Productive Struggle: Why Learning Should Feel Hard

Quick answer

Learning that feels smooth is usually learning that does not stick. The effort of working on a problem just beyond your reach - retrieving half-remembered ideas, making mistakes, trying again - is not an obstacle to learning; it is the mechanism. Educators call this productive struggle, and it is exactly the part AI is built to remove. One prompt and the difficulty dissolves, taking the growth with it. Parents do not need to ban AI to protect this. They need to protect the hard part: a genuine first attempt before any tool comes out, help that explains rather than completes, and praise aimed at effort and strategy rather than speed and polish.

Why the hard part is the learning

Here is an awkward truth from learning science: the feeling of ease is a poor guide to whether learning happened. Re-reading notes feels smooth and produces little. Wrestling to recall something, or to apply a method to an unfamiliar problem, feels slow and clumsy and produces a lot. Learning scientists call these "desirable difficulties" - conditions that make practice feel harder while making the learning more durable.

Struggle also trains something bigger than the topic at hand. A student who plans an attempt, notices where it breaks and adjusts course is practising self-regulation, and the Education Endowment Foundation, whose evidence base is used in Australia through Evidence for Learning, rates metacognition and self-regulated learning as worth around seven months of additional progress. That habit of thinking about your own thinking is built almost entirely inside the hard part of the work. Skip the struggle and you skip the training.

There is a reason good maths teachers resist giving the answer too early. The moment of reaching for it - even reaching and missing - is what makes the eventual explanation land.

What AI does to the struggle

AI tools are engineered to remove friction. That is their entire value in a workplace, and their entire hazard in a classroom. A stuck Year 9 student in 2019 had to sit with the problem or ask for help. A stuck Year 9 student now can dissolve it in seconds, and learn nothing from the dissolving.

Students themselves sense the cost. In RAND's American Youth Panel research, 67% of students said using AI for schoolwork harms critical thinking. The peer-reviewed picture agrees: Gerlich's 2025 study in Societies, with 666 participants, linked heavy AI use to cognitive offloading - the habit of letting the machine do the thinking - and found the association with weaker critical thinking strongest in 17- to 25-year-olds.

None of this makes AI the enemy of learning. Used after a real attempt, it is the most patient explainer your child will ever meet. The difference between the two outcomes is not the tool; it is whether your teen is using AI or learning with it. The habit decides.

How to tell productive struggle from drowning

Not all struggle is productive, and parents sometimes swing between rescuing too early and preaching grit at a genuinely stuck child. The difference is visible once you know what to look for.

The aim is to keep your teen in the first column for as long as it stays productive, and to step in with structure - not answers - when it tips into the second.

How parents protect the hard part at home

You cannot supervise every assignment, and you should not try. A few repeatable habits do the work instead.

1. First attempt before any tool. Ten honest minutes on paper before AI opens. No first thought, no chatbot.
2. Explanations, not answers. The house rule for AI: it may explain a concept or check reasoning, it may not produce the work. What counts as acceptable AI homework help is worth agreeing in advance, calmly, before the 11pm crisis.
3. Praise strategy, not speed. "You tried a second approach" beats "you finished fast". What you praise is what gets repeated.
4. Normalise being stuck. Say it plainly: stuck is what learning feels like from the inside. If you use AI at work, show them where you struggled first.
5. Keep one question in rotation: could you do this yourself? Curious, not accusatory. It is the fastest audit of whether the struggle actually happened.

What to look for in a structured program

If you are choosing formal instruction, ask one question of any provider: where does the difficulty live? A program whose exercises could be completed by a chatbot in an afternoon is teaching your child to operate a chatbot. Look instead for real projects pitched slightly beyond current ability, mentors who coach attempts rather than distribute answers, and work students must present and defend at the end - the structure Edison builds into the Generalist AI Bootcamp and everything above it.

The wider picture - what Australian teenagers are actually doing with AI and how parents can respond - is mapped in AI education for teenagers in Australia.

The recommendation: stop treating ease as the goal of homework and speed as the mark of a bright child. Protect the hard part - a first attempt before any tool, explanations instead of answers, praise for strategy over polish - and let AI in only after the struggle has done its work. A teenager who can sit with difficulty owns every tool that comes later. One who cannot is owned by them.