Emotion is inseparable from cognitive processes and therefore plays a major role in decision making. As a result, it is becoming increasingly important in today's scientific research. The aim of this thesis is to show the advantages of an emotional approach, and to prove that in certain cases computer models equipped with artificial emotions prove to be more efficient than their purely cognitive equivalents. Based on this observation, two emotional models were realised from different study perspectives. They underline the impact of the addition of an emotional dimension in the elaboration of a fast, adaptive and efficient decision. The first developed model uses a graph for strategies representation in order to solve a ten-year-old pupil mathematics exercise called the Cascades problem. Emotion is represented there as weighting values in the graph edges dynamically managed by an ant algorithm. The tests carried out on two versions, one emotional and the other one fully cognitive, show that the use of an emotional model produces a more efficient and adaptive solving. In addition, a second model named GAEA aims at simulating a robot equipped with sensors and effectors and thrown into a prey-predators environment inside which it must survive. Its behaviour is determined by its internal program that evolves thanks to a linear genetic program algorithm manipulating a population of program individuals. Results are promising and indicate that the population produces individuals whose behaviour is more and more adapted, and whose internal activity is analogous to the emergence of relevant emotional reactions.
