Learning researchers have a term for something that predates any current AI debate: desirable difficulty. It describes the finding, replicated for decades, that material learned through some friction — retrieving it from memory, spacing practice out, struggling briefly before getting help — sticks far better than material absorbed smoothly and immediately. The friction is not a bug in the learning process. It is a large part of how the process works.
Why smooth feels like progress and often is not
A clear, immediate explanation feels like understanding, in the moment it is received. That feeling is a poor predictor of whether the material will still be retrievable a week later, under different conditions, without the explanation nearby. Struggle-based learning often feels worse in the moment — slower, more frustrating, less obviously productive — and produces a stronger result precisely because the difficulty forced the brain to do more of the work of building the memory itself.
