Suchergebnisse

There have been two major criticisms of research on vigilance decrement: (1) that there are few troublesome decrements in the real world and (2) that laboratory research on vigilance, often done with comparatively simple tasks, generalizes poorly to complex tasks in operational situations. The first criticism is supported by research that has failed to find vigilance decrement in operationally relevant situations. The second criticism involves the relationship between basic and applied research. Applied scientists who face immediate practical problems should solve them with quasi-realistic simulation. Simulation-based research has limited generality, however, and should not be seen as a substitute for more general solutions that come from basic research and the variables and laws that it establishes.

The correct premise of the topic of human reliability is that system personnel fail in their responding just as equipment does, and so they must be considered along with the equipment in the specification of system reliability. The goal of human reliability efforts as they are presently pursued is to find measures of the reliability of human performance that are expressed in the same terms as measures of equipment reliability and can be combined with them to produce system reliability. The thesis of this paper is that conceptualizing human reliability in these terms raises methodological problems that are likely to prevent this goal from being achieved. What is lacking are: a definition of human failure, units of human behavior whose reliability can be determined, a way to synthesize the reliability of larger behavioral sequences from the units if we could specify them, and a way to integrate human reliability, if it could be determined, with equipment reliability. Monte Carlo modeling is seen as a promising approach to the topic.

The theme of the paper is ways of evaluating flight simulators for aircrew training. The transfer of training experiment and the rating method are the two present-day ways of evaluating the worth of a simulator. The transfer experiment requires the trainee to practice in the simulator and then be tested in the parent aircraft to demonstrate the training value of the simulator. The rating method requires the pilot to be experienced in the parent aircraft and to rate the simulator for similarity to the aircraft. If similarity is high, the training value is assumed to be high. Arguments are presented that both of these methods are flawed. It is contended that a simulator, or any other system, need not necessarily be tested if it is based on reliable scientific laws and the success of other systems, based on the same laws, has been high. Good laws produce accurate prediction, and when outcome can be predicted it is redundant to conduct a system evaluation. With uncertainty about laws, the requirement for system testing increases. The psychological principles underlying simulators are reviewed.

The effects of practice and exposure time were studied for circular and linear (vertically arrayed for this study) dials of the moving pointer type. Type of dial was a between-groups variable, with one group of 11 subjects having the vertical dial and another group of 11 subjects the circular dial. Practice for five daily sessions, and exposure times of 100 or 500 msec, were within-groups variables. The circular dial was superior to the vertical dial for both exposure times, whether per cent correct or amount of reading error was the measure. The practice variable had a minor effect and was statistically significant only for the circular dial at 100 msec. exposure time, and only for the per cent correct measure. Practice tended to alter the distribution of errors for the circular dial.

An experiment was performed to assess the effects of visual display mode and 6 hr of monitoring on performance in a complex vigilance task. The task had 12 stimulus sources arrayed over 60 degrees, and alphanumeric signals that persisted for 6 sec. Each group of 15 subjects had a different display configuration: normally off, normally on, and normally on with noise. Display mode influenced overall mean performance, but not vigilance decrement. The amount of vigilance decrement was small despite the long session, and its magnitude was essentially the same as previous studies whose sessions were two-three hours duration. Earlier conclusions about the triviality of vigilance decrement when tasks are complex were supported.

An experiment was performed to test the hypothesis that knowledge of results (KR) in a vigilance task can be used as a training technique that will transfer positively to subsequent non-KR sessions where feedback is absent. Three groups, fifteen subjects each, participated in a visual monitoring task for four sessions of three hours each. One group received KR about response proficiency after each response in the first two sessions and a second group received only a neutral remark each time from the experimenter. The final two sessions were standard vigilance tasks without feedback. A third control group had the standard task throughout. The KR group had reliably superior performance on all sessions although within-session vigilance decrement was uninfluenced. The neutral stimulation group did not differ from the control condition. It is suggested that KR administered in a training situation can produce improved vigilance during actual system use where KR is absent.