Usability Requirements for Virtual Environments

Kulwinder Kaur, Alistair Sutcliffe and Neil Maiden

Centre for HCI Design, City University
Northampton Square, London
EC1V 0HB, UK

ABSTRACT
This paper reports the development of usability requirements for virtual environments and the results of a study to evaluate their impact on interaction success.

INTRODUCTION
Virtual environments (VEs) provide a computer-based interface representing a real- life or abstract 3-dimensional space. VEs offer new possibilities and challenges to human-computer interface design, however, major interaction problems have been found with current VEs, such as disorientation and difficulty finding and understanding available interactions, which result in user frustration and a low acceptability for the VE (Kaur et al., 1996). Guidance is needed on interaction design for VEs to avoid such usability problems.

THEORETICAL RESEARCH
To inform interaction design guidance, models of interaction for VEs were developed, by elaborating Norman's (1988) general cycle of interaction. The models consist of twenty-one inter-linked stages of activity, describing task-based, exploratory and reactive behaviour, to system-initiated events. The models were evaluated in user studies of interaction behaviour, using verbal protocol analysis (see Kaur et al., inpress). The models were then used to define design properties required to support the user during identified stages of interaction. The properties (46 in total) cover various aspects of a VE: the user task, spatial layout, viewpoint and user representation, objects, system-initiated behaviour, actions and action feedback. For example, the property identifiable objectstates that an object should be easy to identify or recognise, where copied from real world phenomena.

STUDY METHOD
A controlled study was carried out to evaluate the impact of the design properties on interaction success. Eighteen subjects took part in the study and were balanced, according to experience into a Control (C) and Experimental (E) group. The control group was given the original version of a virtual business-park application and the experimental group was given a version with some of the missing design properties implemented. For example, figure 1 shows changes made to the application to implement the properties: distinguishable object, identifiable object, clear navigation pathwaysand declared available action. In the amended version, objects such as walls sharing an edge were made more distinguishable by using textures to emphasise edges; exit doors were made easier to identify by labelling them; areas that the user could not navigate into were marked using 'no-entry' signs (which appeared on approach); and actions to provide information about basic facilities (e.g. lighting) were clearly cued with information signs.

Figure 1: Implementation of four design properties: distinguishable object(top),
identifiable object (bottom left), clear navigation pathways(bottom
left) and declared available action(bottom right).

Subjects interacted with the virtual business-park to complete nine varied tasks, including familiarisation and exploration, investigating windows, opening a loading bay and comparing toilets in a building. Subjects were asked to provide a concurrent, 'think-aloud' verbal protocol and their interaction sessions were video-recorded. Following interaction, subjects completed a memory test on the business-park.

RESULTS AND DISCUSSION
There was a general improvement in interaction with use of the amended version (one-tailed t-tests used for following statistics). The experimental group encountered significantly fewer usability problems (p<0.01; avg. C=134, E=45 problems per subject) and successfully completed significantly more tasks (p<0.01; avg. C=7, E=8.4 tasks). The experimental group also achieved higher scores for the memory test and this difference approachedsignificance (p=0.064; avg. C=46, E=52%).

The results are encouraging and show a 66% reduction in usability problems, leading to subjects being able to complete their tasks better and gain more useful information during interaction. The results appear to indicate that the proposed design properties are important requirements for successful interaction, and a VE interface can be significantly improved by implementing missing design properties. Our future work involves refining the set of design properties in light of detailed results and translating them into concrete guidelines for VE designers. Our overall goal is to address problems of interface design for VEs, using interaction modelling as a theoretical base.

ACKNOWLEDGEMENTS
We thank VR Solutions and The Rural Wales Development Board for loan of the test application and the EPSRC for funding.

REFERENCES
Kaur K., Maiden N. and Sutcliffe A. (1996). Design practice and usability problems
with virtual environments. IMAGE imgs/HCI98CC38.gif
IMAGE imgs/HCI98CC39.gif (IDG Conferences).
Kaur K., Maiden N. and Sutcliffe A. (inpress). Interacting with virtual environments:
an evaluation of a model of interaction. Accepted for publication in Interacting
with Computers: VR special issue.
Norman D.A. (1988). The psychology of everyday things. (New York: Basic Books).