Using scenarios to explain probabilistic inference

Authors:

Marek J. Druzdzel
Carnegie Mellon University
Department of Engineering and Public Policy
(currently with:
University of Pittsburgh
Department of Information Science
and Intelligent Systems Program
e-mail: marek@sis.pitt.edu)

Max Henrion
Rockwell International Science Center
Palo Alto Laboratory
and Carnegie Mellon University
Department of Engineering and Public Policy
e-mail: henrion@sumex--aim.stanford.edu
(currently with:
Lumina Decision Systems)

Abstract:

On theoretical grounds Bayesian probability theory is arguably the soundest approach to uncertain reasoning, but it has been criticized as being hard for humans to understand. The acceptability and effectiveness of decision support systems (DSSs) depends on their ability to explain and justify their conclusions to users. We describe a method of explaining probabilistic inference by considering the relative plausibility of alternative scenarios. Scenarios consist of sequences of events, often appearing as coherent causal explanations of observed evidence. Selecting only the most relevant variables and the most probable scenarios allows control over the simplicity and precision of the explanations. Process tracing studies of human uncertain reasoning suggests that this scheme resembles the way humans normally reason.