Nagaoka University of Technology
   
 

--

Kitajima & Polson (1992)

Kitajima, M. & Polson, P. (1992). A process model of display-based human-computer interaction. CHI'92 Research Symposium.

 

A process model of display-based human-computer interaction

Recent empirical studies on display-based human-computer interaction have provided evidence against standard plan-based theories (e.g. Card, Moran, and Newell, 1983; Kieras and Polson, 1985) of expertise in HCI. Mayes, Draper, McGregor and Oatley (1988) report that experienced MacWrite users have poor recall for the names of menu-items. In addition, Payne (1991) has shown that experienced users do not have complete knowledge about the effects of commands.

These results provide support for theoretical frameworks that assume that sequences of user actions are not pre-planned. Each action is determined making use of display feedback during the course of generating a sequence of action necessary to complete a task. The display plays a crucial role in successful and smooth interaction; the interaction is truly mediated by the display. Howes and Payne (1990), and others have developed theories of skilled performance in which successful interactions are mediated by representations of intermediate states of a task presented in a display. Larkin (1989) called her framework display-based problem solving.

Howes and Payne (1990) extended the task action grammar framework to display-based, menu systems. D-TAG (display-based task-action grammar) is a competence model of users' knowledge of display-based systems used for evaluating the consistency of an interface.

In this paper we outline a process model for display-based HCI based on Kintsch's (1988) construction -integration theory. In a companion paper (Kitajima and Polson, 1992) we describe an early version of the model and the results of three series of simulation experiments. This paper presents an overview of our developing theoretical framework. Our long term goal is to account for both routine, skilled use of a software tool and learning to use a new tool by exploration.

The CI model was originally proposed by Kintsch (1988) as a model of text comprehension. Text comprehension is modeled as a two-phase process of construction and integration. The construction process generates associative networks by using input text as the cues to retrieve relevant knowledge from long-term memory. The construction process is bottom-up; the resultant network includes both contextually relevant knowledge and associatively related knowledge inconsistent with the context. The integration process uses a spreading activation mechanism to extract a contextually relevant consistent interpretation of the text in the form of an activation pattern over the network. The CI model has been generalized to HCI by using the comprehension processes to select an action sequence with the aid of rich knowledge about commands (Mannes and Kintsch, 1991; Doane, Kintsch and Polson, 1990).

 

Download