The micro-structure of use of help

To see whether, from the user's standpoint, muddling through with trial-and-error exploration of the interface actually works as well as consulting help, we analyzed video recordings of 14 subjects using Microsoft Publisher. We segmented their attempts at accomplishing four tasks into episodes, lasting from a few seconds to a few minutes, using four problem-solving approaches: help, recall, and trial-and-error. The subjects' overall rates of success with trial-and-error were higher than with help. We present trends of use of the approaches across the four tasks, review the apparent causes of success and failure when using the approaches, and develop an affordance-knowledge model that helps to explain people's preference for muddling through over using help. We conclude by discussing the model's implications for developers of systems and writers of help.