Towards a Framework for Modelling Variation in Automated Decision Support

It has often been demonstrated, particularly in medical decision making, that groups of experts exhibit both intra-expert variability and inter-expert variability. The process of arriving at a single consensus decision, given the range of opinions obtained from the panel of experts, is a difficult task. In this proposal we aim to establish a framework that will enable us to create decision support systems that model this complex process. In order to achieve this ambitious goal, we propose to bring together, for the first time, three previously disparate strands of research: non-deterministic fuzzy systems - a very recently introduced paradigm in which intra-expert variation is modelled using fuzzy systems that vary over time, type-2 fuzzy models - in which the standard fuzzy membership values are 'blurred' to model the fact that such fuzzy membership values are themselves uncertain, and consensus models - in which a range of expert opinions are combined into a single overall collective decision. This will lay the foundations for an entirely new form of decision-making utilising fuzzy ensembles, which could make a significant contribution to future developments of fuzzy expert systems that model multiple experts reaching consensus.