Modelling multiple criteria-multiple participant problems, the integrative approach

Abstract

Many real-world problems involve multiple criteria and multiple participants. Conflicting criteria have been dealt with in multiple criteria decision making (MCDM) and opposing participants have been modelled in multiple participant decision making (MPDM) which includes areas such as game theory and conflict analysis. In order to deal with multiple criteria-multiple participant problems, MCDM and MPDM have attempted independently to incorporate their missing component by extending their existing methodologies. This extension approach has placed Multiple Criteria-Multiple Participant Decision Making (MCMP) as a subset or application area of the existing domains. As a result, MCMP has so far received only limited attention and has no foundation as an independent research domain. In this thesis, direct focus is placed on MCMP. By doing so, insight is gained from simultaneous consideration of MCDM and MPDM which contributes to the improvement of MCMP significantly.

The research objectives for this thesis are twofold. First, a classification scheme for MCMP solution approaches is proposed with the purpose of building a foundation in MCMP research. Then, new methodologies are developed to improve the quality of decision making associated with MCMP problems.

Three classes of MCMP solution approaches are identified and referred to as reduction, decomposition, and integrative. In this thesis, the integrative approach is shown to possess a high potential for providing improvement opportunities in MCMP. The integrative approach essentially deals with MCDM and MPDM components simultaneously. It can allow inclusion of more preference information in the decision making process than the other two MCMP approaches. Consequently, the possibility of missing a solution that is preferred by all parties, namely a Pareto optimal solution, can be greatly reduced by taking the integrative approach.

Based on the integrative approach, two new methodologies are developed. The first method is referred to as candidate enumeration for solving multicriteria games. Candidate enumeration deals with multiple criteria and multiple parties simultaneously by enumerating all possible solution candidates based on a binary tree model to structure ordinal preferences. When compared to other multicriteria methods, the improvement made in candidate enumeration is the capability to explicitly include criteria in a game model and to utilize criterion importance information.

The second method is called logrolling which is for dealing with multi-issue negotiation. Logrolling is a trade-off procedure in which the exchange of loss in some issues for gain other results in mutual overall gain. Logrolling integrates MCDM and MPDM by considering trade-offs among criteria and among parties simultaneously. The improvement in logrolling results from incorporating criterion weight information and MCDM-MPDM interaction into the trade-off process.

The investigation of logrolling and related concepts has proven to be particularly fruitful. Variations of the logrolling-based negotiation procedure are proposed and the properties of the efficient frontier are studied extensively. Also, interesting implications can be drawn in terms of negotiation behaviour and strategies. In addition, logrolling is shown to be effective in accommodating a specific type of non-linear preference The introduction of non-linearity leads to more realistic human preferences, and therefore, can enhance the applicability and decision making quality of the logrolling-based negotiation methodology. An example is given to demonstrate the application of logrolling and related concepts in a three-issue two-party negotiation context with linear and non-linear preferences.

Many opportunities for future research exist based on the research completed here, including various extensions of the two methodologies. In closing, it is hoped that this thesis will encourage further growth of research in multiple criteria-multiple participant decision making.