A decision support system for group decision-making

Abstract

This research created a decision framework and decision support system specifically for selecting construction designs. They could resolve public conflicts with known options using information and communication technology (ICT). They contributed to the conflict management and multi-criteria decision-making field with the ability to speed up the consensus-building process and make the better course of treatment of uncertain information provided by humans. While the theme of this thesis is about construction, the proposed methods can also be applied into other similar projects, such as product design selection and hotel selection.

In detail, the proposed decision framework involves three major parts: preliminary work, ICT-aided mediation, and ICT-aided arbitration. Preliminary work is the preparation of the latter two parts. ICT-aided mediation and ICT-aided arbitration function like human-performed mediation and arbitration but quantitatively with the aid of ICT and the proposed decision support system, based on two modified methodologies -- MCCM-based Delphi method and VS-TOPSIS method. Also, a new scale was created to capture participants' opinions towards the levels of importance of the selection criteria.

Six experiments were conducted in this research. Experiment 1 showed that humans' levels of resistance to change (RTC) and the extent of change from their initial ratings followed a natural logarithm relationship. Experiment 2 found that people might not provide firm answers when they were asked to make subjective judgements. Experiment 3 concluded that the uncertain information provided by humans would affect the outcomes of the TOPSIS method. Experiment 4 compared the proposed Delphi method with the traditional Delphi method. It found that the modified Delphi method would make the consensus building process faster. Experiment 5 investigated the effect of different objective functions in the minimum-cost consensus model (MCCM). It showed that the consensus-building process would slow down if the objective functions did not match with the relationship between humans' levels of RTC and the extent of change in their initial ratings. Experiment 6 compared the proposed VS-TOPSIS method with the traditional TOPSIS and VS-TOPSIS methods respectively. It showed that the proposed method often generated similar outcomes to the traditional methods. All these results confirmed the claims made in this thesis.

Also, five sensitivity studies were carried out to examine the robustness of the proposed decision support system. Study 1 showed that the modified Delphi method performed faster in the consensus-building process than that by the traditional method provided that the size of participants was large enough. Study 2 found that reserving a gap between the targeted consensus levels and the minimum acceptable consensus levels could speed up the consensus-building process. Study 3 suggested that the consensus-building process would be slower if the minimum acceptable consensus levels were set higher. Study 4 concluded that the threshold levels of uncertainty would not seriously affect the consensus-building speed. Lastly, Study 5 discovered that the degree of similarity between the proposed VS-TOPSIS method and the traditional TOPSIS and VS-TOPSIS methods would be higher if more alternatives or fewer criteria were involved in the decision-making process.