Common Application, to limit the number of colleges students can apply to. This would also force us to really confront our chances from the outside view.
Sure, if you’re that crazy kid who achieved nuclear fusion at 14, you probably can afford to apply to one of the Famous Universities.
For most of us though, the numbers are against us, and we’re probably better off knowing our place like the plebeians we are.
(Socialist uprising, anyone?)
It seems like, to some extent, counselors and schools are already sort of doing this, and it hasn’t been too effective. I also think that lots of people already have fairly realistic expectations. (But they’re not really my target here.)
Also, I expect that the sort of honesty this requires would probably be interpreted as an attack in most cases. People (me included) don’t like being told they’re not special or extraordinary, and this would probably just piss them off.
People might claim that limiting the number of colleges you can apply to restricts freedoms and is patronizing. Famous Universities would probably also fight this because they make several million dollars on application fees alone.
2. Improve Alternate Routes:
I think in competitive schools like mine, there seems to be this general assumption that only Famous Universities breed awesome people. That immediately writes off the other 95% of us who didn’t get in.
I claim that if we want to reduce competition, it’s important to create more choices for higher education, outside of traditional college:
More focus on alternate higher education routes and vocational schools could be helpful. Novel educational opportunities like Fremont’s 42 Coding School, San Francisco’s App Academy, or the Thiel Fellowship all seem like steps in the right direction.
In terms of fighting Moloch, I think this makes competition more spread out and less of the sort of pile-on competition that’s causing our current lose-lose situation. If we have lots of varied choices, we might see less instances of this hyper-competition sucking up people who never wanted to start competing at all.
There’s a two-fold thing I’m advocating here. One is trying to break away from the social assumption / narrative that “Famous University = godhood”, and the other is trying to promote other ways to make our way into the world.
In a similar vein, it’d be good to have more of our national focus shift towards smaller schools who promise benefits like greater attention and novel curricula over the well-established choices. I’m thinking of changes that make this more of a norm and not just a curiosity, which seems to be how we’re currently viewing alternative options.
Other efforts by groups like 80,000 Hours, which provide free career advice to students and focus on practicality over dreams (tying back into #1) also look promising. I myself am a big fan of 80,000 Hours and highly recommend their career profiles.
Changing social norms / narratives is hard, and this may be unlikely to be a very fruitful pathway.
In addition, there’s likely going to be problems with accreditation and validation for new education ventures. We’re back to the original question of “How can we tell if these schools are even teaching anything?” Without oversight, it’s once again a trust issue with the schools actually helping students succeed.
3. Offload It:
If there’s one constant thing we learn from organizational change, it’s that efforts to push for standardization without regards for local nuances can be problematic. This intuitively seems to make sense; those closer to the action tend to know more than higher-ups.
Thus, as a general principle, it seems good to both have systems that are flexible enough to change to meet individual needs. We want any solution we propose to be mutable, easily changed, if it’s not actually working.
It also seems good to give teachers a high amount of autonomy when it comes to leading their classroom. It seems like this is already the case, but there are still standards / standardized curricula. In line with the above, though, giving teachers additional control over what they teach could help with the general goal of learning.
In the best case scenario, we’d have capable people running things such that oversight would be greatly diminished. This appears to generally not be the case for most things.
The proposed solution merely pushes the task of finding a solution lower down the line.
That’s pretty lazy.
4. The Other Things:
This last category is sort of cheating on because I’m using it to represent “everything else”. The above three solutions are all sort of like painkillers for this big, tangled problem; they only address the symptoms, not the underlying problem.
Everything above assumes the current school setup. It doesn’t question anything about how school actually works.
But the whole US schooling system itself seems like it could benefit from lots of changes. For example:
- Improved Teaching Techniques: There is already existing research on ways that help students learn — but they’re not often used in classrooms. More widespread use of better pedagogical techniques, like retrieval practice or spaced repetition seems like it’d be beneficial for students.
- Revamping School: Starting school later in the day and trying out year-round school both seem helpful for improving student performance. In addition the idea of learning six disparate subjects every day for only an hour each seems suboptimal as well. Also, homeschooling as an alternative could be good.
- Separate the teacher and the tester: As things now stand, the teacher has to juggle a bunch of roles in the classroom. Currently, there’s a love-hate relationship between the students where they view the teacher as both their mentor as well as their evaluator. It can be hard to cooperate with someone if they’re also responsible for your failing grade.
- Run Tons of Tests: Classrooms are a huge resource for research: there’s a huge population of students that are constantly given pedagogical treatments and being evaluated. If we tapped into this resource and ran a bunch of randomized control trials on classrooms across the country, we could be gathering a lot more data from schools on what is/isn’t actually working.
A full treatment of the problems with American education would be pretty lengthy, I think. For now, it’s probably good enough to note that we’re operating under a “status-quo framework” which might itself be flawed.
“a)” could be workable, although it isn’t specific enough and doesn’t directly funnel into the whole competition issue. It just gives everyone more powerful tools to study.
“b)” seems like it’d get fairly widespread support, but large institutional change is just generally difficult. Also, there’d likely be pushback on year-round school.
“c)” is not research backed, and it’s already kind of happening with outside tests like the SAT. And we’ve already addressed how SAT can be problematic…
“d)” is very poorly defined and would likely be hard to implement. And research is hard; social science has been having its own problems of late.
School often has a competition-culture; in order to better than everyone else, many students push themselves to their limits. Even though each individual student might want to slow down, they can’t because they have no guarantees about the students around them cooperating. This is leaves us in lose-lose situation where the only stable outcome is everyone working themselves ragged.
Another part of this problem seems to be people who game metrics. In the case of test scores, you could do well by studying or you could just cheat. Cheating tends to be less effort, so it’s still prevalent. Metrics can break down when people find a way to maximize the metric and not what the metric measures, making it less useful.
Many solutions to these problems will only push the issue back and forth, while the problems remain. This is akin to our immune system, where both the viruses and our bodies are locked in a never ending competition to overcome the other side. Thus, we may want our solutions to be more disruptive and focus on trying to shift the overall system of incentives.
Potential solutions are:
- Creating more realistic expectations.
- Focusing on alternatives to Normal College Routes.
- Give teachers more control
- Changing bits and pieces of education entirely.
What The Hell Happens Now?
At this point, I hope you can see the general shadow of how this whole messed-up system in education really does sort of look like a blind god. There’s no intelligence or malice behind it all — it’s just something that happens — but it’s still harmful to everyone in the system.
If you want to take a stab at offering new solutions to the education system (or try slaying Moloch yourself), I invite you to take 5 minutes and really make a plan.
I’d love it if you further critiqued my solutions or left your own suggestions below.
Or, if you think you have the One True Solution (or just want to chat), feel free to get in touch with me at mindlevelup.os [at] gmail.com.
For me, this isn’t really about education.
To be clear, I do think that fixing education is hugely important. Education is upstream of a whole bunch of things like being a good citizen and learning about how the world works.
But I also think that the general shadows of Moloch, with Goodhart’s Law and Red Queens’ Races seem to show up everywhere. From politics to economic development to scientific research, we see misaligned incentives as overwhelming forces that leave us in lose-lose situations no one wants.
My goal in writing all this is to use education as a huge example of examples of this general pattern of how things can go wrong, even if no one’s really being evil.
I’m scared about this whole thing, and I would really like other people to start staring at Moloch. If one of you guys figures out how to solve things, that really takes the load off of my shoulders.
Let’s go god-slaying!
*1 According to the NYT article, AP failure rates did drop, so it wasn’t a total failure. I asked several school admin, though, and no one knows where the data actually came from…
*2 Yes, I’m assuming that these people desperately want nothing more but to go to Famous Universities. I understand this is a controversial assumption, and I address it later.
*3 Apparently SAT scores are correlated with intelligence, but the point I was trying to get across was that the scores can rise with prep, so I think it still stands.