Ask an AI assistant for a hint and you'll get a random walk. Sometimes it names the concept. Sometimes it works out half the problem. Sometimes it gives you the answer wrapped in a question mark. The variance is high and the ceiling is uncontrollable.
A hint ladder solves this by prescribing what each rung is allowed to do. Level 1 names the general area. Level 2 names the tool. Level 3 shows the setup. Level 4 shows the first computational step. Level 5 shows the second-to-last step. The final step is the student's — always.
This structure has two properties that a single-hint system cannot: predictability and preservation. Predictability, because the student knows what they'll get and can decide whether it's worth 10 points. Preservation, because the answer is never leaked at any rung.
The design tradeoff is that early rungs feel unsatisfying to students who wanted the answer. That is a feature. A satisfying rung one would be a rung that gave away too much.
Written by Brainback Research. Published Mar 16, 2026. Filed under Product design.