Tracking Uncertainty Propagation from Model to Formalization: Illustration on Trust Assessment
Résumé
This paper investigates the use of the URREF ontology to characterize and track uncertainties arising within the modeling and formalization phases. Estimation of trust in reported information, a real-world problem of interest to practitioners in the field of security, was adopted for illustration purposes. A functional model of trust was developed to describe the analysis of reported information, and it was implemented with belief functions. When assessing trust in reported information, the uncertainty arises not only from the quality of sources or information content, but also due to the inability of models to capture the complex chain of interactions leading to the final outcome and to constraints imposed by the representation formalism. A primary goal of this work is to separate known approximations, imperfections and inaccuracies from potential errors, while explicitly tracking the uncertainty from the modeling to the formalization phases. A secondary goal is to illustrate how criteria of the URREF ontology can offer a basis for analyzing performances of fusion systems at early stages, ahead of implementation. Ideally, since uncertainty analysis runs dynamically, it can use the existence or absence of observed states and processes inducing uncertainty to adjust the tradeoff between precision and performance of systems on-the-fly.
Origine : Fichiers éditeurs autorisés sur une archive ouverte
Loading...