Digital games are, at their heart, problem solving spaces […].
Learning games present as puzzle games. You can — with effort — prevent this, but as a rule everything from an exercise in DuoLingo to a tutorial level in a shooter will feel like a puzzle. Because you’re trying to learn something that you haven’t learned yet.
An appropriate difficulty curves makes a puzzle game good. To steep, it’s frustrating; too flat, it’s boring. Flow tunnel.
To craft such a curve, you will want to take all those dials big and small that your game has and fine-tune them. Enemy HP, Enemy Speed, Enemy Tenacity. Opponent ELO, Time, Increment, Handicap.
That’s a neat approach, unless you happen to design a learning game. Suddenly, a good chunk of said dials slip out of your control. Trying to teach some language? Say present tense is incredibly complicated, but a rarely used special past tense is trivial. Trying to teach division from zero to hundred? Notice how twelve is incredibly nice to divide into all kinds of things, but thirteen is terribly prime, fourteen is of medium difficult, and fifteen is just kind of built differently. How do you sort the learning content, how do you rebuild a reasonable difficulty curve?
It’s a difficult question, because learning games forbid to just change arbitrary rules. Then again, limitations breed creativity, and see how well Super Algebrawl does on that math problem…