This page is very much under construction. Patience.

Research and Development Projects
Of Interest To
Dr. Brian M. Slator
NDSU Computer Science Dept.
IACC Building, Rm. #258-A6
phone: (701) 231-6124
e-mail: slator@badlands.nodak.edu


* On-going Research and Development

Several projects are underway, in various phases of development. Interested parties and volunteers are welcome to visit and discuss these.

Projects are all in the general area of educational media. Keywords are artificial intelligence, multi-user environments, virtual environments, MUDs and MOOs, educational games, distance education, VRML, role-playing, software agents, software tutors, assessment, software tools.

For details, write slator@badlands.nodak.edu

PROJECTS
Geology Explorer Project
Project Collaborator(s)
Planet Oit Content Development Dr. Schwert, Dr. Saini-Eidukat, Dean Vestal
Planet Oit Assessment Studies Dr. Schwert, Dr. Saini-Eidukat, Ned Kruger, Elizabeth Smith
Planet Oit Graphical Java-MOO Client John Bauer
Planet Oit Graphics Rebecca Potter, Acey Olson
Planet Oit GUI Design Dr. Schwert, Dr. Saini-Eidukat, Dean Vestal, Rebecca Potter, Acey Olson, John Bauer, James Shimer
Planet Oit Software Tutors Krista Dauner
Planet Oit Water Processes Rahul Devabhaktuni
Planet Oit Environmental Effects James Shimer
Crystal Formation Agents in VRML <volunteer needed>
NLP-based Conversational Concept Agents <volunteer needed> (was Brad Vender)
Geology Explorer Workbooks <volunteer needed>
.
Retail Education Project
Project Collaborator(s)
Retailer Game Graphical Java-MOO Client Brad Vender
Retailer Game Graphics Josh Stompro, Matt Carlson, Brad Corradi
Retailer Game Agent-based Economic Simulation Golam Farooque
Retailer Game Economic Forcasting Hong Wu
Retailer Game Market Research Jun Zhou
Retailer Game Employee Agents <volunteer needed>
Retailer Game Case-based Tutoring Kishore Peravali
Restauranter Free Text Query Interface Atif Majeed
.
Virtual Cell Project
Project Collaborator(s)
Protein/Receptor/Inhibitor Agents in VRML Dr. McClean, Dr. White, Brad Vender, Yihe Wu
Virtual Cell Graphical/VRML Java-MOO Client Brad Vender
Virtual Cell Graphical/VRML User Interface Aaron Bergstrom, Mark Rose
Organelle and Instrument MOO Simulation Dr. McClean, Dr. White, Faye Erickson, Lance Holden
Organelle and Instrument Agents in VRML <volunteer needed>
Transporter Proteins in VRML <volunteer needed>
.
Virtual World Building Tools Project
Project Collaborator(s)
Integrated Virtual World Building Tool Kuo-Di Jian, John Bauer
Virtual Abstraction Tool Yongxin "George" Jia (with Kuo-Di Jian)
Virtual Entity Tool Vidyalatha Nagareddy (with Kuo-Di Jian)
Virtual Map Building Tool Umesh Kedla (with Kuo-Di Jian, was Brad Vender)
SLATE Subjective Assessment Tool Atif Majeed, Ned Kruger
Deductive Tutoring Agent Tool <volunteer needed>
LambdaCore Upgrade Engine <volunteer needed> (was Brad Vender)
.
Visual/Virtual Computer Science Project
Project Collaborator(s)
The Visual Computer Program Dr. Juell, Srinivas Kolipaka, Srinivas Kanala
Machines in a Virtual Museum for Computer Science Education Curt Hill

* Details of Current Research and Development Projects

There are several sources of additional information on the following projects. For details, write slator@badlands.nodak.edu

* The Geology Explorer

This project, still in the design stages, is intended to teach college level geology students how to act like geologists. We plan to develop an earthlike planet in a synthetic environment and dispatch small teams of geology students to the surface in search evidence that the planet will support colonization. The first module will involve oil exploration, where students will be expected to plan an expidition, locate and assess potential oil deposits, and survive the somewhat hostile environment in order to report on it.

* The Retailer Project

An existing synthetic environment, designed to teach micro-economics and currently populated by a few human players and an army of extremely simple software agents, will be entirely re-developed with a new set of highly interactive software agents in order to create, for the first time, an interesting environment of "atmosphere", "infrastructure", and "tutorial" agents.

This project will also involve graphical client software development.

* General Descriptions of Research and Development Project Opportunities

The projects described below are fundamental to the entire research programme being pursued here. What's described below are the KINDS of projects that this research is moving towards.

* Case-based Development Tools

Development tools provide platform-independent support for the indexing tools and generalized software modules described below.

* Case/Frame Definition Language

Starting with the published definition of Memory Organization Packets (MOPs) published in Inside Case-based Reasoning (Riesbeck and Schank, 1989), which is a frame language written in Common Lisp, define the equivalent frame language in C/C++ such that it is complete and platform independent (capable of executing on both IBM compatible and PowerMac).

The language must support all the functions for creating new MOPs, creating and storing links between MOPs, and creating hierarchies and ordered sequences of MOPs with branches and loops. Functions must also be provided to save a file of MOPs in compiled form, to load a compiled file of MOPs, and write an equivalent ASCII file of readable MOP definitions.

The language should be developed in such a way that the indexing tools, described below, can create, delete, and modify files of MOP definitions.

* Database Interface Software

Develop a protocol and transport mechanism for storing and retrieving MOPs in a conventional (or object-oriented) multi-user database.

The challenge is that MOPs are hierarchical and dynamic data structures, and the MOP knowledge base (sometimes called a "memory") will be accessed by multiple tools and applications at the same time.

The server-side implementation must be platform independent, there must be client software capable of executing on both IBM compatible and PowerMac.

* Case-based Indexing Tools

Indexing tools are for the construction of knowledge representation structures (knowledge bases). These structures are, in turn, used by educational applications for tutoring, performance support, and game playing.

All tools must save and load network structures from both local MOP files and remote MOP databases.

* Case Abstraction Tools

A graphical user interface and MOP editing tools for creating abstraction hierarchies of concepts, entities, and relations to represent the content of a domain in terms of MOPs.

The tool must allow the user to "partition" their view of the knowledge base so that multiple users can simultaneously create MOPs, and so that a user can concentrate on the part of memory relevant to them without being overwhelmed by the knowledge base as a whole.

The tool must have access to a lexically sound concept library so that new concepts can be related to and defined in terms of previously existing concepts by integrating MOP memory with the structure of a machine-readable dictionary.

* Case Library Tools

A graphical user interface and MOP editing tools for creating MOPs and links between MOPs. The tool must produce arbitrary MOPs with slots and links defined by the user. The MOPs will represent cases and stories, with associated text, and will also represent concepts and procedures.

The links will be of four types:

  1. symbolic -- to represent relations such as IS-A and PART-OF
  2. associational -- to represent "reminding" between concepts and cases
  3. conversational -- to represent the categories of questions
  4. temporal -- to represent increments of time

The tool must have a monitoring and matching function that notices patterns in user defined MOPs and retrieves similar, relevant MOPs for the user's benefit. The user will use the retrieved MOPs to make compare and contrast decisions in order to assist in indexing their case.

The tool must have a graphical user interface and interview-based MOP editing tools for defining the concepts and features of a domain. The user will be a subject matter expert or knowledge engineer who will be defining a domain of knowledge both by cases and by domain feature.

* Conversational Network Tools

A graphical user interface and MOP editing tools for creating converstional networks, where cases and stories are linked by questions. Link types must be either selected from pre-defined sets or defined by the user.

The tool must permit nodes to be linked into abstraction hierarchies and must provide the user with lists of possible follow-up MOPs using simple rules of inference.

* Task Analysis Tools

A graphical user interface and MOP editing tools for creating ordered sequences of MOPs representing the steps, substeps, branching, and nested loops of a structured task. Each node in the task representation should be a MOP, as described above.

The tool must maintain a library of abstract tasks and plans so the user can choose a relevant, similar task structure and modify it to their own purposes. Similarly, the tool must have a monitoring and matching function that notices patterns in user defined task structures and retrieves past cases of task structures for the user's benefit.

The tool must support multiple views of the task: flow charts, lists, timelines, etc.

* Environment Building Tools

Environment building means constructing simulated situations and worlds where learners go to experience a simulated reality and gather skills.

* Exploratory Environment Tools

A graphical user interface to a platform-independent client that allows a user to define objects in a MOO (an object oriented Mud). The user will be able to define rooms and other spaces, complete with a graphical representation of each space, either provided by the user or chosen from a library of stored pictures.

The user will also be able to define objects in the environment, and will be able to link these objects to the concepts in a MOP memory.

* Agent Building Tools

A special kind of object in a MOO will be the agent object. An agent will be defined as the collection of concepts and cases from a particular domain of knowledge that are known to them, where those cases are connected together with conversational links. Agents will be represented graphically as human-like avatars. They will be dispatched to engage environment dwellers when the environment's tutor detects a case based match with that dweller's state. The agent will approach the dweller, and offer to present a case and then discuss it. An agent will also be a repository of relevant documents, pointers to other cases, and problem-solving advice.

* Tutor Building Tools

A tutor will be an agent-dispatching mechanism that monitors the dwellers in an environment and attempts to match a dweller's state against cases in the MOP knowledge base. When a dweller is matched against a case, an agent representing that case will be dispactched to advise the dweller.

Tutor building tools will include a graphical user interface to a platform-independent client that allows a user to define rules of inference and a case-based matcher and retrieval mechanism.

* Outcome-based Simulation Tools

Continuous simulations are notoriously difficult to implement, test, and maintain. For the purpose of teaching with simulated environments, it is cost- and effort-effective to implement discrete simulations, also known as outcome-based simulations.

An outcome-based simulation is a network representation of a task environment. Nodes in the network represent states, places and objects, and links represent events that transition between nodes. In this way, suficiently rich problem spaces are defined that challenge a dweller to achieve some goal, while at the same time constraining the dweller to a finite, albeit large, set of possible actions at any given state.

* Generalized Software Modules

Generalized software modules are the network browsers and case retrievers that actually go in to the building of educational software.

* Generalized Conversational Browser

Given a network of MOPs representing nodes in a conversational network, build a platform-independent brower that will traverse the network, following question links from node to node. At each node, the content, text, video, or both, will be displayed, as well as the question links to other nodes.

* Generalized Case Retriever

Given a case library indexed by surface feature MOPs and abstraction hierarchy nodes, manage a queue of cases such that cases are matched against dweller descriptions and retrieved for dweller inspection.

* Generalized Simulation Engine

Given a network representation for an outcome-based simulation, manage the events and transitions that take a dweller from state to state. Hooks must be provided for integrating case retrieval mechanisms and to support conversational browsing.

Hooks must also be provided to the tools and other resources available in each state of the simulation.

* Generalized Task Monitor

Given a sequenced collection of task-defining MOPs, manage the events and transitions that take a dweller from state to state. Hooks must be provided for integrating case retrieval mechanisms and to support conversational browsing.

Hooks must also be provided to the tools and other resources available in each step of the task.

* Currently Inactive Research and Development Projects

Many of the projects listed above have their roots in the older project ideas outlined below.

* The Polymer Tutor

The Polymer Tutor is a performance support system to assist "formulators" in using the "Formulation Expert" system. This system is for choosing ingredients for highly resistant coatings. The domain is Polymer Chemistry, and it requires highly specialized expertise to successfully create new coatings that are durable and cost-effective. The Formulation Expert system itself requires considerable expertise to use.

The Polymer Tutor is intended to assist specialists in the use of the Formulation Expert system by providing context-sensitive help and advice at every step along the (quite complex) process of formulation, and to provide case-based remindings of relevant stories of success and failure, to assist the formulators in their work.

* The Anthropology MOO

A text-base synthetic environment was initiated in cooperation with the (now defunct) NDSU Anthropology Club. The ideawas to take the approach of first designing and creating an exploratory text-based world, in order to simulate the experiences of a working anthropologist, and then later augmenting that with a graphical user interface.

Last modified: Sat, July 4, 1998, 14:54:41 PM
Last modified: Sun, June 28, 1998, 12:34:41 PM
Write slator@badlands.nodak.edu with comments, suggestions, proposals, or project ideas.