Transcript Computational Organization Theory

Computational Organization Theory
Aaron Drajpuch
WPI, CS525M, Spring 2002





Introduction
What is an Organization
What is Computational Organization Theory
Concepts for Organizational Models
Examples
– SDML
– Swarm

References
Introduction

Organizations differ from Multi-agent systems
–
–
–
–
Exhibit a collective intelligence
Capabilities distinct from individual members
Influence and are influenced as a whole
Structure emerges
• Spontaneous
• Imposed
– Agents must be socially/organizationally aware
What is an Organization

Characteristics
–
–
–
–
–
–
Heterogeneous
Complex
Dynamic
Non-linear
Adaptive
Evolving
What is an Organization

Characteristics cont.
–
–
–
–
–
–
Comprised of multiple agents
Engaged in one or more tasks
Large scale problem solvers
Goal directed
Able to affect and be affected
Have knowledge, culture, memories, history, and
capabilities distinct from any single agent
What is an Organization

Organizations overcome agent limitations
–
–
–
–
Cognitive
Physical
Temporal
Institutional
What is an Organization

Choosing the right organization
–
–
–
–
–
–
Different problems require different designs
Tasks to accomplish
Intelligence and capabilities of agents
Environment
Legal and political considerations
Etc.
What is an Organization

Engineering perspective:
– Find optimal design for given problem

Theoretical perspective
– Specify general principles and tradeoffs
• Specialization
• Division of labor
Computational Organization Theory

Uses mathematical and computational
methods to study organizations
 Research into fundamental principles to
develop better models
 Develop tools and procedures for validating
organizational models
 Helps to formulate explicit and well defined
theories of organizations
Computational Organization Theory

Formal methods for organizations
–
–
–
–
–
–
Modelability
Performance differential
Manipulability
Designability
Practicality
Pragmatism
Organization Concepts

Agents
 Design
 Tasks
 Technology
SDML

Strictly Declarative Modeling Language
 Suited for flat or hierarchical structures
 Cognitive theory-neutral
– Implement agents represented in other languages

Each level in hierarchy is a container
– Lowest levels are cognitive agents
– Agents combine and move up the hierarchy
SDML Organization for Business
Business Organization
Universe
Business 1
Business 2
Business n
Department
Agent 1
Group 1
Group 2
Agent 2
Agent m
SDML Organization for Business



Each level contains the lower levels
Agents could be business processes
Business processes could combine and move up the
hierarchy
– Agents combine to form group business process
– Group business processes combine to form department
processes
– Etc.

Groups work in parallel, agents must communicate
Swarm

Simulation of complex, adaptive systems
 Uses object oriented libraries
 Swarm
– Collection of agents and other swarms
– Schedule of actions
– Representation of time

Can model different levels of detail
Swarm Example
Body
Organ
Organ
Germ
Organ
Cells
Germ
Swarm
Agent
Tissue
Swarm

Implemented with objects
 Inspired from Artificial Life
– Simulates evolution

Behavior emerges from the swarm
References
Kathleen M. Carley and Les Gasser. Computational Organization
Theory. Multiagent Systems: A Modern Approach to Distributed
Artificial Intelligence. The MIT Press, 1999.
 Marcus Daniels. Integrating Simulation Technologies With Swarm. A
paper submitted to the "Agent Simulation: Applications, Models and
Tools" conference held at University of Chicago in October of 1999.
 Nelson Minar, Roger Burkhart, Chris Langton and Manor Askenazi.


The Swarm Simulation System: A Toolkit for Building Multi-Agent
Simulations.
Scott Moss, Helen Gaylard, Steve Wallis and Bruce Edmonds. SDML:
A Multi-Agent Language for Organizational Modelling. Centre for Policy
Modelling, Manchester Metropolitan University. CPM Report No.: 9716, March 1997.
 Stuart Russell and Peter Norvig. Artificial Intelligence: A Modern
Approach. Prentice Hall, 1995.