Levels of incomplete information in group decision models – A comprehensive simulation study

Abstract

We present a comprehensive computational study on the effects of providing different forms of incomplete preference information in additive group decision models. We consider different types of information on individual preferences, and on weights of the group members, and study their effects on conclusiveness, efficiency and fairness of outcomes at the group level. Furthermore, we analyze possible violations of the axiom of independence of irrelevant alternatives (IIA) as well as the impact of problem characteristics, in particular initial agreement between group members. Our results indicate that providing information in the form of a ranking of differences between consecutive alternatives comes close to providing exact cardinal preference information in several outcome dimensions. However, group decision procedures based on incomplete preference information also show a significant amount of violations of the IIA axiom.

Group decision making with incomplete information: a dominance and quasi‐optimality volume‐based approach using Monte‐Carlo simulation

“…This would allow to evaluate the trade‐off between quality of results on one hand, and the additional effort required from DMs to specify differences instead of just ranks on the other hand. The first step in that direction was taken in Vetschera et al ().…”

Preference Disaggregation: Towards an Integrated Framework

“…Following MAVT perspective in MCDA, preferences of a DM are captured by constructing an additive value function. Additive value functions represent the preferences of the DM by computing an overall score for each alternative (Vetschera et al 2014). This comprehensive value indicates the alternative's quality from all relevant points of view considered jointly.…”

Value of agreement in decision analysis: Concept, measures and application

“…Several authors (e.g. Keisler 2004, Salo & Hämäläinen 2001, Vetschera et al 2014 have previously deployed extensive computer simulation as a means to test ideas on preference programming and value of information. To demonstrate the effectiveness of the proposed measures for the Value of Agreement, a large test dataset with 360 instances was generated (see Table 2 and Supplementary Material A).…”