| Human-Computer Interaction (HCI) research has gained prominence due to the need to make computers easier to learn and use. This research (a) develops a HCI framework to structure and review HCI models, (b) develops a subgoal theory that investigates some pieces missing from current models, and (c) tests the subgoal theory.;This HCI framework combines the Action Theory (Norman, 1986) and the SRK Framework (Rasmussen, 1976) to describe the seven stages of HCI activities and the underlying cognitive processes. These activities are classified as knowledge-based, rule-based, or skill-based behaviors, and they are associated with different cognitive controls, stimuli, and errors. This framework clarifies differences in behavior, states hypotheses for investigation, and structures HCI models. A review of the current models reveals that they seldom address knowledge-based behavior.;This research models knowledge-based HCI activities by investigating the subgoal structure as a cognitive control mechanism to overcome bad interfaces. Users employ extra subgoals both when the task and interface operate on different objects, and when they require a different action sequence for task completion. The bottlenecks in HCI arise from the planning, monitoring, and validating of these extra subgoals.;This research concludes with two studies to demonstrate the subgoal theory's explanatory power. In the formula editor study, subjects used a linear editor and a semantic editor to key in formulas. The linear editor operates on character strings and the semantic editor operates on formula structures. Subjects using the linear editor needed many extra subgoals to translate the mismatched structures, resulting in longer keystroke time, extra keystrokes, and higher error rate.;In the Lotus menu traversal experiment, subjects saw one of two menu organizations and received one of two instruction formats to execute simple spreadsheet tasks. The original Lotus menu was redesigned to have a consistent structure. Subjects using this new menu could form consistent subgoal structures to explore the menu. Task instruction formats were also manipulated to either match or mismatch the designed menu structure. Subjects given the new menu with the matching instructions had the best performance as they did not need extra subgoals to buffer the out-of-sequence task actions. |
