Measurement of Preferences in the Analytic Hierarchy Process

摘要

The authors seek to resolve much of the controversy around the legitimacy of theanalytic hierarchy process (AHP) as a decision support methodology. Specifically, they argue that the pairwise comparisons in the AHP should be understood as ratio statements about preference differences between pairs of consequences, because under this interpretation the AHP constitutes a quantitatively meaningful variant of multiattribute value measurement. They also suggest ways to overcome the deficiencies of the standard 1-9 ratio scale, and illustrate that the supermatrix approach fails to eliminate rank reversals unless particular restrictions are placed on its inputs. The remarks in the note indicate how traditional AHP analyses should be revised in order to establish more defensible results; in view of the widespread use of the AHP, they are therefore of substantial practical interest.