Homework is the institution everyone assumes works and almost no one can defend. A third grader spends forty minutes after dinner filling out a worksheet whose contents she could have done in eight. The parents argue about whether to help. The teacher assigned the worksheet because the curriculum guide said to assign something, and the curriculum guide said to assign something because the district expects homework, and the district expects homework because parents complain when there is none, and parents complain when there is none because they believe — without ever having checked — that homework is what makes children learn. The belief is load-bearing and unexamined. When the evidence is examined, it does not bear the load.
The empirical situation is roughly this. In elementary school, the correlation between homework and academic achievement, after controlling for the obvious confounds, is approximately zero. In middle school it is weak. In high school it is positive but modest, with strongly diminishing returns past about an hour or two per night, and with the gains concentrated in narrow procedural skills rather than in deep understanding. The most-cited researcher in the field, Harris Cooper, who is broadly sympathetic to homework, has produced meta-analyses whose own numbers do not support the practices most schools follow. The institution operates as if the research said something it does not say.
What homework reliably produces, on the other hand, is family conflict, sleep loss, after-school stress, and the slow conversion of learning into a chore associated with the worst hours of the day. A child who reads voluntarily for pleasure in second grade can be turned into a child who hates reading by third-grade reading logs that require the activity be recorded and parent-signed. The mechanism is well understood in motivation research: an intrinsically motivated behavior, when made contingent on external compliance, loses its intrinsic motivation. This is not an obscure finding; it is one of the most replicated results in psychology. Schools assign reading logs anyway.
The collective dimension is where the violation of Law 2 becomes visible. Law 2 asks the system to think — to examine its premises, to compare its outputs against its goals, to revise. Homework as an institution does none of these. It persists because of habit, because of the optics of rigor, because affluent parents demand it as a signal that the school is serious, and because no individual teacher or principal has the standing to be the one who stopped. The collective behavior is locked in by a coordination problem: any school that abandons homework looks soft to the families whose children attend it, even if the school's outcomes are unchanged or improved, because the comparison is to peer schools where the practice continues.
There is a second-order problem about what homework signals. For families with two parents at home and the cultural capital to help, homework is a manageable evening project. For families with a single parent working two jobs, or with parents whose English is not the language of the worksheet, or with no quiet place at home, homework is a daily reminder that the school has outsourced its core function to a household that cannot provide the conditions. The achievement gap that homework was supposed to close is, on the evidence, partially produced by homework itself, because homework rewards the home conditions that already correlate with school success. The institution is regressive in a way that its defenders rarely acknowledge.
Finally, there is the question of what is foreclosed by the hours homework consumes. A child who spends two hours after school on assigned work is a child who does not spend two hours playing outside, building something in the garage, arguing with a sibling, reading a book of their own choice, daydreaming, helping cook dinner, or any of the dozen activities that the developmental literature actually associates with the formation of capable adults. The opportunity cost of homework is the childhood that would otherwise have happened. The institution should at least bear the burden of justifying that cost. It has never been asked to.