From: uselton@nas.nasa.gov (Samuel P. Uselton) Subject: Re: PHIL: virtual-worlds ethics Date: 27 Nov 1995 18:45:28 GMT Message-ID: Organization: Numerical Aerodynamic Simulation, NASA Ames From: uselton@nas.nasa.gov (Samuel P. Uselton) In article <484p95$hp6@cville-srv.wam.umd.edu>, weyker@wam.umd.edu (Shayne Weyker) writes: <...deletion> > How do the exclusion of certain aspects of real-life in a simulation > of some part of real-life (for technical, instructional, or > entertainment reasons) shape, and perhaps distort, people's > understanding of the real-world events being represented to them > through VR? <...deletion> In article brownw@atc.boeing.com (William J Brown) responds: These are really interesting questions. I'm sure Fred Brooks and James Coggins at UNC would have a lot more to say about this than me but.... First of all, this is very similar to the debate already going on in the field of Medical Imaging. "How do I know I can trust a digital image as opposed to an X-ray?", "How do I know what processing has been done and what the effects are?" The medical community (at least in the US) has a very stringent testing procedure that must be followed before new technology (of whatever kind) is made available for standard clinical use on humans. These tests include demonstrations that doctors using the new technology are effective at the tasks for which the technology is being tested. I have been concerned with how to make scientific visualization "more scientific" for a couple of years now. My main direction at the moment is to promote tests to determine whether visualization is actually making a difference in particular applications, and to devise mathmatical models to predict whether particular visualizations or visualization features will make a difference for particular uses. Ethically, it seems to me we have a responsibility to represent what we can represent as accurately as possible and, of course, not try to deliberately mislead. I think we also have a responsibility to know what the most important sensory (not just visual) cues are and work down a prioritized list rather than just focus on what's easiest. This last part is, in general, not being done. To avoid deliberately misleading displays is obviously desirable. To avoid displays that we DON'T REALIZE ARE WRONG OR MISLEADING is much harder and of more concern to me. Something Fred Brooks often emphasizes is to focus work on particular users and the tasks they need to accomplish. Making things look "real" or displaying all data available are the wrong goals. Display the RELEVANT data in the way MOST EFFECTIVE for the user to accomplish the task. Sometimes that means diagrammatic line drawings are "better" than images indistinguishable from those captured by a camera. Sometimes it means that a list of a dozen numbers is better than a picture. And sometimes it means we want as much visual accuracy as we can generate. But we need to pay more attention to choosing appropriately. Making the processing history available is also a good "scientific" technique. It corresponds to the "methods" section of lab reports and publications from experiment based science. Depending on the context, this may include information about what numerical technique is used in some particular tool that came from an independent source, as well as all the processing that is done with code generated locally. The "most important sensory cues" is an interesting idea but not well defined. Again I think it is essential to narrow the focus. What specific task is being attempted? What are the relevant characteristics of the users that will use this system? How is success defined? What is most important in a system designed for presenting information to a general audience assumed to have a high school education is very different than what is most important in a system for PhD physicists exploring results of simulations of aeronautics experiments of their own design. (IMHO of course.) Samuel P. Uselton uselton@nas.nasa.gov Employed by MRJ working for NASA (AMes) speaking for myself