We are taking as our primary application domain that of agoraphobia. This has a number of advantages. At the technical level we are constructing virtual cities, cities that aim to be alive and meaningful for participants. In order to understand the effectiveness of such virtual cities, we need a measure of the degree of presence that people experience while in them. Presence in itself is hard to measure. However, for people who suffer from agoraphobia, induced anxiety will be a sign of presence (without presence there would be no anxiety). Anxiety is relatively straightforward to measure.
The existence of the anxiety is perhaps the first clue that VR actually has any real effect at all. Once we have a measurable effect, we can look for variables that determine the strength of that effect. Once we can modulate that effect, we can exploit it. This is an effort to show that although we know anecdotally or emotionally that VE reactions can cover the range of human emotion, we simply don't know why we get those effects or if they were at all appropriate.
We have utilised anxiety measure before in our studies on fear of public spearking and paranoia. Agoraphobia is perhaps an even more interesting topic because, in representing a space, it is even more likely that we will miss subtle cues that mean that we cannot generate anxiety in certain classes of anxiety sufferers. Fortunately a virtual environment is highly malleable, so we can strive to cover all potential anxiety-provoking cues, and even generate certain cues at various "levels" where a different levels might generate different levels of anxiety. For this reason, we are building parameterised models of cities where we can explore the mapping between certain characteristics of the environment and the response.
Why we claim that a virtual environment is a "mixed-reality" is that it only successful, if some aspect of it, such as memory or training, remains with the participant so they can utilise it in the real world. Examples, include calming techniques and exposure therapy. Thus we are trying to relate the virtual and real ultimately for the purpose of treatment. What the participant learns in the VE can (in some sense) be taken out to the real, and used as a facilitation device.
We can emphasise the mixing of the real and virtual in many ways. We can consider not just making a direct visual relationship between the real and virtual, but having physical or virtual tokens that can be carried into and used in both environment. For example, something that "naturally belongs" in both, and which therefore induces a (positive) con-fusion in the mind of the participant as between what is 'virtual' and what is 'real'? They emerge (and merge!) into a single entity, a shared reality, where learning in one 'place' is now also learning in the other. This could be as simple as an audio recorder, for taking notes or for recording relaxing sounds. Or it could be a PDA with an embodiment of the therapist as a talking head. In this example it is important that the PDA work exactly the same so as to emphasise the similarity of the real and virtual. To make the relationship more symetrical we could consider tokens that carry aspects of the real into the virtual as well as vice versa. So a PDA or other mechanism could represent or bring representations of the real world into the virtual. Thus the relationship between the two is made more symmetric. This symmetry could be useful in persuading the participant that if they can cope with one (the virtual) they can cope with the other (the real).
From this we can consider a different type of mixed reality that exploits conditioning. Here we do not mean 'conditioning' in the pure Pavlovian sense, but rather a facility to form associations between similar situations in both experienced environments (real and virtual). Although we have mentioned the transport of media to and from the virtual environment, we can utilise this further by making an association between, say an audio stimulus, in the virtual environment and then exploit that association in equivalent contexts in the real. Once we have built a model of how the k parameters that define the virtual environment generate anxiety, we have a good starting point for investigating minimal cues for anxiety. From that model we also have a good device for measuring progress in treatment or relapse.
Once we have reached this stage we believe we will have given a proof the use of a representation of a space to affect the experience of actual spaces. We can then think about how we might combine this with other threads of work on representations of people and things.
In UCL's agoraphobia work there are many areas where we would need the help of other partners. In particular, real-time on-line measurement of anxiety (or other responses that are associated with clear physiological states) might be thought of as a problem in the realm of wearables. Such systems have recently been proposed as communication devices in City project (see Cliff's UberJournal piece). The design of the devices that might be shared between the different environments would be a matter of paramount importance - we want participants to concentrate on the environment and their responses to it, rather than on trying to work an interface.