Academic journal article Human Factors

Effects of a Psychophysiological System for Adaptive Automation on Performance, Workload, and the Event-Related Potential P300 Component

Academic journal article Human Factors

Effects of a Psychophysiological System for Adaptive Automation on Performance, Workload, and the Event-Related Potential P300 Component

Article excerpt


Recently, there has been much interest in adaptive automation or systems that can adapt to a user's changing requirements in real time. Further, changes in the state of the system can be initiated by either the operator or the system itself (Hancock & Chignell, 1987; Rouse, 1977; Scerbo, 2001; Scerbo, Freeman, & Mikulka, 2000). It has been argued that systems using adaptive automation can provide a better match between the level of operator workload at any moment and the level of automation (Parasuraman, Bahri, Deaton, Morrison, & Barnes, 1992).

Parasuraman et al. (1992) described several techniques that could be used to trigger changes among modes of automation in an adaptive system. For instance, changes among modes could be tied to the presence of specific tactical events that occur in the task environment. Alternatively, models of operator performance or workload could be used to drive the adaptive logic. Another approach would be to assess operator performance in real time and use deviations from acceptable ranges to invoke the automation.

The last method for implementing adaptive automation described by Parasuraman et al. (1992) uses psychophysiological measures to trigger changes among the modes of automation. There are several advantages for using psychophysiological indices in an adaptive system (Byrne & Parasuraman, 1996: Gomer, 1981; Scerbo, Freeman, Mikulka, Parasuraman, Di-Nocero, et al., 2001). First, psychophysiological measures can be obtained continuously and do not necessarily require an overt response from the operator. This is important because in many current systems the operator is often placed in a supervisory role and is rarely required to make any overt responses (e.g., button presses), even though he or she might be engaged in considerable cognitive activity. Also, psychophysiological measures may provide additional information beyond behavioral measures when the two measures are used together. Further, psychophysiological measures have the potential to reflect the state of the operator as well as those functional areas of the brain that are differentially active. Consequently, it is this approach to adaptive automation that is the focus of the present paper.

A Biocybernetic Adaptive System

Pope, Bogart, and Bartolome (1995) described a system in which changes between modes of automation were triggered by the operator's own electroencephalogram (EEG) signals. More specifically, Pope et al. used ah index of engagement based on ratios of EEG power bands (alpha, beta, theta, etc.). The engagement index is based on the idea that the beta band reflects increases in arousal and attention, whereas the alpha and theta bands reflect decreases (Abarbanel, 1999). Moreover, the index captures the simultaneous changes in beta activity along with changes in alpha and/or theta activity in a single variable. Several investigators have reported that EEG power band ratios may be better at distinguishing among different levels of attention than is any single power band in and of itself (Lubar, 1991; Lubar, Swartwood, Swartwood, & O'Donnell, 1995; Streitberg, Rohmel, Herrmann, & Kubicki, 1987).

Pope et al. (1995) studied several different engagement indices from a variety of sites. The engagement indices were computed using a 40-s moving window that was updated every 2 s. Their participants performed the Multi-Attribute Task (MAT) battery (Comstock & Arnegard, 1991), a PC-based suite of activities that includes compensatory tracking, monitoring, and resource management tasks. All of the tasks remained in automatic mode except the tracking task, which shifted between automatic and manual modes.

Their participants performed the task under both positive and negative feedback conditions. Under negative feedback, when the slope of the index function derived from two successive windows was positive (reflecting an increase in engagement), the tracking task was switched to or maintained in automatic mode. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.