My research concerns both theoretical and application work on agents and multi-agent systems. The theoretical work is mainly based on the use of decision theory and logic. What is presented in my HDR could be summarized as follows: A point of view about agent architectures based on a modular structure, where each module is responsible for one of the possible capabilities an agent may have (e.g. problem solving, cooperation, communication, etc.). The agent's behavior is the result of the different modules interaction. This point of view also suggests the integration of another specific module in the agents' architectures, which is dedicated to the agent's personality and shows how this can have an influence on the other modules. The completion of the above point of view with the observation that decision-making is a common characteristic of several deliberation processes involved in the operation of the modules and the idea to propose a unified argumentation based model for their representation. The presentation of this argumentation model which is based on the extension of the Logic Programming without Negation as Failure (LPwNF) framework, the integration of the concepts of roles and context to it as well as its use to model personalities. A dynamic planning model based on graph representation, taking into account changes that are generated not only by the environment but also by the agent himself and its exploitation to build a multi-criteria distributed planning framework. A multi-criteria negotiation, as well as an argumentation-based negotiation approach. The first approach is mainly used in a distributed planning context, but is also able to find a compromise between agents having work and private goals in a more general setting. A logical framework for modeling of complex dialogues, adopting the modular agent's architecture and associating each type of the possible dialogues (i.e. negotiation, persuasion, deliberation, etc.) to a specific module, ensuring the automated generation of dialogues and allowing the representation of embedded dialogues. Applications of the theoretical results in: e-commerce, marketing, information services, diagnostic and some experimental work on agent software engineering.