Drawing reasonable conclusions from information under similarity modelled contexts

We are interested in the process of making reasonable conclusions about the value of a variable. We indicate that reasonableness generally depends on the information we have about the variable as well as the context in which we shall use the assumed value. In order to include a wide range of imprecise and uncertain information, we use granular computing technologies such as fuzzy sets, Dempster-Shafer belief structures and probability theory to represent our knowledge and conclusions. While context is a very diverse idea, in order to provide some structure, we restrict ourselves to the special case where context can be modelled using a similarity relationship. Within this framework, we suggest a measure of the reasonableness of drawing conclusions from information in the context of a similarity relationship. We look at the properties of this measure and investigate its performance in a number of special cases.