From: kirlik@chmsr.gatech.edu (Alex Kirlik)
Subject: Re: Can displays be "too real"?
Date: 22 Oct 90 21:45:47 GMT
Organization: Center for Human-Machine Systems Research


In article <9684@milton.u.washington.edu> cyberoid@milton.u.washington.edu 
 (Robert Jacobson) writes:
>
>First, there is an extensive corpus of literature produced by
>Tom Furness at Wright-Patterson AFB, during the late 1970's and 1980's,
>that demonstrated that pilots DO perform better where there is a 
>feeling of presence. 

Thank you for the pointer, I will have to try to get access to this work.
Also, let me apologize for the 2 redundant postings, as I did not realize
this group was moderated and I though my first posting was lost when it did
not appear.

Also, I'd like to say that I am enthusiastic about the potential of VR and
perceptual enhancement in general to solve certain human-machine interaction
problems.  My own view is that much of the observed cognitively intensive
and error prone activity one sees in complex systems results from the 
operator's attempt to cope with perceptually impoverished (or at least
perceptually innapropriate) interface conditions.  A good amount of time
and effort has gone into exploring aiding concepts and training methods as
potential design solutions, but relatively little effort has been spent on
the alternative, namely creating interfaces that eliminate the need for
certain of these cognitively intensive error prone activities in the first
place.  I think VR, visualization, etc. has the potential to make a real
contribution in this area.

The reason for my original posting was, that when I am asked to defend the
design approach outlined above the objection is sometimes raised that we
might eliminate some of the cognitive-level errors and biases but introduce
a new set of errors and biases, related to both the cognitive and experiential
aspects of perceptual function.  I still think that the perceptual enhancement
approach is on the right track, although I take these objections seriously 
and would like to have more data on this issue.

One reason one must be careful when looking at flight simulation research
(and I don't mean to prejudge the research mentioned above, I am speaking
of other simulation research with which I am familiar), is that it is not
clear how to attribute the causes of enhanced performance to properties of
the simulator interface/environment.  Much of this research uses experienced
pilots as subjects, and I suspect that the reason that performance is better
when subjects report felt presence is that presence enhances the transfer of
previously learned skilled routines to the flight simulator environment.
In these experiments, then, it is not presence for presence sake that may be
enhancing performance, but rather it is the realism that allows for positive
transfer of learned skills.  An interesting issue is whether realism promotes
felt presence and this promotes transfer/performance, or whether both felt
presence and transfer/performance are two independent effects of realism,
although this is a quite different issue.  In those experiments using skilled
subjects in simulations of the environments in which their skills were 
acquired, then, it may be difficult to parcel out the effects of felt presence
on performance and the effects of positive transfer, in my opinion.

I do, though, intend to take a look at the simulation research you mentioned.
When designing new interfaces/systems where we are not relying on the transfer
of previously learned skills, it still may be the case that felt presence
enhances performance, but as I conjectured previously I am concerned that
presence may have the potential to have both beneficial and deleterious
effects. In certain systems, we may even want to discourage the transfer of
innefficient previously learned skilled routines.  In tasks such as locomotion
(driving, flying etc) we are implicitly relying on the transfer of very-
previously learned routines, those evolved to help us get around the world,
and here I am reasonably confident that felt presence will correspond with
good performance when designing interfaces.  For decision-making, problem-
solving, and the like, though one must be pretty confident in how the cognitive
task has been mapped to a perceptual-motor task, and what the new performance
limitations and biases introduced might be.

Thanks again for the pointers and the comments,

Alex 

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