Software Image for Learning by Observation

As it is the case in human societies, software agents could also use learning by observation as an important method of knowledge transference between experts and apprentices even if they have different knowledge representations. However, observation requires that agents and the actions they perform be visible. In this paper, we propose the novel notion of software image that allows software agents, as well as their actions, to become visible to other agents. The software image was designed to accomplish two purposes: allow agents to locate similar experts to observe and provide training examples for the learning by observation algorithm.