Simulation in Virtual Knowledge Communities

1
Simulation in Virtual Knowledge
CommunitiesJacques CALMETIAKS, University of
KarlsruheJulien SUBERCAZEIAKS, University of
KarlsruhePierre MARETLIRIS (CNRS UMR 5205),
INSA de Lyon
1
2
Structure of the speech
Introduction Simulation definition,
implementation Basic measures for the
analysis Experimental results Conclusion
2
3
Introduction

• Virtual Knowledge Communities
• Groups of people with similar interests who
  communicate and interact in an online environment
  John Beech
• A virtual community is a group whose members
  are connected by means of information
  technologies Experts archive question

3
4
Scope

• Multi Agent System approach for VKC
• Agents members of the community, the
  community architecture can also be made of agents
• Communication channels all the possibilities
  offered by MAS internet, telephony

4
5
Agent Oriented Abstraction

• AOA Paradigm Calmet, Maret, Endsuleit
• Agents are made of two components
• Annotated knowledge Knowledge annotations are
  classes or types structuring the knowledge
  possessed by or associated
• to agents
• Decision mechanism process by which an agent
  can reach its assigned goals. It is based upon
  the contents of the knowledge component. A
  decision mechanism is characterized
• by its utility.

5
6
Simulation Why?

• To model knowledge communities and test the
  models
• Corporate knowledge
• Knowledge distribution like Hospital
• Evaluate the performance of the frameworks
• Communication performance
• Stability
• Compare some prototypes
• Knowledge distribution efficiency

6
7
Simulation - Process
Design of the scenario
Execution of the scenario. Collect data
Analyse of the collected data
7
8
Design of the scenario

• A scenario is a tuple ltS,A,C,Z,R,Tgt
• Prototype system S
• Set of agents A
• Set of agent's classes C
• Set of actions Z
• Set of relations between the agents R
• Time length T

8
9
Collecting the data

• Three classes of data can be collected
• Communication data the exchanged messages
• Satisfaction level of the agent
• Knowledge of the agent

9
10
Implementation

• Prototype for VKC at the IAKS Karlsruhe
  Calmet,Hammond,Maret
• Follows the AOA paradigm Knowledge and
  decision
• Knowledge Management using RDF Ontologies
• Virtual communities knowledge exchange using a
  blackboard
• Community founding, searching, joining
• Agents can have their own behaviour, currently
  individual and social behaviours are implemented

10
11
Simulator

• Scenario Editor

11
12
Basic measures for the analysis
For BDI agents, we define the achievement rate
f State -gt 0,1 Combined with the
communication data
12
13
Effective Knowledge Distribution
Knowledge set at the time t KA,t Global
knowledge during the simulation Effective
knowledge distribution
13
14
Experimental Results

• Statistical test for the measures, over 3 trivial
  systems
• Dummy agents send randomly knowledge to others
• Agents ask randomly other agents for knowledge
  matching their interest.
• Yellow page system agents ask other agents
  having knowledge matching their interest

14
15
Experimental Results Achievement rate

15
16
Experimental Results EKD

16
17
Conclusion

• We introduced
• Formal definition of simulation
• Definition of basic measures
• Future work
• Method to collect the data on distributed system
• Collect and analyse data on the prototype
• Local method for efficiency evaluation

17