A Multi-agent System
for Knowledge Management
based on the Implicit Culture
Enrico Blanzieri
Paolo Giorgini
Department of Information and
Communication Technology
University of Trento
Overview
• Implicit Culture framework
• Personal Agents for Implicit Culture Support
• A MAS for Knowledge Management
• Conclusions
Implicit Culture (CoopIS 2001)
Implicit Culture is a relation between a group of
agents G and a set of agents G’ such that the
agents of the set G’ behave according to
the culture of the agents of the group G.
A System for Implicit Culture Support has
the goal of establishing an Implicit
Culture Phenomenon, that is a pair of
group G e G’ related by Implicit Culture.
System for Implicit Culture Support
Observer
Inductive Module
Composer
System for Implicit Culture Support
Observer
Inductive Module
Composer
ε⊆ P ∪ O, P set of agents and O set of objects
System for Implicit Culture Support
Observer stores in a
database the situated
executed actions of
the agents of G.
Observer DB
Inductive Module
G
Composer
ε⊆ P ∪ O, P set of agents and O set of objects
System for Implicit Culture Support
Inductive Σ
Σ
0
Observer stores in a Module
database the situated
executed actions of
the agents of G.
Observer DB
Inductive Module using
the situated executed
actions in DB and the
domain theory Σo, induces G
a validated cultural
constraint theory Σ.
Composer
ε⊆ P ∪ O, P set of agents and O set of objects
System for Implicit Culture Support
Inductive Σ
Σ
0
Observer stores in a Module
database the situated
executed actions of
the agents of G. Composer Observer DB
Inductive Module using
the situated executed
actions in DB and the
domain theory Σo, induces
G’ G
a validated cultural a
σt+1
constraint theory Σ.
c σ”t+1
Composer proposes to
a group G’ a set of b σ’t+1
scenes σ such that
their expected situated
action satisfies Σ.
ε⊆ P ∪ O, P set of agents and O set of objects
The application
Knowledge Management
Make easier the information exchange
among a group of people
Building a Multi-Agent System
based on the
Implicit Culture Framework
Personal Agents
Each agent is implemented with JADE
(Java Agent Development Environment)
Jade Agent
Active Agent
Behaviours Module Inductive
Module Inductive
DB
scheduler Sics theory Σ
Behaviour
Composer
Composer Observer
Observer SICS
Inbox Resources
Resources Group G ’ Group G
Beliefs
Beliefs
Capabilities
Capabilities
SICS inside each single Agent
Each agent is implemented with JADE
(Java Agent Development Environment)
Jade Agent
Active Agent
Behaviours
scheduler
Sics www.fs-on-line.it
http
Behaviour
Inbox Each agent has beliefs about
Resources
a local schema for organizing
Beliefs the information available to
Capabilities the user
SICS inside each single Agent
Each agent uses locally the SICS, which
observes the actions of the user and those of the
other agents
a specific communication protocol allows the SICS to
observe the actions executed by the user and the other
agents in response of its suggestions
induces theories about the user and the other
agents
modify the scenes suggesting
the user
the other agents
the agent itself
The Multi-agent System
User FIPA - platform Directory Facilitator (DF)
provides a yellow pages directory
SICS SICS
service to the agents.
PA1 Pan
Agent Resource Broker (ARB)
provides information about the
SICS
resources available outside the
SICS PAk multiagent system.
PA2
DF
User Personal agent is created and
ARB assigned to each user who
accesses the system by means of a
web browser.
User
Example
platform 1. Request(user1, pa1, ”traintable”)
SICS
PA1
User1
SICS SICS
PA3 PA2
Example
platform 1. Request(user1, pa1, ”traintable”)
SICS 2. Request(p1, p2, ”traintable”)
PA1
User1
SICS SICS
PA3 PA2
Example
platform 1. Request(user1, pa1, ”traintable”)
SICS 2. Request(p1, p2, ”traintable”)
PA1
3. Inform(pa2, pa1, ”info@trenitalia.it”+”pa3”)
User1
SICS SICS
PA3 PA2
Example
platform 1. Request(user1, pa1, ”traintable”)
SICS 2. Request(p1, p2, ”traintable”)
PA1
3. Inform(pa2, pa1, ”info@trenitalia.it”+”pa3”)
User1
4. Request(pa1, pa3, ”traintable”)
SICS SICS
PA3 PA2
Example
platform 1. Request(user1, pa1, ”traintable”)
SICS 2. Request(p1, p2, ”traintable”)
PA1
3. Inform(pa2, pa1, ”info@trenitalia.it”+”pa3”)
User1
4. Request(pa1, pa3, ”traintable”)
5. Inform(pa3, pa1, ”www.viaggi.it”)
SICS SICS
PA3 PA2
Example
platform 1. Request(user1, pa1, ”traintable”)
SICS 2. Request(p1, p2, ”traintable”)
PA1
3. Inform(pa2, pa1, ”info@trenitalia.it”+”pa3”)
User1
4. Request(pa1, pa3, ”traintable”)
5. Inform(pa3, pa1, ”www.viaggi.it”)
6. Inform(pa1, user1,”info@trenitalia.it”+“www.viaggi.it”)
SICS SICS
PA3 PA2
Example
platform 1. Request(user1, pa1, ”traintable”)
SICS 2. Request(p1, p2, ”traintable”)
PA1
3. Inform(pa2, pa1, ”info@trenitalia.it”+”pa3”)
User1
4. Request(pa1, pa3, ”traintable”)
5. Inform(pa3, pa1, ”www.viaggi.it”)
6. Inform(pa1, user1,”info@trenitalia.it”+“www.viaggi.it”)
SICS SICS 7. Inform(user1, pa1, accept(“www.viaggi.it”))
PA3 PA2
Example
platform 1. Request(user1, pa1, ”traintable”)
SICS 2. Request(p1, p2, ”traintable”)
PA1
3. Inform(pa2, pa1, ”info@trenitalia.it”+”pa3”)
User1
4. Request(pa1, pa3, ”traintable”)
5. Inform(pa3, pa1, ”www.viaggi.it”)
6. Inform(pa1, user1,”info@trenitalia.it”+“www.viaggi.it”)
SICS SICS 7. Inform(user1, pa1, accept(“www.viaggi.it”))
PA3 PA2
8. Inform(pa1, pa3, accept(“www.viaggi.it”))
Example
platform 1. Request(user1, pa1, ”traintable”)
SICS 2. Request(p1, p2, ”traintable”)
PA1
3. Inform(pa2, pa1, ”info@trenitalia.it”+”pa3”)
User1
4. Request(pa1, pa3, ”traintable”)
5. Inform(pa3, pa1, ”www.viaggi.it”)
6. Inform(pa1, user1,”info@trenitalia.it”+“www.viaggi.it”)
SICS SICS 7. Inform(user1, pa1, accept(“www.viaggi.it”))
PA3 PA2
8. Inform(pa1, pa3, accept(“www.viaggi.it”))
9. Inform(p1, pa2, accept(“www.viaggi.it”))
The Implicit Culture Phenomenon
SICS
User1 requests information
SICS PA4
to Personal Agent PA 1
PA3
SICS
SICS
PA2 PA1 interacts with
PA6
SICS other agents
PA1
SICS
PA5
PA1 propagate the
information about
the action of the user
User1 to other agents
The Implicit Culture Phenomenon
SICS
There is an improvment
SICS PA4
of knowledge !
PA2
SICS
SICS
PA3
SICS
PA6
PA1
SICS
PA5
Implementation
We have implemented:
• user interface with PHP, Java
• MAS with JADE
• Communication among agents and users
Fipa-request
Extension of Fipa-iterated contract net protocol
• Local schema about the information avaible with XML
Conclusions
• The MAS incorporates a SICS in each Agent
• The SICS is used in order to provide information to the user
and to the other agents
• The SICS observes the local actions of the users and those of
the other agents
• The SICS aims at solving the problem to increase agent’s
performance through tacit knowledge exchange
• We are currently working on the inductive Module and we
want to use the system in a real case study
References
http://www.science.unitn.it/pgiorgio/~ic