NASA is developing new, more agile technologies for modeling human performance. The interfaces that astronauts use are critical to how quickly, accurately, and safely they can do a job. The time it takes for an astronaut to perform a task on a computer, in the cockpit, or on a planet with a robot may mean the difference between the mission's success or failure.
NASA is developing new, more agile technologies for modeling human performance. The interfaces that astronauts use are critical to how quickly, accurately, and safely they can do a job. The time it takes for an astronaut to perform a task on a computer, in the cockpit, or on a planet with a robot may mean the difference between the mission's success or failure.
NASA is developing new, more agile technologies for modeling human performance. The interfaces that astronauts use are critical to how quickly, accurately, and safely they can do a job. The time it takes for an astronaut to perform a task on a computer, in the cockpit, or on a planet with a robot may mean the difference between the mission's success or failure.
Computing, Information, and Communications Technology Program www.cict.nasa.gov
An Agile Tool for Modeling Human Performance
• CORE (Constraint-based Optimal Shafto. “Therefore, we are looking at the Reasoning Engine) introduces new computational tools, the cognitive and Technology Spotlight flexibility in modeling and pre- social systems, and the physical facilities dicting performance of humans and environments that best enable future Technology interacting with complex systems NASA missions to succeed. CORE—Constraint-based Optimal • It’s easy for non-experts to use, Reasoning Engine “The interface is a critical component of while permitting experts to select human-centered systems,” says Shafto. among cognitive architectures and “Until now, system interface design has Function scheduling algorithms been the domain of specialists in human Enables general users to model inter- performance modeling, but those people active performance of humans using The time it takes for an astronaut to per- are usually not experts in designing and new computer/robotic interfaces form a task on a computer, in the cockpit executing actual NASA missions. We need of a spacecraft, or on a planet with a to find an easier way to inject real mission Relevant Missions robot may mean the difference between experience into interface design, and find • Human-Robotic Exploration the mission’s success or failure, and even more affordable and accurate ways to pre- life or death. The system interfaces that dict the efficacy of these interfaces with • International Space Station astronauts use are critical to how quickly, different users. We don’t have the luxury of • Phoenix 2007 Mars Mission accurately, and safely they can do a job. asking astronauts to spend valuable time • Mars Science Laboratory testing new interface designs until we know To develop and evaluate new system inter- they are mission-ready. To meet this chal- Features faces for future space missions, scientists lenge, Alonso Vera and his team at NASA at NASA are developing new, more agile Ames have developed the Constraint-based • “Guided-entry” interface that technologies for modeling human perform- Optimal Reasoning Engine, or CORE.” prompts users for input of task ance. CICT’s Intelligent Systems (IS) descriptions Project is researching the design, develop- Current challenges to modeling • Flexible choice of cognitive architec- ment, and deployment of complex human- Alonso Vera, principal investigator for tures and scheduling algorithms computer systems through its Human- CORE, says, “Current methods for modeling • PERT chart-style graphical output Centered Computing subproject, managed and predicting human performance either by Michael Shafto. require specialized training or are too rigid Benefits to easily accommodate different possibili- Advancing human-centered computing ties. Most other cognitive architectures • Enables affordable, efficient design “Our research teams are studying how require that the modeler understand cogni- and evaluation of complex interface humans, software agents, and robots all tive psychology and have sophisticated pro- concepts without requiring working contribute to system behavior,” says gramming skills. Yet it can still take these prototypes continued on next page • Supports design of software tools used in time-critical contexts • Predicts performance for range of CORE enables almost anyone to users—novice to expert model new human-machine • Provides graphical report on time to interfaces quickly and affordably. complete task, based on skill level, CORE automatically measures how and probability of user error long it would take for humans • Measures speed/accuracy tradeoffs (from novice to expert) to use the • Quantifies and validates theories of interface to perform a task, as well cognition as the probability of errors. At left is an artist’s rendition of an astronaut operating a drill on a Contacts lunar exploratory mission. • Alonso.Vera@nasa.gov • Michael.G.Shafto@nasa.gov INFUSI N Computing, Information, and Communications Technology Program
environmental, and task constraints. They
CORE produces a chart (left) used that approach to model a staff-sched- that breaks the steps of a task ule data entry task for the Collaborative down into its respective Information Portal, which was used suc- “operators,” showing how long cessfully on a daily basis by NASA person- each takes (in milliseconds). This nel on the Mars Exploration Rover mission. chart shows how an expert begins a second step (green) “By defining these constraints separately before the first (blue) is instead of preemptively binding them to complete—called “interleaving.” each other,” says Vera’s colleague Michael McCurdy, “we allow more flexibility in how experts days to develop a model, whereas used (mouse, keyboard, screen, etc.), and the their relationships can be computed. We also we need to get it down to hours or even cognitive strategy for achieving the task, such maintain their independence from arbitrary minutes. And, even after days of expert as moving the mouse, or turning a knob, or constraints imposed by the machine or the development, these models cannot accom- clicking a button. The cognitive strategy can software algorithms used to model the task.” modate multiple skill levels, different archi- also include slow and fast moves depending tectural platforms, multiple interactive users, on how precise a move must be. As a result, expert designers can mix and or display tradeoffs between speed and match cognitive architectures and descrip- accuracy. NASA needs a tool that is auto- CORE then applies embedded (but modifi- tion languages, choosing from the options mated for use by novices, and is flexible able) rules and constraints to the behavioral in the CORE libraries. enough to provide more output options.” elements of the task described, and gener- ates a graphical chart of the optimal timing “CORE is an agile tool,” says Vera. “It pro- Fast and easy modeling of performance and sequence of steps in the task (see illus- vides novice designers with an easy-to-use Vera and his colleagues at NASA Ames, tration above). CORE also predicts the modeling interface and a clear, quantifiable Irene Tollinger and Michael McCurdy, have resource requirements and total time of the report on performance, while still enabling designed CORE to simplify the process of activity (when the steps are done in a par- expert designers to customize the archi- modeling human performance for novice ticular order). A designer can then revise the tecture to meet their own needs.” designers, while providing more options for model to obtain better performance or experienced designers. adjust the tradeoff of speed vs. accuracy. –Larry Laufenberg “CORE enables almost anyone to quickly CORE provides flexibility and choice For more information or stories online, see model and predict human performance of a A key feature of CORE is its flexibility. The www.cict.nasa.gov/infusion system task,” says Vera. “CORE predicts task cognitive architectures used to model human times according to whether a single expert performance include implicit assumptions or user or multiple users of different capabili- rules about human performance. Take, for ties are performing it. CORE can also pre- example, the simple task of clicking a mouse dict task performance given a user’s deci- on a screen icon. Some cognitive architec- NASA CICT Program Office sion on how he or she wants to trade speed tures define the whole move as a single unit Ames Research Center for accuracy, or vice versa.” and proscribe the average time that will take. 650.604.2494 CORE, however, can break that move into its CORE’s X-PRT user interface constituent “operators”—gaze, perception, Program Manager Irene Tollinger, who is designing the X-PRT cognition, and motor activity—and thus Butler Hine interface for CORE, says, “X-PRT will enable accommodate alternative sequences of oper- even novice users to effectively create, ator events, which enables it to track and Deputy Program Manager debug, and visually verify the performance measure the benefits of expert interleaving. Bill Van Dalsem of their models. As a guided-entry inter- face, it assists the user in describing the CORE’s timing of the operators is based on Infusion Editor essential steps in the task without imposing a cognitive architecture—in this case, Larry Laufenberg a particular sequence on those steps. It’s Card, Moran, and Newell’s Model Human llaufenberg@mail.arc.nasa.gov well known that as users learn a task they Processor—but, if expert designers wish, begin to interleave elements. For example, they can select another architecture from CICT is part of the NASA Exploration Systems Mission Directorate’s Human and Robotic before finishing one step, they might the CORE library, or even customize an Technology Theme which represents NASA's com- glance ahead at something they are going architecture by re-defining the fundamental mitment to investing in the technologies and capa- to need in the next step. An expert user can operators and resources, as well as their bilities that will make an ambitious and sustainable perform the steps differently from a novice. parameters and dependencies. 21st century space exploration program possible. The expert may interleave elements of two different steps.” (See illustration above.) The key is constraint satisfaction www.cict.nasa.gov Vera, McCurdy and Tollinger argue that it is The X-PRT interface enables even novices to faster and easier to predict performance describe the task or set of tasks, the devices from a set of architectural, strategic,