From: Stephen Smoliar <ISSSSM%NUSVM@UWAVM.U.WASHINGTON.EDU>
Subject: Re: military information systems (LONG)
Date: Tue, 02 Jul 91 09:02:24 SST
Organization: Institute for Systems Science, Natl. University of Singapore



In article <1991Jul1.215553.17343@milton.u.washington.edu>
gbnewby@rodan.acs.syr.edu (Gregory B. Newby) writes:

>What makes a mil. info. sys. different from your standard virtual
>reality environment is that all data are not physical.  That is, you
>wouldn't just have, say, a map of a theatre of action and locations of
>tanks and oil wells and such....  You would also need to represent
>non-physical data.
>
>Non-physical data?  I mean things such as text (descriptions,
>transactions, communiques), statistical data (weapon status,
>temperature, casualties), and other things which don't have a direct
>physical manifestation.

I am just beginning to get involved with virtual reality, so I apologize if my
remarks may sound overly naive.  However, in thinking about virtual realities,
I often find it helpful to refer back to certain conventions present in "real"
realities as guidelines.  In the "real world," so to speak," we deal with
"non-physical" data by MAKING IT PHYSICAL.  If a military communique has
any import at all, then it is generally rendered in some hard-copy form,
at least long enough for the recipient(s) to absorb it (after which it may
be destroyed for security reasons).  What I am getting at is that if an officer
expects to receive information on a piece of paper in the REAL world, then he
is likely to have the same expectation if you put him into an analogous VIRTUAL
world.  This need not imply that you have to simulate all the physical
characteristics of a piece of paper which has just been torn off of a
FAX machine;  but it probably DOES imply that one of the kinds of objects
which is part of your virtual reality must have some of the characteristics
of paper:  easy to hold, easy to read, can be written on, can be passed around,
can be folded, can be destroyed.

Therefore, for any manifestation of non-physical data, I think it makes sense
to ask:  "How would we encounter it in the 'real world?'"  Nevertheless, this
should probably be only a point of departure, rather than a final answer.  For
example, casualty figures may be just that--a sheet of paper with numbers on
them.  Often it is then necessary to do some demographic analysis.  Where were
the casualties sustained, for example?  Now you are faced with the problem of
integrating data you first received on your sheet of paper with your "physical"
map.  If the map were "real," you might make marks on it to indicate which
losses were incurred where;  again, you would like to expect to do the same
in a virtual reality.  The general principle, then, is to start with how we
would expect to encounter a physical manifestation of the data and then proceed
to integrate those expectations into the capabilities of the virtual
environment.

>To represent non-physical data means we have choices:  how to visualize,
>where to locate, how to decide which information is relevant....  The
>literature on information retrieval doesn't speak to these issues (they
>are almost entirely text-based:  something we will probably want to
>minimize in a virtual information system).

I'm not sure it would be fair to say that we want to MINIMIZE text.  We just
want to use it where it is appropriate, and satisfying the expectations of the
user is one criterion for being appropriate.  For better or worse, the lion's
share of our communication is achieved through text.  (If God had wanted us to
communicate without text, He would have evolved our brains differently!)  If we
become too engaged in trying to translated everything into buttons, knobs, and
levers, we may run the danger of having a different widget for every sentence
we might utter . . . meaning that our repertoire of widgets would be, for all
intents and purposes, impossible to manage.

The real problem, then, lies in the matter of integrating text and non-text in
a common user environment.  My own research is currently concerned with
achieving such integration strictly for information retrieval, without
worrying whether or not the environment is manifested as a "virtual world."
However, I think there is probably some overlap between the issues I have to
address and those being raised by Greg.

>In the ideal case, all information in a virtual information system
>("VIS," to invent a new term) would be located just where someone
>would look for it, and be presented in just the right way so as to
>provide all the information needed, and none of the information not
>needed.  Right....  Well, that hasn't happened in years and years of
>research on AI and IR (information retrieval).

One of the reasons this goal has not been achieved may be that this "ideal
case" simply does not exist.  Let us first consider the phrase "located just
where someone would look for it."  The first item to pick on is the question
of just who this idealized "someone" is.  There is no reason to believe that
any two people would expect to find a given piece of information "in the same
place" and lots of reasons to believe that they actually do not.  (Start by
trying to find your way around in a friend's personal library.)  However, even
if we decide to abstract the universe down to only a single "someone," I would
argue that we are still not out of the woods.  This is because "looking for
information" is no guided by some abstract description of the information we
seek;  rather, it is guided by the goals we are trying to achieve.  For
example, if I am trying to compose a paragraph about problems in implementing
information retrieval systems, I expect to look (in either my personal library
or in an "official" library) around a certain shelf area for books on
information retrieval.  However, when I go off on the tangent of this
particular paragraph, that area is probably not going to do me very much
good, since I am moving into a domain more appropriate to cognitive psychology
and philosophy.  The point is that, as I pursue my task, I am not always
particularly certain where I want to look.  For that matter, I may not always
have a clear idea what I am looking for!  As human beings we are highly
SUBJECTIVE in our approach to information retrieval.  The technologies
of artificial intelligence and information retrieval have yet to provide
an adequate handle on such subjectivity;  and I think this is the key reason
why their results to date have been relatively disappointing (at least for
this particular problem domain).

>So, what can we do?  For a small- to medium-sized domain, we can model
>the organization and representation of the data by just asking people
>to talk about the processes they go through and the situations in
>which they need information:  what questions and goals do they have at
>each phase?  Then, we have a pretty good basis for organizing and
>presenting the data.  This would probably work pretty well for
>decision making tasks.
>
>For larger domains (like you find in an online catalog of bibliographic
>references, for example), both the problem space (of the user(s)) and
>the information space (system domain) becomes too large to map by
>asking people.  We need to think here about automatic means of
>representing the data.

The problem with this approach is that it seems to indicate that we should be
able to "converge" on some "ultimate" SINGLE solution.  However, if we admit
subjectivity as part of the picture, there may be no "fixed point of
convergence."  The "point" moves around, not just from person to person
but also from situation to situation.

Here at ISS Kim Fairchild has been raising an intriguing alternative.  I shall
try to confine myself to a brief summary and let Kim elaborate further if he
wishes.  Kim has been very interested in representing complex information
structures as virtual environments which may then be "navigated."  However,
his vision of the navigator is less that of a passive observer and more that
of an agent INTERACTING with the environment.  What makes Kim's vision
particularly exciting is the idea that the tools which can be used to
manipulate this environment have their origins in the environment itself!
Thus, the environment furnishes the user with either the tools of manipulation
he needs or resources to build other tools which are more appropriate.  In a
rather audacious way, the virtual world becomes what you choose to make it.

The reason I think this is important is because it strikes me as one of the
most viable ways to confront the subjectivity problem.  Rather than an
environment which reflects some objective "solution" (which probably does
not exist), we have one which is responsive to our subjective attitudes and
behaviors.  Through flexibility, it can reflect the changing needs of changing
situations.  The role it plays is closer to that of the cooperative librarian
than that of the library, itself.

Unfortunately, it is not going to be sufficient to solve this problem for a
single user.  Not all military decisions are made by a single individual in
isolation.  A War Room, for example, is inhabited by a small population who
spends as much time communicating among themselves as they do interactive with
the available information.  Therefore, it may be not realistic to imagine
virtual environments which are only inhabited by a single individual.  There
are probably situations in which several individuals may have to enter the same
"world" together, interacting among themselves as well as with that world.

I must confess that I am still not sure how to deal with this problem.  My
approach to subjectivity is tempered by a fundamentally solipsistic stance.
Philosophically, this means that every user has is own virtual environment;
and the presence of other users can only be detected THROUGH that environment.
To fall back on science fiction for a metaphor, this would mean that you could
not build a holodeck (a la STAR TREK) to accommodate multiple inhabitants due
to the subjective vagaries of each individual;  instead, each would have to
have his OWN holodeck within which he would perceive that there are others
present.

From a point of view of system architecture, this may not be as intimidating as
it seems.  It DOES mean that every user requires his own "work-station" for
entering the environment;  but this does not seem to be an unreasonable demand,
at least for the eventual future.  Multiple users would then "share" an
environment through some common "server."  This server would be responsible
for making sure that any individual is aware of others inhabiting the
environment.  Of course we are already familiar with problems of conflict
management which arise when resources are shared in time-sharing and data
base systems;  and I would not want to underestimate the magnitude of such
problems when we step up to the arena of shared virtual environments.  I guess
the personal question I must face is whether these potential problems of system
design are strong enough to shake my current philosophical advocacy of
solipsism.  For now, the answer is "Not yet;"  but I would certainly be
interested in any subsequent discussion on this matter.

===============================================================================

Stephen W. Smoliar
Institute of Systems Science
National University of Singapore
Heng Mui Keng Terrace, Kent Ridge
SINGAPORE 0511

BITNET:  ISSSSM@NUSVM

"He was of Lord Essex's opinion, 'rather to go an hundred miles to speak with
one wise man, than five miles to see a fair town.'"--Boswell on Johnson

--