| 摘要 |
Signals emitted by a computer's input devices are captured in chronological order and identified as occurrences of one or more predefined classes descriptive of relevant activity performed during a work session. To achieve human comprehensibility of final analysis but allow for efficient computer processing these class occurrences are assigned both textual names and unique binary tokens. Individual tokens of the various strings are processed to determine the information quantity embodied by each. To appraise implications of serial correlation among tokens, groups of contiguous tokens are analyzed and information redundancy and information balance determined. Resulting data provide the basis for generating quantitative, objective measures of: (1) overall cognitive effort of employing a software, (2) performance of typical users relative to expert users, (3) ease of learning, and (4) inefficiency due to error. These measures permit usability comparisons: between proposed computer-human interfaces; between extant computer-human interfaces; and between computer operators utilizing a given application software to perform a standard activity suite.
|
