over of tasks, however, is often disliked.
With increasing level of automation operators should be able to rely on it, as their role becomes less prominent, or is even eliminated. Silent systems that cannot be overruled can easily endanger safety. In the field of car driving, a major risk of new in-car technology is over-reliance on these systems (complacency). Complacency is related to reliability of a system, trust in a system builds up over time after it has proven to display predictable behavior ( Muir, 1987), and it is trust in a system that enhances over-reliance. Negative experience with system reliability, e.g., a large amount of false alarms such as found in aviation in original collision avoidance systems ( Kantowitz & Campbell, 1996), also endanger safety as it will lead to ignoring of warnings. In the above discussed tutoring and automated driving studies evidence for complacency was found. Elderly drivers were inclined to trust the tutoring system to provide them with information on speed limits. In the study on the Automated Highway System (AHS) subjects expected the system to react to an emergency, and more than half of them did not reclaim control or did so in a very late phase. In the monotonous environment of the AHS things can get even worse, as stress and fatigue are factors that may enhance complacency ( Matthews & Desmond, 1995).
Adaptive automation (e.g., Scerbo, 1996), implying that mode or level of automation can be altered in real-time, is mentioned as one of the solutions to overcome out-of-the loop behavior. Research on adaptive task allocation shows that temporarily returning control to the human operator has favorable effects on detection of automation failures during subsequent automation control ( Parasuraman et al., 1996). From aviation it is known that pilots worry about the loss of flying skills and consider manual flying of a part of every trip important to maintain their skills ( McClumpha et al., 1991). In the AHS this could be implemented restricting access to the AHS for a limited period of time, and to require manual driving on a non-AHS lane for some minutes before allowing access to the AHS again. However, apart from practical consequences in terms increased weaving in and out of the AHS lane leading to reduced throughput ( Levitan & North, 1995), many drivers may resist this idea (why drive manually again while all is going well?).
The issue of driver acceptance is more and more recognized. Behavioral freedom is of major importance for driver acceptance. In the end it is not technology that determines how we will drive, but what we are willing to. How far we should go in automation is a social choice ( Sheridan, 1996). Since the AHS demonstration in San Diego in 1997 we know that technically the AHS is possible. As take-over of control is generally disliked the question merely is "do we really want it"? Nowadays, technically almost anything is possible. The different studies discussed in this paper show that it is not what is technically conceivable that should determine what is implemented. Most important is the user's reaction to these systems and their behavioral adaptation to them.
Stephen Fairclough and Frank Steyvers are thanked for giving useful comments on an earlier draft of this paper.
Bloomfield J. R., Buck J. R., Carroll S. A., Booth M. S., Romano R. A., McGehee D. V., & North R. A. ( 1995). Human Factors aspects of the transfer of control from the Automated Highway System to the driver. Report FHWA-RD-94-114. McClean, Virginia, U.S.A.: US DOT, Federal Highway Administration, Turner-Fairbank Highway Research Centre.
Brookhuis K. A. ( 1993). Detection, tutoring and enforcement of traffic rule violations--The DETER project. In Vehicle Navigation and Informations Systems, IEEE-IEE Conference proceedings (pp. 698-702). Ottawa, Canada: Trico Printing.
Brookhuis K. A. ( 1995). DETER, Detection, Enforcement & Tutoring for Error Reduction. Final report.