Click for next page ( 23

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement

Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 22
FUTURE METHODS Although we have catalogued a variety of methods to be used in the software design and research process, some needs for information are still unmet. The research needs fall roughly into three categories of needs:. new theories, new representations, and new data collection and analysis methods. TERRIES Three particular kinds of theories are seen as needed Automation theories would tell us what should be auto- mated and what should be assigned to the human processor Such theories would also prescribe an appropriate mix of automation and human control. Some needs of theories art suggested in the field of supervisory control and in office analysis techniques, but a more explicit theory it needed to prescribe the best mix of human and computer processing. Theories of individual differences would tell us about the different kinds of computer support required and desired by different user populations. Special continuing interest focuses on the differences between naive or casual users and expert or dedicated users. Theories of standardization would tell us about which aspects of a system should be standardized for all users (as in the basic control devices in an automobile) and which can be customized for adaptation by and for specific users. In addition, two taxonomies are needed: a character- ization of the kinds of tasks for which software can be built (so that design prescriptions can ' tee tied, perhaps to particular classes of tasks) and a characterization c 22

OCR for page 22
23 the kinds of users that use software applications (related to the theories of individual differences described above). The partial taxonomy of human-computer interface tasks advanced by Lenorovitz et al. (1984) provides a baseline for this effort. REPRESENTATION Many of our analyses outside the testing of a working system with real end users require some specification of what the system can do, what the user knows about how the system works, and how the user conceives of the task. There is thus a need for better representational schemes than those now being used. One such scheme would describe a complex system so that documentation and training could be better designed. Another would represent exactly how a system works--the interface, dialog, communication, or transaction--so that the design could be both analyzed for its fit to users' needs and capabilities and conveyed to those who have to program it. We need techniques for inferring what a user currently understands of a system, a method for extracting the appropriate information from the user and for displaying the resulting understanding or Mental model.. These techniques are as useful in basic research on the per- formance of complex tasks as they are in the applied design process. (A report of the Committee on Human Factors' workshop on mental models in the use of information systems is scheduled for publication in 1985.) DATA COLLECT JON, - ~ES, Ad MUSES Although we have a rich variety of measures to collect from users interacting with a system, we have no direct measures of the user's affect nor do we collect any of the neurophysiological responses that accompany intense work, frustration, and satisfaction. In addition, there is a need for better hardware tools for collecting logging and metering information without slowing the system that the user normally interacts with. More specific methods are needed for analyzing the mountain of data that comes from protocol analysis, not only in deducing how the user i s satisfying h is or her task goals and subgoals, but also in deducing ongoing memory and perceptual loads on the user and how the user compensates for them in per-

OCR for page 22
24 forming the task. Our task analysis methods need to be expanded to include more cognitive aspects of the user's performance, his or her memory, language, and perceptual aspects. Research methods considered most likely to produce high payoff in the near future include: o Representations of the users' understanding of a systems 0 Representations of a dialog to convey the design to programmers; o More comprehensive task analyses that include memory, perceptual , and language considerations as well as timing and error predictions" and o Hardware advances that allow the collection of logging and metering data for tapping the current use of a system.