EQV on AI
AI can make a student seem to understand.
We build for the tools that make them
actually understand.
The question we ask before any pedagogy or engineering decision: does the tool require a student to think, or let them skip it.
A capability, not a teacher.
AI can explain, generate, and complete almost anything a student is asked to produce. That removes the friction that used to force a student to think, whether they wanted to or not.
That is neither good nor bad on its own. It depends entirely on what a school, and a tool, asks the student to do with it. A tool that hands over a finished answer is doing something different to a tool that asks a student to explain their reasoning first, even when both happen to run on the same model underneath.
Two kinds of shortcut, at once.
Self-regulation, the ability to plan, monitor, and judge your own understanding, is not something a student arrives with. It is built through practice: attempting a task, getting it wrong, and working through why. AI that answers on demand can remove that practice before it has a chance to happen, so a student produces work that looks like understanding without ever building it.
A version of the same problem shows up for teachers. Technology that is sold as a way to reduce workload often adds to it, and tools built without much input from teachers or the science of learning tend to constrain the professional judgement that makes teaching sustainable, rather than extend it.
EQV Group survey of NSW high school students, conducted in July 2026.
Three decisions we hold ourselves to.
Ask, don’t answer
A tool that hands over a finished answer bypasses the thinking a task was meant to require. One that asks a student to explain their reasoning depends on it.
Inform, don’t decide
Teachers keep the professional judgement that only they can exercise. What AI surfaces is information for a teacher to act on, not a recommendation to follow.
Evidence, not habit
We test what we build against research on how students actually learn, and we say plainly when something is a hope we’re testing rather than a result we’ve proven.
We don’t think the right response to AI in schools is to ban it outright, and we don’t think the right response is to leave it to sort itself out. Both miss the design question that actually matters: whether using a tool requires a student’s own thinking, or lets them skip it. That question is behind everything we build.