Envision a mouse trying to solve a maze for a piece of cheese, and he has to decide which way to turn at the first cross-section. He takes a look at the three possible routes, thinks for a bit, then turns a sharp left and ran. In a complex labyrinth, the mouse would most likely reach a dead-end by blind guessing.
A regular mouse might get confused (where’s my cheese?!). A smart mouse might think “well that’s all wasted effort, let’s start over” and start again from the beginning. But a smarter mouse might try to backtrack as little as possible, by going back to the nearest intersection and making a different turn.