School is a Broken Loss Function

Every system that trains behavior trains it against a loss function. In machine learning, the loss function is the thing you are minimizing; in life, it's whatever signal tells you that you did badly. The quality of the training depends almost entirely on the quality of that signal. A good loss function is honest, immediate, and resistant to being gamed. A bad one rewards whatever correlates with the real goal cheaply enough to fake. Most of what is called "discipline" or "executive functioning" in high school is actually the shape of a mind trained against a bad loss function for a decade, and the shape is worse than the absence.

This is not a personal complaint about school. It's a claim about what measurement does to the thing being measured, and what that does, in turn, to the students being shaped by the measurement.

goodhart, campbell, and the classroom

Donald Campbell stated it cleanly in 1976: "the more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor." [1] Marilyn Strathern reformulated it a generation later: "when a measure becomes a target, it ceases to be a good measure." [2] The example both authors kept returning to was education. Tests, grades, class rank, GPA. A proxy for learning becomes the thing being optimized, and the underlying learning it was supposed to measure drifts away from it.

What this means in practice is that every hour a student spends on school is an hour spent optimizing a proxy that the institution itself knows has been corrupted. The cheating data confirms it. A Rutgers survey of 24,000 students across 70 U.S. high schools found that 95% admitted to some form of academic dishonesty, 64% specifically on tests, 58% on plagiarism. [3] A loss function that 95% of participants route around is not measuring work ethic. It is measuring something else, probably willingness to perform the ritual, and it is doing it badly.

executive functioning is training for the proxy

The cognitive science meaning of executive functioning, planning, working memory, inhibition, is real and worth having. [4] The thing marketed to high schoolers under that name is different. It is a bundle of compliance behaviors: keep the planner, check the LMS, break the task into subtasks, start early, meet the rubric. These behaviors are framed as discipline, but mechanically they are proxy-optimization. They are the shape your attention takes when you have accepted the measured thing as the real thing.

Self-determination theory has been measuring the cost of this for decades. Deci and Ryan's work across hundreds of studies shows that external controls, grades, rewards, punishments, surveillance, tend to undermine the intrinsic motivation that actually predicts long-run achievement. Autonomy, competence, and relatedness are the conditions under which motivation internalizes; pressure toward a controlled metric is the condition under which it doesn't. [5] The more elaborate the executive-functioning scaffolding around an assignment, the stronger the signal that the assignment cannot survive on its own merit, and students pick up on that signal even when they can't articulate it.

So "good executive functioning," in the school sense, is often a marker of Goodhart's law working on a person. It is what compliance looks like when it is running smoothly.

procrastination is a symptom of a working detector

This is where procrastination comes in, and why the standard framing of it as a character defect is wrong in a specific way.

Chu and Choi, in a 2005 paper in the Journal of Social Psychology, separated procrastinators into two groups. Passive procrastinators freeze, miss deadlines, and underperform. Active procrastinators delay deliberately, prefer working under pressure, and across measures of time use, self-efficacy, and academic outcomes look more like non-procrastinators than like the passive kind. [6] The behavior is identical from the outside. The internal structure is not.

The simplest read of that result, and the one that fits the loss-function frame, is that active procrastinators have a functioning detector for proxy work. Given a task whose real importance is much lower than its assigned importance, the correct response is to discount it, let the deadline do the compression, and then execute under the pressure. The Yerkes-Dodson literature going back to 1908 supports the execution half of this: for well-practiced tasks, higher arousal improves performance, and focus gets cheaper near a deadline because the opportunity cost of distraction collapses. [7] The detection half is just a calibrated student noticing that the proxy is a proxy.

This is distinct from the failure mode. Freezing, avoidance, missing deadlines, lying to yourself about the work, those are real pathologies and they do track with bad outcomes. The defense here is narrow. It is that the specific move of recognizing low-value proxy work, deferring it, and collapsing it into the minimum real effort it deserves is not a failure of executive functioning. It is executive functioning pointing at the right target.

honest loss functions

The contrast worth drawing is with domains where the loss function can't be gamed.

A barbell does not care about your planner. Either the weight goes up or it doesn't, and the feedback is immediate and physical. A wrestling match does not care whether you looked disciplined in practice; if you coasted, you end up on your back against someone who didn't. Code does not care how organized your notes are; it either compiles and produces the output or it doesn't, and the error messages are indifferent to your self-image. These domains are hard in a way that school often is not, and the difference is structural, not a matter of effort. The loss function is honest. You cannot optimize the proxy without also moving the underlying thing, because there is no proxy.

Discipline built against an honest loss function is durable. Discipline built against a gameable one is, almost by definition, a set of habits for gaming. This is why the correlation between "organized student" and "capable adult" is much weaker than the school version of executive functioning predicts. The students who look most disciplined in a high school are, some of the time, the ones who have most completely internalized a corrupted metric.

what to concede

I don't want to overstate this. There are things school measures that track something real. Basic literacy, numeracy, the ability to sit with a hard text long enough to understand it, the ability to produce coherent writing under constraints. Those are worth having, and the measurement of them is noisy but not fake. There are also teachers and classes where the loss function is close to honest, usually because a single person is holding it up against institutional pressure to flatten everything into a rubric.

The argument is not that school is worthless. It is that the specific thing being sold as executive functioning, the planner culture and the productivity scaffolding, is training against the wrong signal, and that the students who partially opt out of it, by procrastinating the busywork and spending the saved attention on work whose loss function pushes back, are often the ones whose minds are calibrated correctly.

the uncomfortable part

If Campbell and Strathern are right, then the cleanest diagnostic for whether a system is training something real is whether it can be gamed. School, by its own 95% admission rate, can be. Gyms, mats, and compilers can't, or at least not in the same way. A mind that notices this and redistributes its attention accordingly is not undisciplined. It is doing the one thing an optimizer is supposed to do, which is refusing to spend more on a proxy than the proxy is worth.