Skip to Main Content
The paper investigates how a group of distributed agents may develop congruent cognitive memories in the form of networks of prototypes. Memories are congruent if they structure reality in a sufficiently similar way for the agents to cooperate and communicate. Each agent develops his own memory independently of the others and based on his own history of experiences. Agent memories are weakly coupled when the agents are in the same environment. They encounter the same sorts of objects and therefore make similar generalisations. Memories can also be more strongly coupled if the agents communicate and use their cognitive memories to structure and conceptualise reality for communication. The paper explores both types of coupling. Experiments are reported where agents build autonomously internal cognitive memories and develop through self-organisation a shared lexicon that makes their memory more congruent.