Raja Parasuraman Catholic University of America, Washington DC
Mustapha Mouloua University of Central Florida, Orlando
Brian Hilburn National Aerospace Lab, Amsterdam
Adaptive automation represents an alternative design approach to the implementation of automation. Computer aiding of the human operator and task allocation between the operator and computer systems are flexible and context-dependent. In contrast, in static automation, provision of computer aiding is predetermined at the design stage, and task allocation is fixed. Several conceptual and theoretical papers have suggested that adaptive systems can regulate operator workload and enhance performance, while preserving the benefits of static automation ( Hancock, Chignell, & Lowenthal, 1985; Parasuraman, Bahri Deaton, Morrison, & Barnes, 1992; Rouse, 1988). The performance costs of certain forms of automation--overreliance, reduced situation awareness, skill degradation, etc. ( Parasuraman & Riley, 1997)--may also be mitigated. These suggestions have only recently been tested empirically ( Hilburn, Jorna, Byrne, & Parasuraman, 1997; Parasuraman, Mouloua, and Molloy, 1996; Scallen, Hancock, & Duley, 1995; see Scerbo, 1996 for a review of earlier work).
Empirical evaluations of adaptive automation have focused primarily on the performance and workload effects of either (1) adaptive aiding of the human operator or (2) adaptive task allocation (ATA), either from the human to the machine (ATA-M), or from the machine to the human (ATA-H). Each of these forms of adaptive automation have been shown to enhance human-system performance, but independently in separate studies. For example, in an early study, Morris and Rouse ( 1986) showed that adaptive aiding (AA) in the form of target localization support enhanced operator performance in a simulated aerial search task. Benefits of AA in a more complex simulation were reported by Hilburn et al. ( 1997), who provided air traffic controllers with a decision aid for determining optimal descent trajectories--the Descent Advisor (DA) of the Center Tracon Automation System (CTAS), an automation aid that is currently undergoing field trials at several air traffic control centers ( Wickens, Mavor, Parasuraman, & McGee, 1998). Hilburn et al. ( 1997) found significant benefits for controller workload (as assessed using physiological measures) when the DA was provided adaptively during high traffic loads, compared to when it was available throughout (static automation) or at low traffic loads. With respect to adaptive task allocation from the machine to the human (ATA-H), Parasuraman et al. ( 1996) showed that temporary return of an automated engine-systems task to manual control benefited subsequent monitoring of the task when it was returned to automated control.
For adaptive systems to be effective, both AA and ATA-H need to be examined jointly in a single work domain. Furthermore, if adaptive systems are designed in a manner typical of "clumsy automation "--i.e., providing aiding or task reallocation when they are least helpful ( Wiener, 1989)--then performance may be degraded rather than enhanced (see also Billings & Woods, 1994). One of the drawbacks of some flightdeck automated systems--for example, the Flight Management System (FMS)--is that they often require extensive reprogramming and impose added workload during high task load phases of flight such as final approach and landing, while doing little to regulate workload during the low-workload