Towards A Methodology For Agent Based Social Simulation Research (Research-in-Progress)

1
Towards A Methodology For Agent Based Social
Simulation Research(Research-in-Progress)
AGENTLINK, February 3-4 2003, Barcelona, Spain

• By A.M. Ramanath and N. Gilbert
• (University of Surrey, UK)

ABSSMSEAS MEETING
2
AGENDA

• Quick Introductions
• Introduction
• Computer-based Simulation as a Research Method
• Multi-agents as a Social Simulation Research
  Stream
• Current Approaches to Multi-agent Model Building
  and Evaluation
• Lessons from Previous Multi-agent Based Social
  Simulation Projects
• Methods derived from Software Engineering
• Methods for Stakeholder Exploration/Evaluation
  from Social Sciences
• One Suggested Approach to MABS
• Conclusions

ABSSMSEAS MEETING
3

• Information Engineering Background
• Industry
• PhD
• Research interests
• Research Fellow

QUICK INTRODUCTIONS
4
SimWeb

• will provide European businesses in the digital
  contents sector with insights and tools which
  will enable them to take informed business
  strategy decisions and become more competitive by
  adapting their traditional business models to the
  new, demanding reality(www.simdigital.com)
• (Partners University of Surrey, CIMNE, iSOCO,
  Universität Koblenz-Landau, Publico.pt, FNAC)

INTRODUCTION
5
Generic Stages in Simulation Research Process A
third way of doing science
Common Pitfalls (from Thesen Travis,
1995) Objectives Failure Unanswerable
Questions Inappropriate Complexity Bad
Assumptions in Model Misinterpretation of
Outputs Budget Time overruns
A Process where quality and productivity
matter.
COMPUTER-BASED SIMULATION AS A RESEARCH METHOD
6
Subtle difference between MAS and MABSIf
the MAS field can be characterised as the study
of or implementation of societies of artificial
autonomous agents, agent-based social simulation
can be defined as the study of artificial
societies of autonomous agents (Conte et al.
1998, page 3).

• Common Constructs
• cognition, behaviour, emergence, second-order
  emergence, flexibility,
• decentralisation, beliefs, desires, intentions,
  self-organisation,
• robustness,intelligence, and complexity, to name
  but a few

MULTI-AGENTS AS A SOCIAL SIMULATION RESEARCH
STREAM
7
More efforts are needed on how to do MAS and
MABS

• For MABS in particular further Research is
  needed in 3 Areas...
• 1) Lessons from Previous MABS Projects
• (from MABS literature)
• 2) Advances in the Neighbouring Field of Software
  Engineering
• (from S. E. Approaches / Methodologies Research
  and Practice)
• Effective Stakeholder Involvement
• (from scientific advances in the social
  sciences)

CURRENT APPROACHES TO MULTI-AGENT MODEL BUILDING
EVALUATION
8
AUTHORS DOMAIN PROJECT (s) ACRONYMS (s) MAS Platform Simulation Steps Covered
D Aquino et al. (2002) Ecosystem Management SHADOC Stratagenes Sylvopast Mejan SelfCormas CORMAS All of them
Gilbert et al. (2002) Renewable Resource Management (Water) FIRMA PHP, PostgreSQL Apache, Javascript, email Conceptualisation through to Publication
Purnomo et al. (2002) Renewable Resource Management (Forests) - CORMAS Conceptualisation through to Publication
Considerations derived from the review of these
studies - Generic - Conceptualisation/Design -
Build/Verification - Evaluation/Publication
iteration, collaboration, exploration (Key
Aspects Expected In MABS Approaches Techniques)
LESSONS FROM PREVIOUS MABS PROJECTS
9
Methodology or Not Methodology?

• Problems that remain (in S.E.) despite
  Methodologies.
• Quality (the capacity of systems/software to
  meet stated requirements)
• Productivity ( the completion of
  systems/software on time and on budget)
• (from Tudor Tudor, 1995)

METHODS DERIVED FROM SOFTWARE ENGINEERING
10
Some Known Methods SSADM PSP/ CMM RAD RUP XP etc
.
Software Life Cycle (Universal paradigm (dates
back to1930s systems theory 1940s
cybernetics and reductionist approaches during
WWII)

• More recently move towards Agility, Speed
  Productivity

METHODS DERIVED FROM SOFTWARE ENGINEERING
11
RUP(Rational Unified Process)Heavyweight
titleversus Lightweight titleXP(Extreme
Programming)
METHODS DERIVED FROM SOFTWARE ENGINEERING
12
Two Particularly Relevant Scenario
Analysis and Participatory Modelling
Mapping out diversity
Policy Exercises
Focus Groups
Scenario Analysis
Participatory Modelling
Advising
Democratisation
Participatory Planning
Citizen Juries
Consensus Conferences
Reaching Consensus
(from van Asselt et al. 2001)
METHODS FOR STAKEHOLDER EXPLORATION / EVALUATION
FROM SOCIAL SCIENCES
13
AN APPROACH TO MODEL BUILDING AND EVALUATION
14
AN APPROACH TO MODEL BUILDING AND EVALUATION
15

• (A conversation said to have taken place
  between GM executives and Bill Gates of
  Microsoft, where Gates compared GM and
  Microsofts achievements as follows)
• If automotive technology had kept pace with
  computer technology over the past few decades,
  you would now be driving a V-32 instead of a V-8,
  and it would have a top speed of 10,000 miles per
  hour. Or you could have an economy car that
  weighs 30lbs and gets a thousand miles to a
  gallon of petrol. In either case the sticker
  price of a new car would be less than 50. GM
  responded Yes, but would you really want to
  drive a car that crashes twice a day?
• (Computer Weekly, 12/06/97).

CONCLUSIONS