The Use of Logic Trees in Technical Documentation Alternative Diagrammatic Approaches

Recently, AT&T was required to reorganize and redesign many jobs within its organization. The reorganization resulted in the development of a large amount of technical documentation. This documentation provided users with an understanding of both what to do to perform their jobs (modified logic tree format) and how to perform each step in the logic tree (supporting reference material).

We are currently investigating ways of providing this documentation on-line. There are unique problems associated with computerizing this type of documentation. One problem, and the major focus of the present paper, is how to maintain the working integrity of the logic tree structure when using standard CRT formats. We observed that the small amount of information available on the CRT screen disrupted the normal information integration & learning process demonstrated by users of logic trees.

A review of the literature provided little help in resolving this problem. The task, then, was to develop & empirically test techniques to provide a maximum amount of information density on the CRT while capitalizing on the cognitive aspect of information assimilation in the use of logic trees.

Three versions of the same logic trees were tested. The types of logic trees were: 1) standard binary decision trees; 2) rule/exception trees; and 3) compressed “logic trees”.

Subjects were tested invididually and given practice using representative sample materials for each logic tree type. Following practice, subjects were given the experimental logic trees and supporting reference document (the reference document was constant across all conditions). Subjects were given problems to solve using the logic trees. Twenty-seven questions were given in the main portion of the study, nine per logic tree format. The questions were given in a random order with the order of logic tree usage being counter-balanced across subjects. The presentation of the questions response timing, and procedural accuracy was computer controlled. Following the main portion of the experiment, subjects were asked to write their preferences regarding the logic tree formats.

The mean time taken by the subjects to answer questions using a compressed logic tree format was 45.69 seconds; for the rule/exception format it was 52.35 seconds; and for the standard format it was 60.91 seconds. A repeated measures analysis of variance was run on the data, and it was found that a difference did exist among the means (F=3.93, df=2/22, p<05). A comparison between the compressed format and the other two formats showed that the questions were answered more quickly when the compressed logic trees were used. Subjects rated the compact (compressed) logic tree format as an “unfriendly” format and it was the most disliked format of the three.

The surprising data indicates that, for novices, it is possible to provide “compressed” procedural documentation. Presently, research is being conducted to examine the effect of extended practice with compressed logic trees. Those results will indicate the usability of the compressed format across all populations.

The present data is discussed in terms of the impact on current views of how to provide diagrammatic documentation. More importantly, the discussion centers around the need to evaluate the type of task and the length of experience the target population has with given documentation as the driving forces behind diagrammatic documentation design.