One of the clearest trends in the late twentieth century has been the increasing growth and complexity of automation in modern society. The automation of mechanical systems that accompanied the industrial revolution in the nineteenth century was primarily directed at increasing the physical capability of humans with the goal of increasing manufacturing productivity. The automation of information systems in the twentieth century has generally been directed at increasing the cognitive capability of humans, again with the goal of increasing productivity in the information age. Unlike mechanical automation, which reduced human workload, advanced information systems appear to have increased the mental workload through increased complexity.
The aviation community became aware of serious automation problems in 1992 as the result of a mode confusion accident in Strasbourg, France. The Airbus A-320 was the most highly automated civil aircraft flying at the time, with a "fly-by-wire" control system and an advanced Flight Management System (FMS). The FMS is essentially a sophisticated autopilot capable of flying the aircraft along a pre-programmed path from takeoff to touchdown. Investigation of the Strasbourg accident suggested that the flight automation system flew the aircraft at a steeper descent angle than the pilots had planned. Basically, the flight crew thought that they had programmed the aircraft to descend at a 3.2 degree angle in the "Flight Path Angle" mode, but they had actually programmed a 3200-foot-per-minute descent in the "Vertical Speed" mode.
The Strasbourg accident was the leading edge of a wave of automationrelated incidents that occurred in the 1990s. While the overall safety level was improving, the percentage of accidents attributed to human error remained constant at approximately 70 percent. What was changing, however, was the nature of the errors. Previously most accidents were caused by problems with the physical skills involved in flying the aircraft or with errors of judgment. The new problems involved issues of management of the complex aircraft and automation systems. The role of the pilot had shifted from being a manipulator of the controls to being a manager of the aircraft systems. The automation also tended to exacerbate variability in pilot workload, increasing the demands during the busiest phases of flight (takeoff and landing) while reducing the load during the high altitude cruise phase of flight.
STUDYING THE CAUSES
Since the 1990s, we have been involved in investigating the causes of aviation automation problems. We define the mode awareness problem as what happens when the automation executes an unexpected action (an act of commission) or fails to execute an action that is anticipated or expected by one or more of the pilots (an act of omission). Figure 1 shows the results of an analysis of 300 anonymous pilot reports of automation problems sent to the NASA Aviation Safety Reporting System between 1990 and 1994. The data indicate that programming errors, mode confusion, and insufficient understanding of the automation were the leading causes of difficulty among flight crews. If these highly trained, motivated, and alert pilots were having problems understanding their automation systems, there was concern that aviation might be a leading indicator of broader problems with automation.
In a comprehensive technical review of flight automation systems in the U.S. commercial-jet fleet, we found that the current systems were sufficiently complex that neither the pilots nor the designers could provide a simple articulation of the structure of the automation systems. No clear model of the automation was apparent in any training material we reviewed. In the absence of a clear model of the system, the pilots appear to develop ad-hoc mental models based on their experience. While these ad-hoc models work well in normal operations, there is no certainty that they will work well in abnormal or emergency situations. …