Proceedings of the 22nd Annual Conference on Behavior Representation in Modeling and Simulation (BRiMS), Ottawa, Canada (July 2013)
In order to extract information from a complex, multi-agent situation, where hard facts are not readily known, it is necessary to determine how much trust a user should place in each information source report. In order to usefully aid a human in making this assessment, an algorithm must be able to rapidly assign trust values to multiple agents, taking into account a priori assessments by the user based on data that cannot be accurately represented in the model. To this end, we have designed a graph-theoretic algorithm for assessing appropriate trust to place in each agent of a multi-agent situation, based on information provided by the agents, the scale of agreement between this information, using the Katz centrality metric. This algorithm takes user beliefs into account while simultaneously allowing for rapid recalculation to test alternate interpretations and helps users avoid information order bias by computing metrics that remain consistent irrespective of the order in which data is entered. This method serves to make the problem of assigning trust to sources more tractable for analysts, and has the potential to greatly accelerate the process of making accurate decisions.
For More Information
(Please include your name, address, organization, and the paper reference. Requests without this information will not be honored.)