RADAR: A Framework for Human in the Loop Planning and Data Based Decision Support

Abstract

As the Navy s vision network centric warfare is realized, the commanders, analysts and warfighters alike will often have to make complex decisions under various types of uncertainties, while following precise or imprecise command policies when available, and based on the input of a multitude of structured, semi-structured and unstructured information sources. To ensure that the correct decisions are made, human decision makers will require a decision support system that (i) recognizes the plans and goals of the decision makers, (ii) actively provides decision support through action recommendation and explanation under uncertainties and (iii) proactively supports their information needs through retrieval, rectification, alignment and aggregation of the information. All these activities must be performed continuously throughout the entire mission. The recognition is needed so as to identify the plan and goal of the decision makers, on which the other two activities are critically based. Active decision support can provide the decision makers with a set of action recommendations along with their explanations, under the uncertainties of the current situation (e.g., location of a ship) and the decision makers preferences, as well as the inaccuracy of available command policies. Rectification is needed to improve the quality of uncurated data (with missing information of multiple modalities). The alignment is needed to support seamless querying and browsing of structured and unstructured data (e.g., align captioned/annotated images, micro-blogs and stored records with the appropriate segments of a text document such as military doctrine). The technical objective of this work is to develop, implement and evaluate RADAR1 (see Figure 2), a framework for proactive decision support, that can track the goals, intents and plans of the human decision maker to provide action recommendations and explanations under various types of uncertainties, as well prefetch high-value information from a multitude of incomplete, inconsistent, imprecise and unreliable sources, in order to help them make the correct decisions. RADAR will have the ability to (i) recognize the higher level plans of the human decision makers, (ii) suggest context-specific decision alternatives based on the goals, intents and plans of the decision makers under uncertainties, (iii) learn and adapt to user inputs to improve the current system, and (iv) predict the decision makers information needs, and to proactively query and fetch high value information for supporting decision making. Additionally, its information gathering process would have the ability to select trustworthy and relevant sources, process queries over the structured data that explicitly takes into account data incompleteness, inconsistency and query imprecision, and align the structured and unstructured data to support a seamless querying and browsing experience.

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 08, 2016
Source ID
N000141512027

Entities

People

  • Subbarao Kambhampati

Organizations

  • Arizona State University
  • Office of Naval Research
  • United States Navy

Tags

Readers

  • Defense Acquisition Program Management
  • Geospatial Intelligence and Artificial Intelligence Analytics
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.