InCOA: Human-Machine-Human Collaboration in Sketch-Based Mission Planning

摘要

A fundamental challenge in exploiting machine intelligence in human collaborative teams is the barrier between natural human-human communication (through speech and sketch) versus the symbolic structured discourse necessary for machine interpretation. Our hypothesis was that a layer of machine reasoning, exploiting situational context and doctrinal models, might induce a sufficiently rich symbolic description of the human-human discourse to enable machine-based reasoning in context of the collaborative problem solving. Traditionally, when outside sources, such as simulations or machine intelligence, need to be leveraged, there must be a break in the collaboration to bring that information into the environment. It requires a user to leave the system in order to collect the required data and then integrate it back into the system so other users can benefit from it. Our goal was to create a system that supports seamless user collaboration and human-machine interactions showing the feasibility of human-machine-human interactions in a collaborative environment. Our vision is a mixed-initiative collaborative adaptive planning system that helps a distributed user base manage complex decision making by including machine intelligence in the decision-making process. The Interactive Course of Action (InCOA) system described in this paper sought to prove that it was possible to induce the high level human thought process of a military plan from the normal multi-modal input currently used during the creation of a collaboratively sketched course of action. This system leverages qualitative reasoning, template classification, and mission-aided plan elaboration to develop the content necessary to meet the stated objectives and to lay the foundation for future research into human-machine-human collaboration.