Eureka! Institutional Repository

Current research in Artificial Intelligence has shown a great interest on emotioanal agents. The truth is that they have implementations to a wide variety of areas. However it is not yet clear when agents that they are called emotional always deserve their properties. In short it is at best an open question whether and when these agents express real feelings. The purpose of this thesis is to "match" the Interface Agents with Emotional Agents in a model based on agent negotiation protocols and in the history of their interactions in the context of a Multi-Agent System. These agents will be referred here after as Emotional Interface Agents (EIA). The methodology used for the development of the model of EIAs regarding the mapping of emotions to theme is also based on binary search trees and decision trees because they support the processes of search and insertion effectively. To implement the above methodology, JAVA technology was used and JADE, which is a open source objectoriented software development platform for developing agents, actually a set of JAVA libraries for the specific purpose. As a case study to apply the model developed we used a Multi-Agent System for solving the Timetabling Teaching Program. The results of the application appear to be at least promising. We have reached our first goal, which was to observe different behaviors in different actions of the agents. For example, an angry and upset agent will be expressed nervously and without due respect, while a happy and focused agent will try calmly through consultation to satisfy the teachers' demands. Furthermore, we observed in larger scale of the overall system that there are differences in both the efficiency of work and the “psychological welfare” of agents and this is directly related to the composition of their teams. The simulation has already led to changes that favor the completion of negotiations in solving the timetabling problem in such a “stressful” environment.