Advanced Decision Architectures Collaborative Technology Alliance An Interactive Decision Support Architecture for Visualizing Robust Solutions in High-Risk Military Environments
Advanced Decision Architectures Collaborative Technology Alliance Medina Wasl
Advanced Decision Architectures Collaborative Technology Alliance Vignette Receive a high likelihood intelligence report that a bomb maker has set up shop in one of three locations within the town. –Uncertain which of three locations –Reported armed hostiles –Busy public areas –Risk for both civilian and friendly casualties –Risk of losing the bomb maker You must decide how to allocate the resources of the company to find and capture the bomb maker.
Advanced Decision Architectures Collaborative Technology Alliance You are faced with a decision involving risk and uncertainty. –Do you trust that the assumptions of the world state information is correct and choose the COA that would maximize the expected value of the outcome (i.e., a normative decision)? –Do you acknowledge that the world is complex, dynamic, and uncertain; and therefore, choose to explore contingencies and alternative parameter assumptions leading you to select a COA that would still result in a desirable outcome if the assumed world state information turned out to be incorrect (i.e., a robust decision)?
Advanced Decision Architectures Collaborative Technology Alliance Robust Solutions What are they? –Solutions produced by simulation models that result in successful outcomes under a variety of supportable assumptions based on the idea that planning should aim to produce plans that have satisfactory performance even when reality fails to match the assumptions and models used in planning. What they are not? –Not normative, or optimal, solutions. –Not solutions that assume a defined probability space and a rational actor. –Not based on maximizing the expected value.
Advanced Decision Architectures Collaborative Technology Alliance Why Robust Solutions? How important is it that the decision maker understands the impact of false planning assumptions and has explored contingencies for those cases? How does the decision maker respond to situations when the assumptions of an optimal solution don’t actually fit the bounds of reality vs. the decision maker already aware of the performance of the choice under a variety of real world constraints? Robust decisions remain sufficient even when reality is not as expected, because robust decision making acknowledges models of the world are incomplete and human decision making error prone.
Advanced Decision Architectures Collaborative Technology Alliance Sensor Allocation Sensor allocation is a fruitful research area Risk and uncertainty involved Complex due to increased use of unmanned resources in military operations Multiple criteria and goals to consider ( Reliability, Accuracy, Survivability, Mobility)
Advanced Decision Architectures Collaborative Technology Alliance DSS R&D Prototype of an interactive system to help develop sensor allocation plans. –A decision support architecture developed by The Ohio State University –Visualization concepts developed by Alion Science and Technology Designed to help commanders determine how to reallocate a limited number of sensors in response to a changing threat. The decision support architecture consists of: –a Generator-Evaluator –a Pareto Filter
Advanced Decision Architectures Collaborative Technology Alliance Generator - Evaluator Generates plans and evaluates against multiple criteria. –E.g. timeliness, coverage, longevity, survivability, etc. Based on an evolutionary algorithm –Survivors of each generation composed of the Pareto-optimal set with respect to selected criteria.
Advanced Decision Architectures Collaborative Technology Alliance Pareto Filter Determines the Pareto-optimal set of survivors from each generation of plans. –Each plan takes on values from several criteria. –Pi dominates Pj if Pi is better than or equal to Pj in all the criteria. –Pj can be discarded. No plan dominates another plan. The only way to choose between Pareto-optimal plans is to perform trade-offs.
Advanced Decision Architectures Collaborative Technology Alliance Visualizations The Visualizations have three linked components: –a terrain (spatial) view –a data view –a timeline view
Advanced Decision Architectures Collaborative Technology Alliance Interaction with the Data View
Advanced Decision Architectures Collaborative Technology Alliance NAI 2 Maximum Coverage
Advanced Decision Architectures Collaborative Technology Alliance NAI 1 Maximum Coverage
Advanced Decision Architectures Collaborative Technology Alliance NAI 3 Maximum Coverage
Advanced Decision Architectures Collaborative Technology Alliance Interaction with the Spatial View
Advanced Decision Architectures Collaborative Technology Alliance Modified Plan
Advanced Decision Architectures Collaborative Technology Alliance Interaction with the Timeline
Advanced Decision Architectures Collaborative Technology Alliance
Current Situation
Advanced Decision Architectures Collaborative Technology Alliance Specific Plan Comparisons
Advanced Decision Architectures Collaborative Technology Alliance Plan Comparisons
Advanced Decision Architectures Collaborative Technology Alliance Visualization Studies Several studies conducted to explore visualizations: –Information presentation –Information format Other factors: –Training, feedback, time pressure. Relationships to cognitive biases and decision strategies –Risk tendencies, information focus, decision strategies Work with SMEs to understand: –critical cues, important COA criteria, tradeoffs, alternative solutions, and the reasoning/strategies.
Advanced Decision Architectures Collaborative Technology Alliance Integrated DSS Takes computational load off the planner Facilitates visualization of alternate plans. Understanding alternative outcomes. Allows application of tacit preferences or assumptions The result is an integrated suite of technologies and visualizations for: –Generating military sensor allocation plans –Simulating military sensor allocation plans –Multi-criterially comparing military sensor allocation plans –Assessing the robustness of military sensor allocation plans
Advanced Decision Architectures Collaborative Technology Alliance Scenarios Sensitivity analysis Assessing missing model information Contingency planning (reference scenarios in the chapter)
Advanced Decision Architectures Collaborative Technology Alliance Recap Interplay between robust solutions, normative models, and the ability to understand the risk environment. Assuming dynamic and imperfect information, we anticipate more robust solutions if various decision outcomes and consequences are better understood. Consequently, we developed a DSS that that allows commanders to understand the consequences of various courses of action while maintaining flexible options in cases of sudden combat reversals.
Advanced Decision Architectures Collaborative Technology Alliance Major Accomplishments Able to show implications of a plan under different planning assumptions. Also able to show that how information is presented, and interacted with, matters for uncertain decision making situations. Two fundamental roles for military commanders in the planning and decision making process: –Facilitation –Prevention
Advanced Decision Architectures Collaborative Technology Alliance Future Directions Currently working towards: –Developing specific measures of robustness. –Designing an experiment to demonstrate robustness.