To increase the gains which can be acquired for a long period of time by leading the relationship between the combinations of actions set against the future states and the gains based on a change rule showing the changes of gains due to actions and evaluating the future changing states of gains which are caused by actions to build these changing states into the decision standard for selection of actions.

This device includes a storage 1, an input/output device 3, an arithmetic unit 5 which performs the utility calculation, and a communication device 7 which performs communication with the computer of the opposite party side. A change rule of gain vector is newly added to a data base of game trees of the storage 1. When the gains are described to the combinations of actions, the changing states of gains due to actions are expressed by means of the change rule. Then the relationship is calculated based on the change rule between the gains and the combinations of actions set against the future states, and this relationship is led. At the same time, the future changing states of gains which are caused by the present actions are evaluated when a specific action to be decided is calculated. These evaluated changing states are built into the decision reference for selection of actions.