Leveraging Human-Computer Collaboration for Decision Making in Complex Systems Mary (Missy) Cummings Humans & Automation Lab Aeronautics & Astronautics MissyC@mit.edu
Focus Areas Supervisory Control Humans vs. automation in complex systems Mixed initiative approach to decision making How to apply interactive decision aiding strategies to allow a human to explore the automation decision space? Information Visualization Cost functions, constraints & variables Sensitivity analysis Smith, Layton, McCoy My research Mica and Kaber Lower sometimes better but there has to be another way!
Human Supervisory Control Controls Sensors Computer Task Human Operator (Supervisor) Displays Actuators Planning a computer-based task Communicating to the computer what was planned Monitoring the computer’s actions for errors and/or failures Intervening when the plan has been completed or the computer requires assistance Human & computer learn from the experience Bandwidth Trust Machine/computer metaphors
Research Motivation: Tactical Tomahawk Project
Proposed Tactical Tomahawk Missions Primary (Default) Target Default Mission Flex Mission Preplanned Health and Status points Emergent Mission Guidance Waypoint Alternate (Flex) Target Branch Point Default Target Loiter Pattern Time-critical (emergent) Target Launch Basket
Automation Description Sheridan & Verplank’s 10 Levels of Automation Automation Level Automation Description 1 The computer offers no assistance: human must take all decision and actions. 2 The computer offers a complete set of decision/action alternatives, or 3 narrows the selection down to a few, or 4 suggests one alternative, and 5 executes that suggestion if the human approves, or 6 allows the human a restricted time to veto before automatic execution, or 7 executes automatically, then necessarily informs humans, and 8 informs the human only if asked, or 9 informs the human only if it, the computer, decides to. 10 The computer decides everything and acts autonomously, ignoring the human.
Supervisory Command & Control Operators effectively controlled up to 12 missiles Original “guestimate” was 4, FAA results similar Automation bias and communication management were issues
Human Supervisory Control & Network Centric Warfare Appropriate levels of automation Information overload Adaptive automation Distributed decision-making through team coordination Complexity measures Decision biases Attention allocation Supervisory monitoring of operators
Information Overload
Adaptive Automation Dynamic role allocation System constraints Mixed initiatives System constraints Skill Rule Knowledge-based behaviors Cueing mechanisms Psychophysiological Noisy Decision theoretic
Complexity Measures in HSC Complexity of Environment Complexity of Goals Complexity of Procedures Complexity of Displays Cognitive Complexity Direct relationship Indirect relationship
Human Interaction with Autonomous Vehicles Human interaction with anytime path planning algorithms Time-critical domains Windows into automation Multiple vehicle task management decision support Levels of automation Preview times Stopping rules Swarming behavior
Command & Control: Rapid Replanning
Resource Allocation
Formation Flying in the Future