EVALUATION OF SYSTEM PERFORMANCE FROM RANK-ORDER DATA

An approximate model is proposed for predicting the rank-order of system failure probabilities. This approximate model, based on a previous exact one, uses rank-order input data. Rank-order form simplifies data gathering while sacrificing only a slight amount of rigor. Further, a wider range of informants may be used to obtain useful system information than when numerical probabilities must be requested.

SYSTEMS ANALYSIS OF DRIVING SIMULATION

System analysis procedures as applied to driving simulators must accomplish several bask tasks: (1) they must identify the system components and interactions that are to be simulated; (2) they must specify the degree of simulation of each component in the system. In this case, the extent to which the automobile and the environment are to be reproduced by the simulator must be adequately established; (3) the input and output to the simulator, plus the process of introducing variability and measuring the response to it, must be subjected to careful analysis; (4) finally, the simulation process requires that a major portion of any driver-related systems analysis be devoted to an examination of each component in the system, its input, functioning, and response characteristics.

The basic components of the driving situation were analyzed from a systems point of view. In order to summarize some existing simulation attempts, the particular focus of each study with respect to the component or method of handling a component, was utilized. The driving simulation literature was reviewed to provide an integrated conceptual framework of the accomplishments of research in this field

The basic component relations were explored and the primary functions of simulation were then analyzed. These primary functions were: (1) the introduction of experimental variation into the system; (2) the representation of the components of the system; (3) the measurement of component response; and (4) the measurement of system performance.

COMPUTER-BASED METHODOLOGY FOR SYSTEM DEVELOPMENT: SITE PRODUCTION AND REDUCTION SYSTEM

A biological model, the nervous system, is applied to a man–machine system in order to stimulate interdisciplinary thought. Conceptual similarities between engineering, communication, and psychology are examined as they relate to the application of this model. If system designers incorporate afferent (Exercise) and efferent (Evaluation) subsystems so that feedback capability exists, system “learning” becomes feasible. This biological model is operational in the Site Production and Reduction System (SPARS), a computer system providing support for system training in the Semi-Automatic Ground Environment (SAGE) System. SPARS problem production (Exercise) and data reduction (Evaluation) capabilities are described and implications for future system technology are presented.

MEASUREMENT OF ANALOG SEQUENTIAL DEPENDENCIES

A statistic sensitive to sequential dependencies existing between responses was proposed. Two studies were undertaken to determine the validity and practicality of this measure. In the first study, results included: (1) the distribution of this measure (λ) is sufficiently normal to permit parametric analyses; (2) λ sensitively reflects both individual differences and the course of learning; and (3) only a low positive correlation exists between λ and average error. The second study, in the area of controller sensitivity, indicated that λ is more sensitive to small changes in task parameters than either average error or root-mean-square. Future applications of the λ coefficient to both analog and digital data were suggested.

PROBLEMS IN EVALUATING SYSTEM MANNING REQUIREMENTS ESTIMATES AND ESTIMATION TECHNIQUES

System manning requirements information is desired prior to the time when it is scheduled to become available under the Personnel Subsystem concept. Rather than proceed from the conclusion that a new manning estimation method is needed to obtain manning information earlier, it is assumed that present methods can be used. Support for the assumption was to be obtained by validating manning requirements estimates for several systems, and assessing the adequacy of the estimation methods. Problems that arose in conducting the study are discussed. The shortcomings of the Mace B and the Bomarc manning requirements analyses are discussed along with the inappropriateness of system test data as criteria against which to evaluate manning estimates. The human factors system analysis approach, modified to account for pertinent quantitative data, is recommended for estimating manning requirements during the initial stage of system development.

THE EFFECT OF FLASH DISTRIBUTION AND ILLUMINATION LEVEL UPON THE DETECTION OF LOW INTENSITY LIGHT STIMULI

The proportion of light flashes detected by naive subjects was determined as a function of two flash groupings and two levels of flash intensity. One flash grouping, the “massed” condition, consisted of two groups of six flashes. The flashes were presented at the rate of one flash per second with about 1° of arc separations. The groups were separated by approximately 90 sec of time and 90° of arc. The second, or “distributed” condition, of twelve flashes presented at the rate of one flash for each 10° of arc and 10 sec of time. The illuminance level of these two conditions was equivalent to 0.13 kmc. One group of subjects was run under the “distributed” condition when the illuminance was increased to 0.935 kmc. There was no significant difference in the proportions of subject detecting flashes where flash distribution was the independent variable. A greater proportion of flashes was seen under the “distributed” condition than under the “massed” condition. More subjects made detections when the stimulus was 0.935 kmc than when it was 0.13 kmc.