On this page
Should you study solved problems, or struggle with problems on your own? Research gives an answer that surprises many students: it depends on how much you already know. Beginners learn faster from worked examples; once the basics are in place, solving on your own wins. Here is the evidence, a plan that moves you from one to the other, and where AI step-by-step solvers fit.
What is the worked example effect?
In a series of experiments, Sweller and Cooper gave algebra students either worked examples or conventional problems. Worked examples took much less time to process, and students then solved similar problems faster and with fewer errors (Sweller & Cooper 1985). Their explanation: a beginner solving from scratch spends mental effort searching for a method, and that search slows down learning the pattern of the solution.
The effect has held up. A 2023 meta-analysis of 55 studies in mathematics found a medium average effect of worked examples on performance (g = 0.48), whether used for first learning or for practice (Barbieri et al. 2023). Correct examples on their own worked better than incorrect ones. One surprise: in that analysis, adding self-explanation prompts to examples was linked to smaller effects, so how best to study an example is still being worked out. We cover the theory in John Sweller and cognitive load.
When is solving from scratch better?
The benefit does not last forever. Kalyuga, Ayres, Chandler and Sweller called this the expertise reversal effect: techniques that are highly effective with inexperienced learners "can lose their effectiveness and even have negative consequences" with more experienced ones (Kalyuga et al. 2003). Once you know the method, reading yet another full solution adds little. Producing the solution yourself is what builds speed and transfer.
Alexander Renkl and Robert Atkinson proposed a bridge between the two, called fading: start with complete examples, then remove solution steps one at a time until the learner solves whole problems alone. They reported empirical evidence supporting the procedure (Renkl & Atkinson 2003).
A four-stage plan for a new JEE chapter
Here is fading applied to a new chapter, say rotational motion:
Study two or three full worked examples of the core problem type. Read slowly, then cover the rest of the solution and try to say the next step before you read it.
Complete partly solved problems. Take an example with the last one or two steps hidden and finish it. Then one with only the set-up shown.
Solve similar problems alone, checking against the solution only after a full attempt.
Switch to unfamiliar and mixed problems from different chapters, with no examples nearby. Expertise reversal suggests this is where most of your time should go before the exam.
If you get stuck at stage 3, step back one stage for that problem type, not to the start of the chapter.
Where do AI step-by-step solvers fit?
An AI solver that shows every step is a worked-example machine. Used at stage 1, on a problem type that is new to you, that matches what the research supports. Used at stages 3 and 4, the moment a problem feels hard, it removes the very practice that builds skill.
A field experiment with nearly a thousand high school maths students showed the risk. Students with unrestricted access to a ChatGPT-style assistant did better on practice problems, but once access was removed they scored 17% lower than students who never had it; a version with prompts designed to safeguard learning largely avoided that drop (Bastani et al. 2025).
Practical rules:
On any problem type you have seen before, attempt the problem yourself first, for at least ten minutes.
Ask for one step at a time, and try the next step yourself.
After reading a full solution, close it and redo the problem from a blank page.
Note the step you could not find. That step is your real gap, and our post on why you understand theory but fail numericals explains how to practise it.
The Fermi Problem Library has worked problems with a step-by-step solver that explains the reasoning rather than printing an answer, which suits stages 1 and 2. For stages 3 and 4, attempt the problem on your own first and open the solution only to check.
Your next step: pick one chapter this week and run the four stages on its main problem type.
