Software Theories for Machines: Open-World Semantics for the Knowledge in Machines

The knowledge needs of an agent may be met by exploiting the knowledge environment. In particular, knowledge can be obtained from other agents or a variety of knowledge repositories such as databases. Humans readily satisfy much their knowledge needs in the way just mentioned, but there is still a lack of a conceptual framework to open the way for machines to use knowledge more extensively and intelligently. A major part of the difficulty lies in the difficulty of accessing knowledge in the environment, but even if that were to be solved we would still need to revise our concepts of how machine agents are to use information. The latter problem is explored taking the viewpoint that an agent is a user of knowledge. Humans and programs could be considered knowledge using agents. We can evaluate their techniques and effectiveness in using knowledge. Comparing and contrasting the ways humans and intelligent programs use knowledge helped us identify and better understand and better understand one of the sources of human management abilities. A major part of management involves intelligently seeking, obtaining and assembling knowledge. The incorporation of some of these abilities in automation sytems will increase their autonomy and management capabilites. A simple way for an agent to begin actively tapping into external knowledge that involves modifications to the reasoning cycle is suggested and research in the area of environmental intelligence is continuing.