Every student with a phone already has access to an AI that will solve their homework in seconds. That's exactly the problem. If the goal of homework is to learn, a tool that removes the need to think is working against the student, however helpful it feels in the moment.
At Classigoo we start from a different principle:
AI should reduce the difficulty of learning, not remove the act of learning.
Helpful is not the same as useful
A worked answer is useful when you've already struggled with a problem and want to check your thinking. It's useless, or worse, when it replaces the struggle entirely. The productive part of learning is the effort of connecting what you know to what you don't yet know.
So the question we ask isn't "can the AI answer this?" It always can. The question is "what kind of help would move this student forward without doing the work for them?"
Context changes the right answer
That's why we're developing an Active Learning and AI Consumption model. Instead of immediately answering a homework problem, the system is being designed to consider things like:
- whether the activity is practice, homework, a quiz or an examination;
- what level of assistance the teacher allows;
- whether the student has actually attempted the problem;
- which concept the student seems to misunderstand.
With that context, the AI can offer a hint, an explanation, a guiding question, a worked example of a similar problem, or a short review of the underlying concept. What it shouldn't do is quietly complete the student's assignment.
Teachers stay in charge
Teachers know their class and their goals. A practice set might welcome generous help; an assessment should allow none. Putting the assistance level in the teacher's hands means AI follows the classroom's rules rather than rewriting them.
What this looks like for a student
Imagine a student stuck on a momentum problem. Rather than the final number, they might get a question back: "What's conserved in this collision?" If they're still stuck, a reminder of the formula. If they're still stuck after that, a worked example with different values. Each step gives a little more support, and each one leaves the last step for the student.
Where we're going
This is a direction, and we're building toward it carefully. We think the schools that get the most from AI won't be the ones that use it most. They'll be the ones that use it to make students better at learning.
