Justintime Knowledge Flow for Distributed Organizations using Agents Technology

1
Just-in-time Knowledge Flow for Distributed
Organizations usingAgents Technology

• This presentation will probably involve audience
  discussion, which will create action items. Use
  PowerPoint to keep track of these action items
  during your presentation
• In Slide Show, click on the right mouse button
• Select Meeting Minder
• Select the Action Items tab
• Type in action items as they come up
• Click OK to dismiss this box
• This will automatically create an Action Item
  slide at the end of your presentation with your
  points entered.

• R. Brena, J.L. Aguirre, A.C. Treviño
• ITESM, Mexico

2
Who are we?

• Joint project between a lab in AI and a Bussiness
  research center at the ITESM, Monterrey
• Center for Artificial Intelligence
• Center for Knowledge Systems

3
Summary

• Just-in-time Knowledge
• Agents technology
• Architecture
• Ontologies
• Services
• Contexts
• Discussion
• Conclusions

4
Knowledge economy

• Increased relative weight of knowledge and
  information in the production of value
• Knowledge is critical for shrinking cycle time
  for competency-base renewal
• Pressure for most organization to cope with
  massive flood of unstructured information

5
Knowledge life cycle
K acquisition / discovery
K distribution
K use
6
Knowledge flow

• Often K is created in one place and needed (often
  not used) in another
• Many large distributed organizations suffer from
  a lack of K circulation both in vertical and
  horizontal directions
• Good dissemination of relevant pieces of
  information and knowledge is very important

7
JITIK concept

• Just-in-time Information and Knowledge
• Give support to K circulation in the organization
  by connecting the right K with the right person
  at the right moment

8
Right Knowledge

• Knowledge pieces are characterized as points in a
  multidimensional space
• Dimensions are class hierarchies
• These come from ontologies
• Examples
• Classification of users interest areas
• Structure of the organization

9
Right person

• Right knowledge and Right person are
  reciprocal
• We use the same categories for characterizing
  users and knowledge
• This makes possible to link pieces of knowledge
  to corresponding users

10
Right moment

• The time for sending a piece of knowledge could
  be
• When that K is generated
• When it arrives to the organization or is
  discovered
• When a particular users needs it
• The user knows he/she needs it (point-and-click
  search)
• The user is not aware of needing some information
  (K is diffused to some specific users)

11
Agents technology

• Long-lived autonomous processes
• Reactive and proactive
• Cooperate and compete
• Users delegate tasks to agents (electronic
  assistants)

12
Point Click / Delegate

• Old paradigm Point and click
• Computer does just what user directs it to do
• New paradigm Delegate
• Computer takes care of tasks and reports results
  to the master

13
Search / Diffuse

• Usually K is searched by users
• (point and click paradigm)
• In JITIK K is diffused to users as it becomes
  available / relevant
• (delegation paradigm)

14
Architecture
Directory agent
Parasiticagents
Clusteragents
Personal Agents
15
Cluster agents
InferenceEngine
DB
Workingdata
Communicationinterface
16
Inference engine

• Forward-chaining rule-based
• Rules of the form
• Distribute K satisfying Char(K) to U satisfying
  Char(U) when E
• Distribute www_library_regulation satisfying
  relevant(library_regulation, new_users) to ?x
  satisfying new_registered(?x) when NOW.

17
Directory agents
Directory agent

• Help finding other agents
• Like DNS services
• Give great flexibility to the systems
  (re)configuration

18
Personal agents

• Take care of services specific to an individual
• Examples
• Monitoring specific web pages
• Checking ranges for values in databases
• Reporting articles from netnews related to
  interest areas
• Maintain a user profile

19
Parasitic agents

• Allow JITIK to gather information from other
  systems
• Parasitic agents are fastened to conventional
  programs
• PA report to cluster agents
• Identity of the PA
• Event being reported
• Associated data
• PA find cluster agents using DA

20
Knowledge/Users specification

• Combination (and) of several class hierarchies
• Example
• User in the sales division (organization
  classification), and
• User is executive in charge of a department
  (level classification), and
• User involved in best-practices support (tasks
  classification), and
• User expert in customer satisfaction (competence
  classification)

21
Ontologies

• The set of class hierarchies is a form of
  ontology
• Terms are defined by their place in class
  hierarchies
• Expression of ontologies in an ontology-oriented
  language is under way

22
JITIK services

• Alerts
• Messages
• K from bussiness processes

23
Alerts

• Alerts are brief notification of events
• A new relevant piece of information has arrived
• An important information has changed
• A deadline has been reached
• They are reported to personal agents, which
  notify their user according to his/her profile

24
Messages

• Messages are created by a human user
• The delivery time could be now or when some
  conjunction of conditions become true
• Messages are delivered in a form chosen by the
  user (email, instant messaging, web)

25
K from bussiness processes

• Parasitic Agents are attached to key points of
  bussiness processes
• Information from PA is associated to a
  specification of relevant K
• Relevant K is then distributed among concerned
  users
• Example New user registers at a library.- Send
  him/her regulations

26
JITIK in context

• JITIK could be applied to a variety of
  organizational contexts
• Big distributed / multinational / vertical
  entreprises
• Virtual organizations and networks
• Task forces

27
Best-practices support

• Best practices are hard to generalize
• Quality circles run without much automated
  support
• JITIK can distribute best practices to relevant
  users
• Could help to break geografical / organizational
  distances

28
Virtual organizations

• Users are spread over the world
• They share common interests / activities (e.g.
  Medical specialists)
• JITIK could help to distribute relevant alerts,
  information and knowledge

29
Discussion

• Personal computers or server networks?
• PC are not running and connected around the clock
• Permanent tasks could not be delegated
• Personal agents need to take care of their
  delegated tasks on a permanent basis

30
A word on methodology

• Currently there is no methodology to create rules
  relating knowledge to users and events
• Rules are themselves a form of K about the
  organization (e.g. Intellectual capital)

31
Project status

• Web-based client-server technology version
  running
• Java programming
• Jess inference engine
• Full agent-tech version under development

32
Future work

• Broader scope in K life cycle
• Tighter integration as KM support tool (e.g. K
  maps, K monitoring)
• Sophisticated AI tools (Case-based reasoning,
  Data mining, Knowledge repositories)
• Integration with workflow systems and CSCW
• Integration with XML and related technologies

33
Conclusions

• JITIK is a method for relating a piece of
  knowledge or information, specied as mentioned,
  to a set of users, also characterized in an
  abstract way, when an event is produced
• JITIK provides diffusion of the right K to the
  right person at the right time