Virtual reality systems are getting more and more attention. Even people who are not interested in computing have been persuaded that this technology can bring us great benefits. The possibility that we can see photo-realistic representations of complex structures, move around them, perhaps manipulate them - all without leaving the comfort of home - is truly seductive. My own experience with a rather simple VR system (see Newsletter, Vol. 5, no. 2, Aug., 1992, "Computer Graphics Developments," pp. 6-7.) was instructive. I found that comparatively unsophisticated system to be quite compelling, but the increasing power of our computers will make far more realism possible. With increasing realism will come ever more compelling virtual worlds.
Douglas Gann's illustration of Homol'ovi in the last Newsletter (Fig. 6) was one example of the results that can be obtained. The best renderings I have seen were published in Progressive Architecture for September, 1993 (pp. 81 ff.). The images are of Louis Kahn's never-built Hurva Synagogue in Jerusalem, and they were made by the author of the article on the synagogue, Kent Larson. Mr. Larson also constructed the CAD model of the synagogue that underlies the renderings.
This is exciting, but some trends are troubling. Not all who have used the technology have started from the kinds of carefully constructed models used by Mr. Gann and Mr. Larson. As more and more people have realized the potential of the technology, some have allowed the power of the presentation to overcome concerns about the accuracy of the information being presented. When we are in awe of the technological legerdemain, we may forget to question the scholarship that underlies the model. As noted by Ms. Lukesh (see Imaging the Past), this problem was "heatedly discussed" at the London Imaging the Past conference.
VR worlds rely upon models - models built with CAD systems or similar modeling tools. The models can be very simple, or they can be amazingly complex. In either case, however, the model need not be accurate to yield a visually compelling rendering or VR presentation. That is, a model of a building such as the Pantheon could be made with measurements to the nearest meter or the nearest millimeter. When rendered or made into a 'virtual Pantheon,' each model would yield a different visual impression, of course, but only someone familiar with the building could tell which model more accurately reflected the building itself. At small scales, in fact, even an expert would have some trouble identifying the more accurate model.
Does that mean a model made from measurements to the nearest meter is as good as one made from measurements to the nearest millimeter? Or is there a difference? Obviously, the more accurate model is better. Dimensions from the model can be used. (All dimensions from a CAD model can be retrieved.) Subtleties can be examined. If the representations are large, subtleties can even be seen as they would in a good photograph.
But what about the less accurate model? In some cases, as when a building being modeled is no longer available for examination, a model of uncertain accuracy may be the only one possible. In that case, should it be made at all? Is it illuminating or misleading to make a visually compelling model that is based on poor information?
More to the point today, there are projects involving large and complex models being proposed - a 'virtual' ancient Rome and a 'virtual' Pompeii, for instance. How will the models be made? What levels of accuracy will prevail? What kinds of resources should be spent to create models using gross measurements - but models that will be visually stunning? This is an especially thorny question in our current situation, with diminished resources for work of any kind.
I have asked a good many questions but offered few answers. Now let me suggest some responses that may lead toward answers. First, a good model may serve both scholarly/educational and entertainment purposes, but a poor model can only entertain. Insofar as it teaches, it is likely to mislead and, therefore, to teach badly.
Second, models created as part of a scholarly project must meet standard demands of scholarship for accuracy and thoroughness. Models of archaeological material, for instance, should be based upon the kind of accurate measurements one would use for any other archaeological work.
Third, a poor model may do no harm so long as it used only at small scale and with no access to the underlying geometry. Such a model would be little different from a good sketch or hand rendering of a standing monument.
Fourth, scholarly models will be more expensive and time-consuming to create. If a model is to be used for a transient purpose, the cost of the scholarly model may not be justified.
Fifth, it makes no sense to make a good and a poor model of the same thing. But people must be prepared to share models if that duplication is to be avoided.
Sixth, no model that misleads should be tolerated. Some very poor models may not mislead if they are used on systems that restrict the scale of reproduction; better models on more flexible systems may be very misleading.
I have suggested a few ideas. I surely do not have all the right answers, but I do think I have the right questions. I hope those who are thinking about VR work or photo-realistic rendering will ask themselves these questions before they begins.
For other Newsletter articles concerning the use of electronic media in the humanities, consult the Subject index.
Next Article: Finding Numerical Relationships in Architecture
Table of Contents for the February, 1995 issue of the CSA Newsletter (Vol. 7, no. 4)
Table of Contents for all CSA Newsletter issues on the Web
Go to CSA Home Page