Beckett Park, Leeds LS6 3QS, UK

Life-long learning and training-on-the-job is of increasing importance in today’s society. An exponential expansion of distance learning is expected within a decade. The British Open University, for example, declares it attempts to create the "Big Bang" in tele-education for its students [Hawkridge 1995]. The recent rapid increase in information sources and computer resources available for education may have the pedagogical potential to offer a new kind of learning experience, particularly in terms of the Human Computer Interface employed. The contention of this paper is that this may be achieved through a hybrid design incorporating the evolving technologies of avatars and agents to establish a continuous presence of users in CVEs.

A virtual environment (VE) is an interactive simulation of three-dimensional structures in a virtual world. It can function as a tool for visualising information. A VE supports natural aspects of human perception by extending visual information in three spatial dimensions. A central component of any VE system is the ability to interact by direct manipulation [Mann, Mon-Williams 1996]. The user-interface of a collaborative VE (CVE) must not only allow engagement with objects but, additionally, communication with other users, since by definition, a CVE actively seeks to support human-human communication in addition to human-machine communication. A virtual body, known as an ‘avatar’, is the interaction device in a CVE.

In the field of education, there are applications that benefit directly from breaking the bounds of two-dimensional representation (e.g. medical training, architectural visualisation, flight simulation, molecular models) because of their natural affinity with 3D-objects. Other educational applications of CVE technology could benefit indirectly, by offering the student a stimulating shared learning environment, in which various forms of interaction and communication methods are available.

Interaction is of key importance for the process of learning. There are three general types of interaction in education, characterised by interaction taking place between student-content, student-teacher, and student-student [Moore 1993]. The interaction between student and learning content is well researched. Frameworks exist for analysing educational media in terms of how they support the teaching-learning process and the different modes of learning [Laurillard 1993]. The ‘guided discovery approach’ (see figure 1) is widely valued in the academic community to support reflection and therefore the integration of the students experiential and conceptual knowledge [Laurillard 1995]. The student-teacher as well as the student-student interaction can be described as communication between the participants in the teaching-learning process. As humans, we do not rely upon a single mode of communication in order to impart information to others, we use a myriad of devices and channels choosing that which is most appropriate at the time. Factors affecting our preferred choice of communication include to whom we are communicating, the willingness to communicate, communication methods available, spatial proximity and time [Morphett 1996].

In a CVE some students may create tools or objects (e.g., work aids, idea prototypes, discussion maps) that everyone uses and therefore contribute towards a common goal. Those responsible for creating and administering learning environments are urged to put more emphasis on students learning from each other, since there is good evidence that in such systems less well qualified students can learn from the better qualified and the latter can also gain valuable skills [O’Malley 1995].

Computers increasingly facilitate co-operative activity between users. This has lead to the development of interfaces to allow Computer Supported Collaborative Learning (CSCL). It has been argued that three-dimensional CSCL is best conducted in a shared virtual environment: since in this setting computers can provide the same type of collaborative information that people have in face-to-face interactions, such as communication by object manipulation and gesture [Wexelblatt 1993].

Figures 2 and 3 illustrate the process of learning through guided discovery within a CVE. The learning content in the virtual world is teacher constructed. Interaction with the content as well as communication with other participants is avatar-mediated. By reflecting on experienced interactions with the content and by reflecting on student-teacher and student-student discussions a student builds up a model of the virtual world and self and adapts his actions accordingly. This model of the virtual world and self is part of the student’s model of the real world and self and therefore becomes part of a students conceptual knowledge. Figure 3 shows how teacher, student and co-students are interconnected within the CVE.

What distinguishes CVE technology from all preceding technology is the sense of immediacy and control created by immersion: presence or the feeling of ‘being there’ [Psotka 1995]. The engagement and excitement that is part of the VE phenomenon is an obvious benefit of immersion in education and training [Bricken, Byrne 1993]. Considering immersion as the ‘key added value’ researchers have just begun to analyse what immersion is, what cognitive variables are connected to immersion, how it is generated in multi-user VEs, and what its benefits for education and training are. Immersion is the cognitive state of being mentally absorbed. The state of cognitive immersion is an attitude based on a temporary willingness to accept the illusion offered by some medium as real. It is a suspension of disbelief. People can become immersed in a movie or a book or any other medium, by some kind of interaction with this medium. Based on the Theory of Distal Attribution [Loomis 1992] four variables to the experience of immersion can be identified (see figure 4). The Theory of Cognitive Immersion states that the degree of immersion which a user will be able to experience is dependent on the quality of the avatars and interactivity, which are factors of the CVE, as well as on the ability to become absorbed in the activity and the amount of experience with the virtual world, which are user characteristics [Tromp 1995].

When information about an environment is presented to an individual, that individual may have the sense of being present in that environment to a greater or lesser extent. Presence is the effect of immersion and refers to the specific sense of self-location in an environment.

Co-presence describes the awareness of the existence of others within an environment. It has been argued that there is an inherent logical connection between the degree of presence and the virtual body [Slater, Usoh 1994]. Furthermore, we not only identify with our own body we also recognise the existence of others through their bodies. Avatars can provide presence and social facilitation of all participants within a CVE.

If members of a group are not co-present, there is a lack of community feeling [Huxor 1998]. This leads to a demand for co-presence and presence-in-absence within CVEs. There has been little research into the psychological dimensions of co-presence and its relation to the process of cognitive immersion. However, a permanent representation, resulting in continuous presence of teacher and fellow students seems to be very important for students within educational CVEs; it would give other participants a persistent proxy for communication and collaboration.

Looking at the process of cognitive immersion, it is obvious that embodification is highly significant for the psychological phenomenon of presence. Further, presence itself seems highly significant for the process of learning in CVEs. Aspects, relevant to embodification within CVEs, of the concepts of avatar and agent technologies are introduced in brief, followed by a discussion of the potential benefits of a hybrid system of these technology.

An avatar is a proxy for the purposes of simplifying and facilitating the process of human communication. The word ‘avatar’ derives from Hindu mythology. A God called Vishnu is believed to have visited earth nine times to curb evil. For each visit he took a different incarnation, called an avatar. Researchers and developers of virtual reality systems have produced a rich variety of definitions. A synopsis of all these definitions leads to a characterisation of an avatar as the embodification or representation of a user’s awareness and identity within a multi-user computer environment.

Avatars have several potential properties: identity, presence, subordination, authority, social facilitation. Avatars provide a way for other users to better understand the intended persona of the underlying user, his identity. They help establishing a feeling of "being there", a form of self-location, presence. They imply subordination, i.e. being under the direct control of the user, without significant control over their own actions and internal state. Avatars act with the authority of the user. They can provide social facilitation by giving a proxy for human communication and by facilitating interaction.

Software agents are meant to carry out tasks for the user and serve as another layer of mediation within the system [Maes 1995]. Reviewing the state of definition, it is proposed [Franklin, Graesser 1996], that software agents can have several properties: reactive, autonomous, goal-oriented, temporally continuous, communicative, learning, flexible. Thus agents not only simply act in response to the environment, they can be pro-active, purposeful, can exercises control over their own actions to a certain predefined degree. Agents can have a believable personality and emotional state. They can communicate with other agents and avatars. By applying artificial intelligence technology, such as expert systems, neural networks, and genetic algorithms, avatars can be adaptive and learning. Instead of scripted actions their behaviour can be based on their previous experience. Agents may be crucial for interacting and learning successfully in CVEs. In connection with avatars and CVEs, agents may play an important role concerning the psychological phenomena of immersion and presence as discussed.

"Information overload is the problem that agents alone can solve. Users are increasingly dealing with vast amounts of information that is unstructured and very dynamic. In order to keep track of everything, and in order to find the information relevant to them, they will have to use software that knows their interests and can act on their behalf." [Maes 1997]

In a CVE for tele-education the need for both synchronous and time-independent, asynchronous forms of communication and collaboration will arise and continuous intelligent presence of all participants involved in the teaching-learning process would be beneficial. A hybrid avatar/agent model is seen as a potential means of achieving this form of permanent presence in CVEs [Gerhard 1997]. In the absence of the underlying user, the avatar can function as an intelligent agent, perceiving and responding within the environment, giving, receiving and filtering information in order to fulfil a predefined task.

The autonomous character of agents and the subordinated character of avatars might seem to be contradictory. However, considering that a hybrid avatar/agent model can represent the user in presence as well as absence it is not really paradoxical; with the user being present the avatar operates under direct control, but, with the user being absent, the avatar, extended with agent technology, can exercise control over its own interactions with other users and the environment. Intelligent agent technology allows the user to customise the avatar’s behaviour for the times of absence and therefore influences the avatar’s operations towards a predefined learning goal.

With respect to tele-education the emergence of the World Wide Web (WWW) is certainly a key factor; accessibility and performance is increasing steadily and makes it a very promising platform for delivering distance learning via CVE technology. The WWW has had an enormous impact on the way in which information can be shared. Therefore our research is focused on Web-based CVE technology.

Recent years have seen the emergence of the Virtual Reality Modelling Language (VRML). VRML can trace its origins to a fusion of science fiction, virtual reality, 3D graphics, and the phenomenon of the WWW. The goal of VRML is to provide an open, extensible system that supports 3D interactive multi-user virtual worlds on the WWW. As the language evolved, initial 3D scene description capabilities have been enhanced with behaviour and logic, incorporated into the world via the use of script nodes that contain a reference to a piece of code written in a programming language such as Java. Animation and interaction features were added to scenes using sensors which trigger events. The official launch of VRML 1.0 was in April 1995, the specification of VRML 2.0 was completed in August 1996. The process of ratification was completed in December 1997, and the current version is known as ISO/IEC 14772, or informally as VRML 97 [htp://www.vrml.org].

In the meantime a myriad of browsers and various authoring tools and multi-user servers were developed. An estimated 30 million VRML browsers are installed world wide. Shared 3D virtual environment technologies on the Internet, such as Active Worlds [http://www.activeworlds.com], Blaxxun [http://www.blaxxun.com], Sony’s Community Places [http://www.spiw.com/vs], and The Palace [http://www.thepalace.com] are currently investigated within the AIMS group at LMU. The Web’s current bandwidth limitations however rigorously constrain the design of usable VRML avatars for these worlds, so for acceptable performance of web-based CVEs avatars should not have more than 500 polygons and no more than one texture [http://www.blaxxun.com/].

VRML is not the only way to create 3D virtual worlds for the Web. Some of the recent competing techniques are Microsoft's Chrome, SUN’s Java3D and MetaStreams by MetaTools. The future of VRML in particular and Web-based CVE technology in general is still open. VRML is the most promising technology, but has not found the support and acceptance to make it highly profitable; the expectations of the VRML community have not been met. "[…] none of the big players - not Netscape or Microsoft or IBM or SUN or even SGI - are standing in VRML-land anymore" [Pesce 1998]. However, if need arises, and a new technology should emerge, switching from VRML based CVEs to any other form of implementation is simply a matter of using different tools. The existing work, geometrical and behavioural objects will not need modification.

Work is under way at LMU to evaluate the educational utility of continuous presence in educational CVEs. A series of experiments using different web-based prototype CVEs and different forms of user embodiments will be conducted to find empirical evidence for the benefit of such presence.

The virtual body is widely seen as the key factor for presence in CVEs, however, measuring presence is not a trivial task. Asking questions that measure only the subjects’ perception of the technology that contributes to immersion can be easily confused with actually measuring a subjects’ feeling of ‘being there’ or their behavioural responses to events in the VE. It has been argued that measuring presence makes sense only when speaking about the degree of presence in one environment relative to another [Slater, Usoh, Steed 1994].

Despite the difficulties concerning the measurement of cognitive variables such as presence, the role of virtual body representation will be further analysed, using avatars and agents and focusing on co-presence and presence-in-absence. The outcome of these experiments is expected to encourage the use and guide the design of persistent avatars in educational CVEs.

[Bricken, Byrne 1993] Bricken, M., Byrne, C., Summer Students in VR: A Pilot Study on Educational Applications in

[Hawkridge 1995] Hawkridge, David, The Big Bang Theory in Distance Education, in Open and Distance

[Franklin, Graesser 1996] Franklin, S., Graesser, A., A Taxonomy for Autonomous Agents, in Proceedings of the 3rd

[Gerhard 1997] Gerhard, Michael, A Hybrid Avatar/Agent Model for the Representation of Users within CVEs, in

[Huxor 1998] Huxor, Avon, The Role of 3D Shared Worlds in Support of Chance Encounters in CSCW, in

[Laurillard 1993] Laurillard, Diana, Rethinking University Teaching: a framework for the effective use of

[Laurillard 1995] Laurillard, Diana, Multimedia and the changing experience of the learner, in British Journal of

[Loomis 1992] Loomis, J.M., Distal Attribution and Presence, in Presence: Teleoperators and Virtual

[Maes 1995] Maes, Pattie, Artificial Life Meets Entertainment: Life like Autonomous Agents, in

[Maes 1997] Maes, Pattie, in "The Future of Software Agents", in Roundtable Iconline,

[Mann, Mon-Williams 1996] Mann, John, Mon-Williams, Mark, What does Virtual Reality Need?: human factors issues in the

[Moore 1993] Moore, Michael G., Three Types of Interaction, in Distance Education: New Perspectives, edited

[Morphett 1996] Morphett, Jason, Presence in Absence: Asynchronous Communication in Virtual Environments, in

[O’Malley 1995] O’Malley, C., Computer Supported Collaborative Learning, Springer, Berlin, Germany, 1995

[Pesce 1998] Pesce, Mark, What's in Store for VRML?, in InternetUser, Ziff Davies Net, Vol May, 1998

[Psotka 1995] Psotka, Joseph, Immersive Tutoring Systems: Virtual Reality and Education and Training, in

[Slater, Usoh 1994] Slater, M., Usoh, Body Centred Interaction in Immersive Virtual Environments, in Artificial Life

[Slater, Usoh, Steed 1994] Slater, M., Usoh, M., Steed, A., Depth of Presence in Immersive Virtual Environments, in

[Tromp 1995] Tromp, Jolanda G., The Cognitive Factors of Embodiment and Interaction in Virtual

[Wexelblatt 1993] Wexelblatt, Alan, The Reality of Co-operation: Virtual Reality and CSCW, in Virtual Reality: