Charles River Analytics presented research on Bayesian belief networks at the Sixth Bayesian Modeling Applications Workshop during the 24th Conference on Uncertainty in Artificial Intelligence (UAI 2008). The conference took place on July 9, 2008, in Helsinki, Finland. The theme of this year’s Bayesian Modeling Applications Workshop was “How biased are our numbers?” Dr. Jonathan Pfautz and Sean Guarino served on the program committee. Mike Farry and Eric Carlson presented Charles River’s research.
Mike Farry, a Senior Scientist at Charles River, presented “An Experimental Procedure for Evaluating User-Centered Methods for Rapid Bayesian Network Construction,” regarding a recent BNet.Builder™ usability study. BNet.Builder is Charles River’s tool for quickly and easily creating Belief Networks. Farry explained, “This study, conducted on a group of 20 university students, confirmed that users who are not experts in mathematical modeling could understand, create, and reason effectively with our novel Causal Influence Modeling techniques. The study yielded great insight into user interaction with the current BNet.Builder user interface and will inform future UI developments.”
Eric Carlson, a Scientist at Charles River, presented “Methods for Representing Bias in Bayesian Networks” using BNet.Builder. Carlson explained, “Bias is intrinsic to observation and reasoning, whether done by humans or automated systems. The consumers of data that includes biases must be aware of these biases to correctly use the biased information. Through Charles River’s work on encoding expert knowledge into intelligent systems, automating sensor processing, and data fusion, we have developed a suite of methods for representing biases and integrating these representations into reasoning processes. We presented those methods along with discussion of their appropriate use and strengths and weaknesses.”
Charles River Analytics has developed belief network modeling tools with its BNet products. The BNet product family includes:
BNet.Builder™ for rapidly creating Belief Networks, entering information, and getting results
BNet.EngineKit™ for incorporating Belief Network Technology in your applications