The MODIST User Interface

We interact with MODIST through a questionnaire view. Once models have been constructed for our project, we need know nothing of the underlying Bayesian network technology in order to run the models.

The tool comes with a default questionnaire, but the application builder can hide questions, modify them, or augment them. Questions are annotated with detailed guidelines on when, how, and whether they should be answered in a particular context.

We therefore just answer the questions presented, to the best of our current knowledge, and then request recalculation of all the model's variables.

We can select which variables to view in monitor panels on the screen.

MODIST displays the predicted values of variables as distributions in either line graph or histogram format, depending on whether the variable has discrete or continuous values.

Decision support involves a lot of 'what if' -style investigation. Frequently we want to compare the predicted outcomes of 2 or more alternative courses of action. In MODIST, we can designate these as distinct scenarios based on the same model, and display the results side by side.