Use a KIF like language (SUO-IEEE) to represent shared active episodic memories.  Java, Prolog and Visual Basic libraries integrate with a first person VRML/MOO system to provide debugging and agent facilities.

Libraries are provided that you need to use SUO-KIF (Standard Upperlevel Ontology - Knowledge Interchange Format) and DAML (DARPA Agent Markup Language) content into an episodic simulation simulation of a virtual world first person VRML/MOO system.  The programming languages used where Prolog, Java, and Visual Basic.  Combining powerful tools: Microsoft Virtual Worlds and two Edinburgh style of Prologs SWI-Prolog/B-Prolog we can represent episodic models of memory, diagrams for belief engines, or fun 3D role playing games.  SWI-Prolog is a stable, free, professional distribution of the prolog virtual machine and compiler.  Microsoft Virtual Worlds provides a complementary graphical environment: It supports rapid prototyping of distributed multi-user applications of 3D virtual communities with streaming media and chat.  But most importantly it supports a persistent, changeable world state that preserves object level information. B-Prolog is a recent addition thanks to work being done by Feng-Fa Zhou.

The Virtual Worlds Group http://vworlds.research.microsoft.com is an organization of individuals who are dedicated to helping advance the body of research on online multi-user applications. The founding members of this group are employees of Microsoft Research and have created a software platform that enables users to build these applications and quickly prototype user interfaces. They have released the source code of the Virtual Worlds software in conjunction with the binary builds in an effort to actively foster research. The platform is very general purpose and, as such, can be used to build almost any type of multi-user application. You don't have to know C++; you should have a basic knowledge of scripting. Interface and object behavior is programmable in languages like VBScript, Jscript, DHTML and now available through the Prolog compiler.

PrologVirtualWorlds Architecture Background
The Microsoft Virtual Worlds Service is an object remote brokering system: It is based on many clients and a server sharing references to instances. (COM/ActiveX Objects)  PrologVirtualWorlds takes advantage of this by connecting to virtual worlds as an (in)visible Avatar to programmatically alter objects and invoke methods.  It can create or remove properties, design new methods, and manufacture new object types.  Other agents may assist in work and run from different machines or languages.  An agent itself is an object and can alter it's own interface members, or of others.  The server maintains the consistency of the objects and facilitates news that may have affected an agent. This allows us power beyond conventional object orientated programming.

A simple example of agent orientated programming inside a Moo:

• A Rain Agent that enters a room object and emotes that it is raining.  It then begins adding a 'wetness counter' property to each object in the room.  The longer it stays in the room the more it increments. (This does not have any functional impact on the current workings of the objects in the room)

• Later you may want to write another agent who enters the room called 'Sun Agent' who's duty is to look for objects that have the property 'wetness' to decrement.  

• A third agent called 'Common Cold' can go thru and look for both a 'health property' and a 'wetness property' and then decide to alter the 'smile method' from emoting 'beaming smile' to 'tries to smile from behind a stuffed up nose'.  

Why is this important?

As since all of these actions are moving thru representation terms  (it can be graphically depicted) another agent may in tandem try to analyze what had just happened and build a 'story script' object. The properties of such a script are in an action-level-ontology language the represents an instantiated scene ('story script' objects have a method called replay).  Since information about the methods and properties were public at the time the system may at later replay the scenario placing different objects in place of the originals in a non-deterministic way. 

Sometimes we do not have all of the information to heuristically search for a given situation.  As we discover new facts, the situation we learn was not correctly represented. Since the situation is persistable we can re-instantiate the episode and put better-educated agents to the task of understanding.  This one day I hope will further our technology and searchablity of 'scenario' and will allow better rule styles to be designed.

Brokering Model (sites are not places on the internet they are memory spaces that the object resides in)