We can think of value iteration in RL as actually doing backward induction, and thus get a better intuition of what it does.