A simulation study concerning the effect of errors on the establishment of knowledge spaces by querying experts (Q1916537)

Adaptive procedures for the construction of knowledge spaces by querying experts recently have been developed [see, for example, \textit{C. E. Dowling}, ibid. 37, No. 1, 21-48 (1993; Zbl 0811.92031)]. In these procedures an expert judges assertions on the set of underlying questions. The procedures minimize the number of such assertions necessary for the construction of a knowledge space by drawing inferences from the judgements already obtained. The question is what happens if an expert judges some of these assertions incorrectly? As our simulations have shown, errors in experts' judgements can lead to severe mistakes in the resulting knowledge spaces. Evidently, assessment procedures can only be efficient and economical in connection with well validated knowledge spaces. Therefore the spaces resulting from the querying procedure have to be validated, for example, through a comparison with data patterns observed in an experiment, before they can be used as a basis for the assessment procedure.