Jon Marsh, Christian Heath, Mike Fraser, Steve Benford, Chris Greenhalgh
An increasing number and variety of organizations are finding innovative uses for virtual reality (VR), uses that encompass such fields as design, engineering, and medicine. Indeed, in future years, virtual scenes could well become everyday "workplaces" for remote colleagues. For example, an architect could remotely guide clients around a virtual building design, or doctors could remotely discuss surgical techniques over a virtual body. However, most technologies to support real-time, distributed collaborative work have not yet proved to provide satisfactory domains for remote working. It has been argued that this weakness derives from the inability of systems to assist individuals in working flexibly with a variety of workplace objects (paper documents, images, etc.). This would seem critical to the development of virtual environments in which remote colleagues would need to discuss virtual objects and scenes and also to collaboratively coordinate their navigation around and within the world. Therefore, to realize the vision of a VR workplace, we believe that we need a richer understanding of the organization of interaction between individuals in virtual worlds and the resources they require to take part in "seamless" interaction around objects and artifacts.
Intriguingly, it is often assumed that problems of recognizing what views and scenes are available to the other are naturally overcome in 3-D worlds. However, we undertook trials with a desktop collaborative virtual environment (CVE) called MASSIVE-2 and found various problems faced by participants even when referencing simple virtual objects. We used a basic design task in order to encourage object-focused interaction, in which participants were encouraged to collaboratively arrange the layout of furniture in an application we called Furniture World (see Figures 1 and 2).
The difficulties revealed by our observational analysis include the following:
- The constrained field of view possible on a desktop monitor (55 to 60°) provides a "blinkered" view of the virtual world. As a result, participants encounter the world in "fragments"; they see an object or an acting avatar (e.g., pointing) but rarely a coherent scene in which both are visible and can be easily interrelated. So, in cases where an object is referenced, there are examples in which participants rotate their view 360° in order to find the other's pointing avatar only to return to look at the very same object that they were facing in the beginning.
- Participants have problems accurately assessing and monitoring the perspectives, orientations, and activities of their colleagues even when the other's embodiment is in full view. Confusion arises from the difference between the actual field of view for a participant and that anticipated by observers when looking at their pseudo-humanoid embodiment. This reduces opportunities for participants to design and coordinate their actions for others even when the other's avatar is visible on-screen.
- The difficulties of establishing mutual orientation to particular objects distract the participants from arranging furniture in the world. As a result, securing adequate reference to objects becomes a topic in itself rather than merely a resource for task production.
In the light of these observations, the paper reconsiders the design of the virtual world. For example, we question our use of pseudo-humanoid embodiments. The findings suggest that the straightforward translation of human physical embodiments into CVEs, however detailed, are likely to be unsuccessful unless participants can also be provided with the perceptual capabilities of physical human bodies, including a very wide field of view and rapid gaze and body movement. We argue that, to compensate for the curious views on, and possible behaviors within, virtual environments, more exaggerated embodiments are required. Moreover, in mimicking interaction in the material world, CVE designers currently tend to present actions (e.g., pointing gestures) solely on the acting avatar. However, our findings would suggest that actions may best be represented across disparate features of the environment to improve access to the sense of an action.
The paper provides further details on how the trials were undertaken, the research methods adopted, specific examples of problems encountered by participants, and the proposed alterations to the system. More generally, the paper begins to question the relevance of the vision of VR to simulate the physical world and highlights the problems that such an approach can have for social interaction when the illusion of a common virtual world is encountered through the physical constraints of current VR interfaces.
Jon Hindmarsh and Christian Heath
Mike Fraser, Steve Benford, and Chris Greenhalgh
University of Nottingham
Article published in ACM Transactions on ComputerHuman Interaction 7, 4 (December 2000).]
©2001 ACM 1072-5220/01/0500 $5.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2001 ACM, Inc.