In this paper a decision making system for autonomous and social agents who live in a virtual world is presented. This world was built using a text based multi-user game: a MUD (Multi User Domain). In this world the agents can interact with one other, allowing social interaction, as well as interaction with the other objects present in the world. In this paper, the usefulness of using this kind of text based multi-user games as test beds for designing decision making systems of artiﬁcial agents, is proved. The proposed decision making system is composed of several subsystems: a mo-tivational system, a drives system and an evaluation and behaviour selection system. The selection of behaviours is learned by the agent using reinforcement learning al-gorithms. The dominant motivation is considered as the inner state of the agent. In order to simplify the learning process, the states related to the objects are considered as independent from one another. The state of the agent is a combination between his inner state and his state in relation with the rest of agents and objects. This sys-tem uses happiness and sadness, deﬁned as positive and negative variations of the wellbeing of the agent, as the reinforcement function.