A learning device includes a plurality of individual learners. Each of the individual learners includes a planner configured to generate information for defining an operation of the operation subject corresponding to itself, and a reward deriver configured to derive a reward obtained by evaluating information to be evaluated including feedback information obtained from a simulator by inputting information based on the information for defining the operation of the operation subject to the simulator. The planner performs reinforcement learning based on the reward derived by the reward deriver, and at least two of the plurality of individual learners are different in the operations of the operation subject in which the reward derived by the reward deriver is maximized.