Solutions to the problem of the “Rank reversal” effect in the hierarchy analysis process

The author once again continues the cycle of publications in the direction of improving the accuracy of measurements, eliminating errors and overcoming limitations in the analytic hierarchy process (AHP) using various modifications. This time, an extended study of the “Rank Reversal” (RR) effect, known in science for more than 30 years, was conducted. As a result of the implementation of the control experiment, it turned out that the RR effect in the case of using the traditional form of AHP T. Saaty is nothing more than a set of logical errors in measuring matrix estimates that violate the principle of transitivity, intentionally or unintentionally committed by the decision maker (DM). In this case, the consistency benchmark may be less than 0.1, especially if the matrix does not exceed the dimension of 5 5 and the DM uses small estimates. Eliminating the RR effect “manually” by correcting “mistakenly” exposed matrix estimates that violate the principle of transitivity in the form of simple formal logic when comparing objects is not always possible, but when used leads to even greater distortion of measurements throughout the hierarchy in AHP. Another experiment was conducted, which confirmed the complete exclusion of any probability of the appearance of the RR effect if the DM uses a modified AHP. At the end of the article, the ways of the next inCRement of scientific knowledge through further modification of AHP are desCRibed.