Agents can't force others to perform a certain action, nor write data onto their internal status. But they can perform communicative actions in order to influence other agents.In Multi-Agent Systems, communication therefore forms a basis for agent interaction as well as social organisation: in the absence of communication,each agent would be closed into its own loop.Thus, it is needed to ensure agents cooperate and coordinate their actions, as well as negotiate in situations of conflict. This makes it possible for tasks to be carried out in a joint manner too.
Most of the communication in Multi-Agent systems is based on the speech act theory, which treats communication as action: it assumes that agents perform speech actions just like any other action, to further their goals and intentions.
There are numerous ways in which the communication can take place, the different modes of communication being classified according to their functions:
In the early 1990s, the US-based DARPA-funded Knowledge Sharing Effort (KSE) was formed, with the objective of allowing exchange of knowledge between autonomous information systems, and hence cooperation between agents. As a result, it generated the Knowledge Query and Manipulation Language (KQML), and the Knowledge Interchange Format (KIF).