Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!gatech!news-feed-1.peachnet.edu!insosf1.infonet.net!newshost.marcam.com!uunet!in1.uu.net!psinntp!relay.lehman.com!florida!fsimkin
From: fsimkin@shearson.com (Fred Simkin)
Subject: Representation of Knowledge in War Games and Conflict Simulations
Message-ID: <D5LKLB.I9A@lehman.com>
Keywords: 
 
Sender: news@lehman.com (News)
Nntp-Posting-Host: florida
Reply-To: fsimkin@shearson.com
Organization: Lehman Brothers, Inc.
Date: Fri, 17 Mar 1995 18:27:10 GMT
Lines: 30


Having devoutly wished, hoped and prayed for the forum I now take fingers to keys and post this my first message. I have two hopes one to turn the conversation away from the "gosh did you see the neat beast in Nazi Blowfish from Heck" type of message which  have appeared here along with the " does anybody know algorithm for...", towards a 
more general discussion of the nature of and proper application of "Intelligence" in games.

In fairness to readers (and in order to let them set kill files accordingly) I believe it behooves me to start by stating some personal biases:
I am a boomer (born 1950) ZIMA drinkers beware. :-)
I make my living developing commercial applications of that subset of AI referred 
to as Knowledge Based Systems. 
I am what is sometimes derisively referred to by those in academia and research as a "tool-bender". That is I make use of  high level development tools rather than languages in most (but not all) of the work I do.
The area of gaming that I am most interested in is wargamming/Conflict Simulation.
I do not think this is the right forum for a) "Is this where I find out about AI". or b)"Can somebody tell me where I can find game X".

Having  stated all that (and donned flame retardent clothing)let's get on with it.

The first thing I'd like to say is that I believe that important to separate "intelligence (knowledge)" from the construct commonly called AI in games. What game manufacturers call AI is in reality the artificial opponent.  This may seem to some 
to be a politically correct act of hair splitting but in fact I believe that until we separate these concepts we cannot look at the application of intelligence in games in 
a meaningful way. I believe  that the majority of existing "AO's" (at least in war games) are designed to act on data rather than to reason about data and in that intent lies the difference between truly intelligent systems and conventional systems.  


Next  I'd like to open the discussion on is the representation of knowledge in games more specifically war games and conflict simulations. Over the last several years I have become increasingly aware that even in those instances where a single source of domain expertise exists that no single knowledge representation schema can provide solutions across even a limited problem space. I have found during knowledge acquisition that the domain expert may approach seemingly similar problem models with different o
r mixed solution methodologies. This has led me to believe that any adequate intelligence (knowledge )representation in war games or conflict simulations must incorporate multiple representation schema. As I have examined the problem of incorporating intelligence into war games and conflict simulations I have found that the same is true of military expertise (see any book on Rommel in the Western Desert). Thus an intelligent component of either the AO or the player adviser might use neural nets to recogniz
e patterns of activity. When certain activity fall into a pattern the net hands off the output to a fuzzy rule base which determines the patterns degree of membership in an activities set finally the defuzzified results are sent to a crisp rule base or case base and course of action is recommended. I realize that the concept of a hybrid system is not new but I have as yet to see any evidence of its application in commercially available war games or conflict simulations.

What I'm asking the group for is their thoughts on this concept of a hybrid knowledge solution and perhaps some thought on how one would create a meta knowledge system to determine how the various sub systems would determine what type of problem was appropriate for their particular type of representation.

O.K.  you may fire when ready

Fred Simkin
AICando@aol.com

