The authors of this paper describe a set of experiments to study the effect on a user of placing links to deep knowledge and explanations on the various screens of a knowledge-based system.

The broad hypothesis is that a user will access background knowledge more readily if the links are placed in context, during problem solving, rather than in a more traditional help system, where effectively, all links are provided from every screen. Through having the ability to get help in context, the user can improve both performance in using the program and learning of concepts where the system demands it. The experiments are well thought out, and were carried out with full control of independent variables where possible. The results bear out the hypotheses.

This paper is of interest to anyone writing user interfaces and program documentation, especially where the system carries the user through a sequence of steps, where each step has to be justified. The mention of hypertext as a knowledge representation scheme is unnecessary, since the paper stands without it, and in any case is not usually included in discussions of such schemes.