Searching for the Impossible in Multi-agent Systems

A multi-agent system (MAS) is a computational society of intelligent computer programs, called agents, that exchange information in order to solve problems that are provided by end-users. Agents are assumed to operate in the background: independently, and without further human help or intervention. The combination of these two objectives puts heavy demands on designers of agents and MAS-architectures. Many different MASs have been proposed, and most of them are rather complex and ambitious. Designers include various kinds of features into their system in order to arrange a full-fledged MAS with rational and autonomous participants. Much seems to be possible in this respect, because few MASs have actually been realized at this stage. This paper proposes to consider the other side of the issue. Earlier research, in other scientific disciplines, has proven that negative results may help to set bounds to an otherwise `liberal' discipline. This paper presents a negative result, that is aime..