An estimation procedure to detect and remove unintentionaljudgemental bias (UJB) in the analytic hierarchy processmethodology

摘要

This paper provides an estimation procedure to detect and toremove unintentional judgmental bias (UJB) in the analytical hierarchyprocess (AHP) methodology for multicriteria decision making (MCDM). UJBreflects a “cyclical hierarchy” because of the existence ofa structural dependence of alternatives on criteria. This violates theprinciple of hierarchic composition embodied in the AHP method. Thejudgment data for comparing the different sources of risk aresusceptible to the “framing effect” bias. The reference AHPand the supermatrix approach can be used to detect for the presence ofUJB in the original AHP data. The estimation and removal of UJB can beachieved by an L2 norm estimation procedure using the Stein estimator.This involves use of the generalized inverse theory and methodology. Thefeedback data matrices provide a “dual” formulation for theoriginal AHP hierarchy. The supermatrix approach is based on astationary Markov system involving the feedback matrix and the originalAHP matrix local priorities at each level of the hierarchy. Once theoriginal AHP judgment data has been revised to remove the UJB influence,it can then be used to provide the correct AHP sensitivityanalysis