An agent based model for trust and information sharing in networked systems

Vast amounts of information are generated, shared, and processed in tactical networks. In such systems, human cooperation is a crucial component for effective processing of information. However, human behavior is often mediated by social and organizational relationships, i.e. trust between team members and system level characteristics, such as network delays. The impact these different processes have on each other is not well understood. In this paper, we develop an agent based model for information sharing that incorporates trust into decision making. The nodes in our model are decision makers and are primarily responsible for the disambiguation of received information to obtain correct situation awareness as quickly as possible. Additionally, each node must share information with other nodes to enable the network to attain shared situation awareness. In our model, team members make trust evaluations for fellow team members that they cooperate with throughout a task. Most existing trust models concentrate on whether trust exists or not, and do not consider what task for which trust is being used. In contrast, we consider a new model of trust that incorporates two components: trust for competence and trust for throughput. In time constrained environments, both types of trust are crucial to mission success. Furthermore, this model allows us to study the impact of communication delays on overall trust. We also show how these trust values can be converted to labels that control agent's decision making behavior. To test out this proposed model, we use a command and control experiment platform called ELICIT (Experimental Laboratory for Investigating Collaboration, Information-sharing and Trust). We give initial experimental results that show how trust of nodes change in an ELICIT information sharing task for various settings of initial team trust, the biases of the nodes and possible communication channel disturbances.