Explanation in probabilistic systems: Is it feasible? Will it work?

Authors:

Marek J. Druzdzel
University of Pittsburgh
School of Information Sciences
and Intelligent Systems Program
e-mail: marek@sis.pitt.edu

Abstract:

Reasoning within such domains as engineering, science, management, or medicine is traditionally based on formal methods employing probabilistic treatment of uncertainty. It seems natural to base artificial reasoning systems in these domains on the normative foundations of probability theory. Two usual objections to this approach are (1) probabilistic inference is computationally intractable in the worst case, and (2) probability theory is incomprehensible for humans and, hence, probabilistic systems may be hardly usable. The first objection has been addressed effectively in the last decade by a variety of efficient exact and approximate schemes for probabilistic reasoning, applied in several practical systems. In this paper, I review the state of the art with respect to the second objection.

First I argue that the observed discrepancies between human and probabilistic reasoning and the anticipated difficulties in building user interfaces are not a good reason for rejecting probability theory. On the contrary --- they provide motivation for a normative treatment of uncertainty. I point out that probability theory rests on qualitative foundations that capture essential properties of a domain along with such concepts such as relevance and conflicting evidence. In addition, graphical probabilistic models, as opposed to rule-based systems, integrate numerical and structural properties of a domain and provide a natural representation of causality. Finally, availability of a full quantitative specification of a model allows for manipulating the level of precision for both reasoning and explanation.

Keywords:

Reasoning under uncertainty, user interfaces, explanation