Vpeljava združitveno-razdružitvenega principa v integrativne pogajalske procese : doktorska disertacija

slovenian

doktorska dela

2009

Univerza v Mariboru, Fakulteta za elektrotehniko računalništvo in informatiko

In the dissertation, an original mechanism for integrative multi-agent negotiations is defined. It is based on the dichotomic sorting analysis, and on the elicitation of preferences in the form of pairwise comparisons, fuzzy values, holistic information, pseudo-criteria and outranking relations. It is the first available approach to apply the aggregation-disaggregation principle to highly autonomous negotiation and group decision-making processes. The aggregation-disaggregation analysis reduces the cognitive load, improves the ability of learning about the problem domain, and automatically adapts individual preferences to match the consensual solution. It is an integral part of the agent interaction protocol, which resolves disagreements among group members and allows for the convergence of opinions by observing the group's direction, and ensures the credibility, reliability and rationality of the consensual decision by introducing robustness metrics that are implemented with several mathematical optimization algorithms. In order to model correlations between preferential parameters, the aggregation-disaggregation approach is extended with two distinct methods thatderive criteria weights according to the influence of veto thresholds. The introduced methodological solutions are evaluated on the basis of a comprehensive experimental and theoretical model comprising of twenty three hypotheses and over twenty variables. The results of a simulation performed on large sets of sampled statistical data that reflect a wide spectrum of problem situations, and theoretical proofs confirm fast convergence, the ability to autonomously guide the decision-makers and their personal agents, the efficiency of processing preferential information, as well as the robustness and credibility of both the suggested decision and the inferred criteria weights.

07.08.2023