Towards a Collection-Based Knowledge Representation: Application to Geo-political Risks and Crisis Management

The paper aims at presenting the use of a collection based knowledge representation to improve a geopolitical risks and crisis management system. Initially, the system was built onto classic ontological knowledge representation for the use in a multi-agent system. Events occurring were classified using object matching with model events. After comparisons with type scenarios, the system could predict if a crisis can happen. A better understanding of crisis appeared and the system was completely working but, only if the main agent which has the task of questioning the user was replaced with a human because an ontological model is not well adapted for such work. This gives us motive to reconsider the role and the design of the system, stressing the need of a better knowledge management. A collection based knowledge representation seemed to be a good alternative to gather and organize critical information without losing time and information in a type matching process. It is one of the first times this concept of collection, well known in the artistic domain is used in an IT system. We will explain this particular knowledge representation through the successful example of the CHEOPS Project.