Imagine you are lost in a maze, with thick dark hedges all around you. In the distance, poking above the greenery, you can see a tall post with a sign at the top saying EXIT. How do you get out of the maze? A naive approach would be to walk towards the sign as much as possible. This is a form of greedy algorithm: trying to solve a global problem by making a sequence of always-improving moves.
But a greedy algorithm fails for complicated problems: in your maze, you will initially get closer to your goal, but you will sooner or later get stuck in a dead end. To escape the maze, you will have to abandon your simplistic approach of heading directly towards the exit, and instead, you will need to move away from your goal in the short term. The path to success has many twists and turns.
More sophisticated search algorithms employ a combination of exploration (wandering around looking for a good approach and avoiding the dead ends) and greedy exploitation (finding something good and locally making it better). The skill is in balancing these two processes: too much exploration and you never achieve anything except by chance; too much exploitation and you get trapped in dead ends.
Stanley and Lehman are computer scientists who have taken this insight—that to achieve your goal, heading directly towards it is rarely the best approach, and in some cases may even be the worst approach—and applied it more widely. The real world is hugely more complicated than a puzzle maze, yet many management practices employ greedy algorithms: reach your (complicated, ill-defined, distant, changing) goal by moving directly towards it. Yet no matter how much you improve candlewax and wicks, you will never achieve electric lighting; no matter how many trees you climb, you will never reach the moon.
This slim book is one long discussion of why our current obsession with objective setting, with a pure greedy exploitation-only approach is not a good idea. We need more exploration in our management, research and education, if we are not to get stuck in dead ends, and the authors set out how this could be achieved.