The effect of splitting objectives in the Analytic Hierarchy Process

This paper experimentally investigates the effect of splitting objectives into more detailed levels in Analytic Hierarchy Process for a multi-criteria decision making problem. A school selection problem is considered for the computational experiment. Subjects were asked to weight the unsplitted as well as splitted attributes at a level of hierarchy. Judgements have been taken from various categories of people. It is observed that details of attribute specifications enhance attribute weights, i.e. people are, in general, prone to overweighting bias. This overweighting bias has been tested statistically and it has been proved that the overweighting bias can change the ranking of alternatives, as expected.