Social Learning Case Study

As is mentioned in [10] and [24], Social Learning is usually introduced as a learning technique in environments where more than two agents act autonomously, where each one has its own information about and knowledge about the world and the other agents.

Manly Social learning, like reinforcement learning, was developed using some traits of human behavior, i.e. the ability of humans to learn from environments that are rich in people, interactions and lots of unknown information. For a game agent, the social environment is a game with all its components and entities [11]. As is shown in Figure 4, there are two extremes and three same levels of the extremes, in studying multi agent systems [10][24]. The two extremes regarding the interactions between agents are:

• Cooperation: agents cooperate for a common goal, they can share a utility functions and knowledge. Example, industrial robotic systems cooperate to create production lines and RoboCup teams co-operate to win a match.

• Competition: an agent can only win when some other agent loses (often referred to as a zero-sum game). Example, as in our case study, creating social events-tournaments with many different agents. Figure 1. MAS Social Organizations

The presented three levels are common for both extremes due to their

In the leftmost board in Figure 1, pointed with the arrow pawn, demonstrates a legal (“tick”) and an illegal (“cross”) move. The illegal move being due to the rule that does not allow decreasing the distance from the home (black) base. The rightmost boards demonstrate the loss of pawns, with arrows showing pawn casualties. A “trapped” pawn automatically draws away from the game, like in the middle board the black pawn pointed with the arrow. Another example of a trapped pawn is presented in the right most board, where there is no free square next to the base of the white pawn, the rest of the white pawns of the base are