|
|
|
|
An agent which operates in an environment with poor knowledge behaves suboptimally.
If a group of agents act in the same environment, information
about their actions can be used to improve the knowledge and the behavior of each agent.
The problem can be solved by providing knowledge or learning capabilities
to the agent, but in some domain the knowledge description can be very hard
and learning from behavior observations requires a more complex agent.
Our approach is
to make the agent behave as a member of the group would do, i.e. as he/she implicitly
belongs to the same "culture", without extra-effort or direct interaction.
We introduce the concept of Implicit Culture and propose a general architecture for
Systems for Implicit Culture Support. The main component takes as
input the current agent's behavior and a set of past actions performed
by the group and suggest actions coherently.
These "cultural actions" are representative of the behavior of the group's
members and suggested in an implicit way, namely the overall process does
not require to be known by the agents. |
|
|
|
|
