Abstract
The measure of specificity is introduced and shown to be an appropriate measure for the amount of information contained in a fuzzy subset. We then turn to fusion of fuzzy observations. It is shown that if the data being fused is non-conflicting then the maximal information is obtained by simply taking the intersection of the fuzzy observations. When the data is conflicting the use of the intersection can result in a fused value having less information then any of its components. In order to increase the information content in this conflicting environment some meta-knowledge must be introduced to adjudicate between conflicting data. Two approaches to address this problem are introduced. The first approach considers the possibility of using only a subset of the observations to construct the fused value. The basic rational of this approach is to calculate the fused value from a subset of observations that are not to conflicting and consisted of enough of the observations to be considered a credible fusion. Central to this approach is the introduction of meta-knowledge in the form of a measure of credibility associated with the use of different subsets of the observations. The second approach is based upon the introduction of a prioritization of the observations. In this approach an observation is essentially discounted if conflicts with higher priority observations.