Ratio-Scale Elicitation of Degrees of Support
38 Pages Posted: 15 Oct 2008
Date Written: 1993
During the last decade, the computational paradigms known as inflzcence diagramsand belief networks have become to dominate the diagnostic expert systems field.Using elaborate collections of nodes and arcs, these representations describe how propositionsof interest interact with each other through a variety of causal and predictive links.The links are parameterized with inexact degrees of support, typically expressed as subjectiveconditional probabilities or likelihood ratios. To date, most of the research in thisarea has focused on developing efficient belief-revision calculi to support decision makingunder uncertainty. Taking a different perspective, this paper focuses on the inputs of thesecalculi, i.e. on the human-supplied degrees of support which provide the currency of thebelief revision process. Traditional methods for eliciting subjective probability functionsare of little use in rule-based settings, where propositions of interest represent causally relatedand mostly discrete random variables. We describe ratio-scale and graphical methodsfor (i) eliciting degrees of support from human experts in a credible manner, and (ii) transformingthem into the conditional probabilities and likelihood-ratios required by standardbelief revision algorithms. As a secondary contribution, the paper offers a new graphicaljustification to eigenvector techniques for smoothing subjective answers to pair-wiseelicitation questions.
Suggested Citation: Suggested Citation