The reward an agent receives is proportional to how much it has learned about the world, which may require exploration and space exploration to gather information. Some agents, such as AI algorithms, can have their rewards optimized by proving stronger results for them.