My friend Michael Petrilli just wrote a great essay about the “homework gap.” He sets aside the upper-class question (stressed out kids with too much of it) and steers readers to the gap (some kids do what’s assigned, some kids copy and cheat, and some kids skip it).

“Our goal when it comes to homework,” Petrilli writes, “should be to get more students to do more of it—at least the valuable, productive kind...and that means addressing the barriers that some low-income students face.” Obstacles include fully-functioning laptops, quiet places at home, and help when you’re stuck. I agree.

My take is that, in addition to Petrilli’s points, the key to addressing the homework gap is to find and examine “outlier teachers”—those inside any high school who generate unusually large amounts of productive homework *completion*, compared to other teachers in the same school.

These outlier teachers are able to change the behavior of students, the same kids who—for other teachers—tend to skip assignments or just copy from friends (or AI). But somehow, in these outlier classrooms, the exact same students show up with work they actually did on their own, despite the same noisy home with crappy Wi-Fi and no parent push.

What do kids say when they describe such teachers? I’ve learned three common themes from students, from my time operating and observing schools.

1. The teacher nails the homework basics, meaning assignments are clear, reasonably useful, and graded fairly and quickly. Moreover, these outlier teachers “patrol” for kids who might cheat, perhaps using pop quizzes where students must answer “live” the exact same questions they supposedly did ten hours before.

2. The homework task is “plausible”—meaning that most students exit today’s class actually able to do tonight’s homework when they get home (assuming they deign to try). This outcome is not easy to achieve. Broadly speaking, it means “better teaching,” specifically legitimate “checks for understanding” and adapting on the fly—including sometimes changing the assignment five minutes before the bell rings.

What happens instead? Imagine a weaker teacher who nominally taught kids today about the quadratic equation. By nominally, I mean presented the topic, but it’s clear that the median student didn’t grasp it. Now the teacher faces a tough choice. Should she assign typical quadratic equation homework? If she doesn’t, then someone could say she lowered expectations. Or should she assign really Mickey Mouse–type problems, easy ones that her median student can do but that fail the “reasonably useful” aspect of theme 1 above. Doing those easy problems doesn’t mean a kid has even moderate understanding of quadratic equation.

3. Finally, when students describe these outlier teachers, the ones they “try harder” for, often they describe the relationships. The same kid who is inclined to skip homework for most of her teachers wants to try for the outlier teachers because she feels known and cared about and doesn’t want to disappoint them.

Note that these are three separate ways to win. A teacher doesn’t have to be amazing at all of them to generate more productive homework completion—and more learning—compared to other teachers in the same school. The combinations vary. A struggling student who adores his English teacher? Perhaps he produces a weak essay rather than no essay at all; this gives the teacher a starting point for feedback. A moderately motivated student who feels accountable to an “old school” science teacher might write her science lab report and skip the social studies essay because that teacher never checks to see if work is done.

Math, to be sure, is different than the other subjects. Particularly with literature and history homework, a good faith effort often satisfies both student and teacher. In math, a struggler can make a good faith effort but get absolutely “stuck” on a question and have no path forward. Hence the quadratic equation example above rather than, say, the causes of the Civil War. Maybe he gets help from a friend or teacher for a single question but is then is absolutely “stuck” on several more, unable to proceed.

I think that’s why affluent parents purchase math tutoring for their kids way more than they do for all other subjects combined. I say that as someone who would like to democratize access to (human) math tutoring. Even the best math teachers—those who outperform their colleagues in motivating students—can’t always be “on call” for long periods to help kids who are “trying but stuck.” And these less motivated kids are, so far, resistant to math AI tools; there’s no evidence that AI is shrinking the homework gap.

Like so much in education, how well and how much kids learn from homework depends on the strategies and skills of their instructors. The good news for individual teachers is they have much within their control.