Manejando información incompleta en problemas de toma de decisiones en grupo en contexto difuso

Resumen

Decision making consists on a thought and cognitive process of selecting a logical and best choice from the set of available options. This is a pervasive task in human beings every day routine. Indeed, we make choices ranging from quotidian elections, such as the type of coffee, to more complex and transcendentals selections, such as the best investment. Therefore the study of decision making mechanisms to obtain the best solution has attracted extensive research attention in very diverse areas ranging from Economy, Psychology and Sociology to Artificial Intelligence, and Engineering. When it comes to the case of a complex choices, in the majority of the occasions, the decision is made by a group of people, also known as group of experts. This kind of decision making processes involving more than one person is formally known as Group Decision Making, GDM. In this situations, even though experts, may have their own opinions and background approaching the problem from different perspectives, they share the common interest in reaching agreement on selecting the most suitable options. The main aim of this dissertation lies in the study and development of new group decision making approaches under highly uncertainty environments with missing information. When dealing with multiple experts in decision making situations, in this new demanding environments some specific research challenges arise. In the following we briefly describe those challenges and explain how this dissertation aims to improve the state of the art of the current research efforts in these lines. ¿ Preference representation formats :The way in which the experts enunciate their opinions highly affects the decision process and so it has attracted extensive research attention. There are multiple ways of enunciating the preferences, ranging from crisp values to linguistic preference relations based on fuzzy sets. Depending on the type of decision making process and the degree of uncertainty involved, it could be better to use one type of preference relation or another. For instance, intuitionist preference relations allow the user to express certain degree of hesitation when enunciating their opinions. Therefore in highly uncertain environments they can be of great help. In this dissertation we will carry out a critical analysis of the different types of preference relations that has been proposed in the literature pointing out their main strengths and weakness. ¿ Missing information Decision making situations where all experts are able to efficiently express their preferences over all the available options might be considered the exception rather than the rule. Indeed,this scenario requires the experts to posses a precise or sufficient level of knowledge of the whole problem to tackle, including the ability to discriminate the degree up to which some options are better than others. These assumptions can be seen as unrealistic in many decision making situations, especially those involving a large number of alternatives to choose from and/or conflicting and dynamic sources of information. In this contribution we present a thoughtfully review of the main methodologies proposed to deal with missing information in GDM, for the most extended types of preference relations. Moreover a new methodology designed to deal with incomplete information in highly uncertain environments is proposed. ¿ Consensus: When many experts interact providing their opinions it is natural that they have different point of views. However, in general, it is desirable or even mandatory to reach a decision accepted by the whole group. Therefore the inclusion of mechanisms ensuring that some agreement have been obtained is more than justified. These methodologies are known as consensus processes and, in general, are designed as iterative negotiation processes. In this contributions we analyze the proposed consensus approaches under highly uncertainty environments and we introduce a new approach that leverage the uncertainty inherent in the expert's opinion to increase the agreement taking advantage of granular information without the necessity of going over a multi-stage negotiation. ¿ Information aggregation: Obviously a key issue when dealing with the opinions of multiple experts is how to combine them. For instance, there are situations in which the same degree of importance is given to all the people involve in the decision making, this is the case of the political elections. Nevertheless, there are situations in which it makes sense to allocate more importance to those experts that presents more meaningful answers. That is, less contradiction in their opinions. However under uncertainty situations, the experts confidence on the enunciated opinions also may play a key role. In other words, the opinion of an expert who is hundred percent confident on his/her answer could be more valuable than the opinion of the one that is doubtful. Therefore, in this contribution we present a new GDM approach that calculates the experts degree of confidence on the provided solutions and allocates more importance to those ones that are more confident with their answer. ¿ Software tools to automatically carry out GDM approaches: With the inclusion of new technologies the complexity of the decision making processes have increase involving in many cases a huge number of experts considering a wide set of alternatives. To that aim effective software tools to deal with this complexity, being able to estimate the missing information and at the same time providing meaningful graphical representations needs to be presented. In this sense our aim is to propose a new open source software library that automatically deals with decision making processes.