BNet.Builder - Novice User Interface

Background

Bayesian belief networks are an incredibly powerful tool, but they also require a great deal of technical knowledge to set up and manipulate. Charles River Analytics developed a powerful graphical interface called BNet.Builder to help experts quickly build, test, and implement belief networks, but the tool was still overwhelming to novices. In an effort to bring the benefits of Bayesian belief networks to the widest audience possible, a simplified user interfaces was researched, tested, and designed.

Project Summary

BNet.Builder was designed to be an easy to use, easy to learn graphical user interface that lets users create models quickly, without fussing with frustrating and time-consuming dialog boxes, edit and compile modes, nested menus, and other irritating features so common in other belief network modeling packages.

In addition, a simplified user interface mode was created that hide many of the less intuitive aspects of building belief networks, such as merging the separate true and false evidence values for a boolean node into a single linear scale, and greatly improved the ability of novices to use the system. Similar modifications were made to ordinal and categorical nodes. These changes allowed novice users and subject matter experts to leverage the benefits of belief networks without needing lengthy training.