Human-Aware Planning & Decision Support for Collaborative Complex Decision-Making
Abstract
Mission planning in modern Navy scenarios such as those faced by PACFLT and PACSAG involves complex decision making with aggregation and analysis of large amounts of constraints and data. Currently, much of this is handled manually, and is thus often prone to human errors due to cognitive overload and loss of situational awareness. Most prior attempts to introduce automation have not paid adequate attention to the human-decision-system interaction, which hindered their adoption by mission planners. It is thus critical that decision support systems be designed from ground-up to be human-aware". Among other things, this will require the decision support systems to effectively track the mental model of the human mission planners, and supporting the systems suggestions and recommendations with explanations that are comprehensible to the humans.In our recent work, we have been making a sustained attempt at developing the foundations of human-aware planning and decision support. Encouraged by promising results from this research, here we propose an ambitious project to develop a naturalistic and human-aware decision-support system for mission planners. The primary challenge we identify is the mental modeling of the human(s) in the loop which enables the decision support system with a suite of capabilities such as the ability to recognize and anticipate human intent while also allowing it to explain its recommendations or even generate suggestions that are easier to understand or explicable to the human. We propose detailed research tasks on (i) extending the traditional automated planning frameworks to handle the differing models of the task that the human planners and decision systems have (ii) developing representations and algorithms to recognize the intent of the human mission planners (iii) supporting multiple human decision-makers in the loop and (iv) techniques for handling model incompleteness and model uncertainty and to learn the models efficiently. With the help of the proposed technologies in human-aware automated planning and decision making, we hope to achieve improvements in speed, efficacy and quality of the complex decision making processes of human decision makers such as in the PACFLT (e.g. event planning for the Pacific Partnership). We also propose to evaluate and demonstrate these improvements through principled human studies, both in synthetic domains, and (unclassified) scenarios of naval interest, including PACFLT and PACSAG.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 17, 2018
- Source ID
- N000141812840
Entities
People
- Subbarao Kambhampati
Organizations
- Arizona State University
- Office of Naval Research
- United States Navy