Sonification of probabilistic belief networks

We describe a working sonification system for design and implementation of real time data driven auditory display. Sonification is applied to enhance the visual display of an interactive decision support system consulting a Bayesian Belief Network (BBN). The sonification case presented in the paper employs the concept of auditory signature. The auditory signature is attributed to the nodes that observers wish to keep track of, particularly for monitoring the dynamics of internal nodes. The objective is to provide fine gradients of auditory information to help observers be aware of the relative contribution of internal nodes to the final outcome. For implementation of the prototype system, we developed a task based model of BBN dynamics. This model provides criteria for the design of a sonification architecture. The early development of a prototype architecture allows the research team to identify constraints presented by the visual display and interactivity of the BBN, and to develop alternatives early in the project cycle.