TaskICDA: A methodology for designing multi-agent decision support systems

Decision support systems (DSS) are computer-based systems that support the domain experts in semi-structured problem solving process. It is not necessarily easy even for experienced software designers, to create a DSS that improves users' performance consistently. These specialized decision-making tasks are very complex and difficult to master. Domain experts usually have a tremendous amount of training and experience through which they have acquired domain knowledge. DSS designers need a step-by-step guideline and a methodology to help them identify users' needs in domain tasks, and to design appropriate functions and information exchange between agents, in order to create a useable and efficient DSS.; The primary contribution of this thesis is: proposing TaskICDA (Task structure based Inter-agent Communication Design Assistant) methodology. TaskICDA methodology uses the task analysis and information dependency analysis to assist the DSS designers in designing the inter-agent communication and human computer interaction of a DSS. TaskICDA methodology consists of two models: (1) task and information dependency analysis model, and (2) agent and inter-agent communication design model.; The secondary contribution of this thesis is: applying TaskICDA methodology to CoRAVEN (military intelligence analysis DSS) system development process by designing the following two software components: (1) IntellView —a digital map software that presents the spatial view and overlay the conclusions of the intelligence analysis on the spatial objects (military symbols on the map). (2) VisQuery—an inter-agent communication mechanism that provides a direct interface for the user to view and query the intelligence analysis information between all three CoRAVEN views, spatial, temporal and logical. The usability studies showed that IntellView and VisQuery can reduce the information overload and allow the analysts performing intelligence analysis more quickly and accurately.