Los Angeles Unified has imposed a districtwide moratorium on student use of generative AI on district devices for the current school year. All grades.
I understand the concern. Everyone is talking about it. We can’t not. The New York Times features an article this week about a new report from MIT suggesting that AI is causing “cognitive surrender” in college students, as well.
It’s clear that AI makes it extraordinarily easy for a student to produce work they did not actually do. It can write the paper, summarize the reading, answer the questions, and create a polished presentation in seconds.
We need to be worried about what happens when young people routinely outsource the thinking they need to learn to do themselves.
But banning AI will not solve the cheating problem. Cheating is a predictable response to an educational system that too often prioritizes compliance and completion over relevance, agency, and meaningful engagement.
Denise Pope and Victor Lee at Stanford's Graduate School of Education have tracked academic integrity for more than 15 years. Their surveys found that 60 to 70 percent of students were already engaging in at least one form of cheating in a typical month, long before generative AI existed. When they surveyed again in 2023, after ChatGPT arrived, that rate stayed about the same, or dropped slightly.
AI changed the speed at which one could fake learning. It did not create the behavior.
The more important question is why so many students were already looking for ways around the work.
Pope and Lee point to familiar reasons: too much homework, not enough time, pressure to get good grades, and assignments that feel like meaningless busywork. They also found students are less likely to cheat when they feel respected, connected, and engaged in work that has a real purpose.
So yes, we should protect young people from letting AI replace the struggle through which learning happens. But if our response is limited to blocking a tool, we've missed the larger problem. Students were phoning in schoolwork long before AI made it easy.
Conventional schooling asks young people to comply first and care later. Complete the task because it was assigned. Learn this because it's on the test. Move at this pace because everyone else is moving at this pace.
That approach may produce completed assignments. It does not reliably produce curiosity, ownership, or deep learning, and in fact, undermines them.
What excites me about this moment is what it's forcing into the open: producing an assignment has never been the same thing as learning. Real engagement is hard to manufacture around work that feels arbitrary, disconnected from a student's interests, or irrelevant to anything they care about, and no ban fixes that.
At LightHouse, we are choosing intentional use over a ban. We have built AI into our new school operating system in a way that is optional, transparent, and visible to staff. We are rolling it out slowly and with care. Students may eventually choose to work with an AI agent inside our system, but those conversations are logged rather than private, and the system can flag language that may require adult attention.
The goal is to make the useful parts of AI available without allowing the tool to quietly become a therapist, confidant, or substitute for adult support.
We approach the question of whether and when to use AI through what is sometimes referred to as backward design: start with the learning goal, then work backward to the experiences and evidence that will help a student reach it. What should this student come to understand? What should they be able to do? What capacities are they trying to develop? Only then do we ask what role AI should play in the process.
Sometimes AI can help a student generate possibilities, get unstuck, receive feedback, organize information, or pursue a question further. Other times the struggle is the point. If a student needs to develop their own argument, wrestle with uncertainty, learn to write clearly, solve a problem, collaborate with another person, or discover what they actually think, handing that work to AI defeats the purpose.
That determination will not always be the same for every assignment or for every student. In a personalized, competency-based learning environment, it doesn't have to be.
AI Fluency is now one of the competencies we teach, and it becomes a graduation requirement beginning in 2028. Knowing how to use AI matters. Knowing when not to use it may matter just as much.
At LightHouse, we start from a different place. We get to know our students well, collaborate with them on choosing and developing work that feels interesting and relevant, and give them real choices in how they do it. We build the relationships and support that make hard things possible.
AI makes it easier than ever to fake compliance. It also makes it harder to pretend compliance was ever the same thing as learning, and that's the discussion worth having.
Learn more about the LightHouse approach to AI >>