Graphical Design of Complex Intelligent Systems
AgentWorks from Charles River Analytics allows you to design, build, and test complex intelligent systems through an intuitive graphical user interface.
AgentWorks provides a robust set of modeling and analysis tools to support complex computational reasoning and an intuitive visual editor that lets you design and build intelligent systems without writing code. Its composite construction paradigm also allows you to incorporate legacy agent models and software systems to quickly field new systems to solve the next set of challenges.
AgentWorks is built from the ground up to be an extensible system. Anyone can contribute custom modules written in Java into its plug-in architecture to make them available as first-class tools that can be efficiently integrated with existing and developing computational modules.
Military commanders need tools that support planning and execution in effects-based operations, where commanders evaluate the effects of a given set of Diplomatic, Information, Military and Economic (DIME) actions within the Political, Military, Economic, Social, Information and Infrastructure (PMESII) context. In this domain, it is essential to model all the PMESII relationships between entities that both affect the outcome and are affected by the outcome of a particular action.
Social network modeling and analysis provide the conceptual framework needed to model the PMESII environment. But since the nature of the intelligence used to specify the nodes and link of a network is imprecise, any tool must be able to reason across the social network using methods that are effective at capturing significant uncertainty and dealing with probabilistic inferencing.
The Charles River Analytics Solution
After our scientists and engineers at Charles River found that available tools for solving these "fuzzy" modeling problems were lacking in the critical functionalities needed for a practical solution, we turned to AgentWorks. It provided us an ideal hybrid artificial intelligence (AI) approach to performing this type of probabilistic social network analysis, incorporating Bayesian Networks (BNs) and other AI technologies needed for rich semantic analysis.
We developed an AgentWorks application to design, construct, and evaluate models of nation states within a relevant DIME/PMESII scenario. We used Connect™ (a social network modeling tool within AgentWorks) to model the relationships between the key entities that impact PMESII analysis, including governments, populations, and insurgent groups in this scenario. We also constructed a model of the objectives, vulnerabilities, and potential courses of action (COAs) for each nation state in the scenario.
In this figure, the logic editor in AgentWorks displays a simple model of the potentially destabilizing effects of specific insurgent group activities.
We then used BNet.Builder™ (a Bayesian network modeling tool in AgentWorks) to construct analysis models that evaluated aspects of our PMESII model to analyze the threat of each vulnerability to U.S. allies and adversaries in the scenario. These models provided details about which allied vulnerabilities the commanders should be protecting and which adversary vulnerabilities could be exploited through DIME actions.
In this figure, the BNet editor displays the BN used to model the effects of PMESII variables on the overall importance of an insurgent group.
Learn More about AgentWorks
AgentWorks components include Connect™, our Customizable Ontology-Based Network Component Toolkit, and BNet.Builder™, our desktop application for rapidly creating Belief Networks, entering information, and getting results.