Supporting Collaborative Activity

This thesis has two parts. The first part presents SharedActivity, a model for collaborative agents acting in a group. The model suggests mental attitudes for agents with different levels of cooperation and allows modelling of groups in which members are motivated to increase individual benefits. Unlike previous models, SharedActivity is suitable also for groups that do not have a joint goal. The model defines key components of loosely cooperative activity and provides a platform for developing tools to support such activity. We have studied the behavior of the model in a simulation environment. Results show how the benefit attained by cooperation is influenced by the complexity of the environment, the number of group members as well as the social dependencies between the members. The results demonstrate that the model covers social behavior both in settings previously addressed, as well as novel settings. The second part presents an algorithm for solving the problem of iterative search in a closed group. Our solution takes into account the load on agents and the agents’ willingness to help other agents. It also manages reciprocity between agents. The proposed algorithm supports limited-resource platforms. We evaluate the behavior of our algorithm in a simulation environment and compare it to the random walk algorithm. Results show an advantage for our algorithm regarding the random walk algorithm in most environmental settings. These advantages are expressed in retrieving times of wanted objects and fairness in task distribution.