Interfaces and Models for Human-in-the-loop Supervisory Control
Humans increasingly use computing systems not only to store information and perform calculations, but to sense, explore, and control the real world. Systems in energy, health care, manufacturing, and vehicle-based exploration and transportation have all seen increasing automation in recent decades. However, the real world is dynamic and ever-changing; safety-criticality and rapidly-changing objectives still require a human operator or administrator to remain in the loop to solve problems in cases of emergency. These users require reactive, real-time command and control to overcome bugs and "brittleness" inherent in automated systems.
In these environments, human operators follow written procedures designed to make them act consistently and predictably. As "human software" written in natural language, procedures are not always executed as intended; humans are inventive and creative problem solvers, but they are also prone to errors in interpreting instructions. I will discuss models we are developing to understand procedure-guided user actions using statistical natural language processing methods. These are applied to understanding human subject behaviors in computer-based training for nuclear power plant control room operations. I believe the results may provide insights generally for the effectiveness of online learning. I will also discuss our work to build richer human interfaces to two robotic motion planning algorithms, the Rapidly-exploring Random Tree (RRT) algorithm and the Spectral Multiscale Coverage (SMC) algorithm.
Bio: Jamie Macbeth is a postdoctoral research associate at the Humans and Automation Lab at MIT and holds a Ph.D. in Computer Science from UCLA. He has a wide range of experience in software engineering and UI design: work on listening and communication training for the hearing impaired at Neurotone Inc., on professional audio mixing interfaces at Euphonix (now a division of Avid), and on digital media distribution at Sony Music Entertainment and Liquid Digital Media. Prior to that he earned an M.S. in Physics from Stanford University and a B.S. in Physics from Brown University. His broad interests include human- computer interaction for supervisory control systems and human factors of software engineering.