Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!swrinde!ucsd!ogicse!milton!hlab
From: kirlik@chmsr.gatech.edu (Alex Kirlik)
Newsgroups: sci.virtual-worlds
Subject: Re: Virtual Reality, Helmets, and Gloves
Message-ID: <1991Apr4.021328.12844@milton.u.washington.edu>
Date: 4 Apr 91 01:05:47 GMT
References: <1991Apr3.231354.27540@milton.u.washington.edu>
Sender: hlab@milton.u.washington.edu (Human Int. Technology Lab)
Organization: ISyE, Georgia Tech
Lines: 89
Approved: cyberoid@milton.u.washington.edu



In article <1991Apr3.231354.27540@milton.u.washington.edu> esz001@cck.
coventry.ac.uk (Will Overington) writes:
>
>My reason for asking is that it would seem possible to be able to
>get a lot of work done without helmets and just using a PC as
>the picture producing device. Maybe one could still use a glove,
>or maybe this could be an option.
>
>Would work of such a nature be a reasonable facet of discussion
>in sci.virtual-worlds or would this be debasing the reality aspect
>of virtual reality?

My opinion is that a discussion aimed at defining and bounding the set
of human-machine interaction problems for which VR is an appropriate
solution is surely a welcome topic in this newsgroup (and in the VR
community for that matter).  I have seen relatively little discussion
of this matter here or in the open literature.  Without a good
understanding of the types of problems for which VR will be an effective
solution, I am concerned we will see VR being force fit as a solution
to non-problems, or as a solution to what may turn out to be very
small problems in certain cases of human-machine interaction, leaving
the real problems unaddressed.  Or, in analogy to other primarily
technology-driven approaches to interface design (e.g., expert systems)
it may even be the case that VR technology may have the potential to
introduce its own novel class of interaction problems.

Now if I was a VR technologist I wouldn't spend a lot of time worrying
about this issue; I couldn't see why it would pay me to do so (in the
short term at least).  Too many "real" engineering problems to deal
with now.  If, on the other hand, I was primarily interested
in designing a cockpit or control room and wanted to add VR to my tool
box, defining its scope and limits would be one of the first things I'd
want to do.

I'll take a stab at it in the hopes of hearing what others think.  It
strikes me that VR will be appropriate for solving two sorts of 
interaction problems.  The first set of problems are those where
restricted perception-action access to the environment is the major
factor constraining human-machine system performance.  Say we observe
an aspect of human behavior that we'd like to aid, eliminate, etc.
through better design.  I'm suggesting we ask ourself whether the
observed behavior is due to restricted perception-action interaction
with the environment (as that interaction is mediated through the
existing interface), or whether that aspect of behavior is in some
way inherent in the system control task.  I suspect that many of the
"situational awareness" problems and memory problems are indicative
of overly restrictive perception-action access.  Here we may want to
use VR to provide the operator with a less opaque window to the task
at hand.  Many interaction problems, though, do not fit this schema.
A decision support system that allows the decision-maker to "walk"
though a space of possibilities and to create possible worlds may
sound neat but it will (or perhaps should) never reduce the sometimes
painful need to clearly examine value structures, etc.

Secondly, I suggest VR will be appropriate for that class of problems
where we can invent a novel perception-action solution to a problem
that had previously never been performable in a perception-action
format.  Unlike the problems mentioned above, here the issue is not
so much the "reality problem" of portraying an enhanced veridical
system representation, but rather the "mapping problem" whereby we
try to ensure that performance on the substituted problem is better
than performance on the original problem.  As a gross simplification,
assume that people have two "modes" of selecting behavior, one a fast
and possibly parallel mode based on tight perception-action relation
to the external world, and the other a slow possibly serial mode based
on using representations to detach oneself from the here-now to expand
the range of information considered.  The issue is to map the first 
type into the second by providing concrete representations where
abstract ones were once required.  Not only do you have to be pretty
clever to do this, you also have to make sure that the inherited
error forms characteristic of perception-action processing do not 
lead to bigger problems than you started with.  I usually don't mind
the "slips" characteristic of skilled driving behavior (taking the
path of least resistance and forgetting to stop at the bank on the
way home) partly because my world is relatively friendly to them.
But now take the same kind of behavior as an engineered analog for
controlling a nuclear power plant and those kinds of slips can be
disasterous.  Maybe in certain cases the fluency of skill-based
solutions to cognitive tasks can come at too high a price.

That's my two cents; I'm interested to hear what others may think.

Alex 

UUCP:   kirlik@chmsr.UUCP
        {backbones}!gatech!chmsr!kirlik
INTERNET:       kirlik@chmsr.gatech.edu