In plain words
Failure-based learning is the idea that you learn fastest by making lots of attempts, many of them bad, and correcting as you go. It works best where failing is cheap, which is exactly what games and play provide.
See it
A story often told about a pottery class: half the students were graded on one perfect pot, half on how many pots they made. By the end of the year, the best pots came from the quantity group, who’d learned from hundreds of mistakes.
Why designers care
It’s how players master your game, and it’s how you’ll master designing games. Make failure cheap and informative for players, and give yourself the same deal by building lots of small prototypes instead of one perfect plan.