Haifa, Israel. MODES IN AUTOMATED COCKPITS: PROBLEMS, DATA ANALYSIS, AND A MODELING FRAMEWORK

Today’s modern automated flight control systems employ a variety of different behaviors, or modes, in order to allow the crew flexibility in operating the aircraft. These modes lay along a continuum starting from fully-manual to fully-automatic control, with many levels in between. While developments in cockpit automation resulted in workload reduction and economical advantages, they also gave rise to a special class of human-machine problems. These ill defined problems are sometimes referred to as “automation surprises.” These phenomena usually involve confusion about the status of the autoflight control system—in particular its modes—and the subsequent behavior of the aircraft. Recently, there have been five fatal airline accidents involving, in one way or another, this class of human-machine problems. Our research involves a two-way approach: (1) summary analysis of mode usage data, and (2) the development of a methodology for describing human-automation interaction. The mode usage data was collected during a field experiment that documented how pilots, flying a Boeing B-757/767, engaged the various modes in the automated flight control system. Analysis of the data indicated which modes are commonly used, and the different paths that pilots use while transitioning from one mode to another. Based on these and other analyses we developed a modeling framework for describing human interaction with automated control systems. The framework allows us to formally describe a given system and to trace down the features in the control system or interface design that harbor the potential for mode confusion. One example of mode confusion is described and modeled. The particular design feature that led to this mode confusion is identified and explained. The implication of this framework and our research approach for mapping human-automation interaction are briefly discussed.- 2