Agent-based Modeling and Simulation for the Social Scientist

1
Agent-based Modeling and Simulation for the
Social Scientist

• MAIA

Amineh Ghorbani, Virginia Dignum, Pieter Bots,
Gerard Dijkema, Bert Belder
2
Goal

• Framework for agent-based conceptualization and
  simulation
• Rich enough to capture a diverse range of social
  systems
• Support developers with little/no
  programming/software engineering knowledge
• Application areas
• Policy design / public goods problems
• Social systems complex behavior / discrete
  entities
• Approach
• Collaborative modelling
• Institutional analysis (Ostrom)
• Model driven engineering (MDE)
• meta-modeling and semi-automatic code generation

3
Applications

• Domains

• Commonalities

• Wood-fuel market
• E-Waste recycling
• Consumer lighting
• Basic income grants
• Family-based care

• Domain characteristics
• What-if analysis of policies
• Problem-owners /domain experts had limited
  simulation knowledge

4
Common characteristics

• Effect of incentives / policies
• Social networks and institutions
• Individual interests
• Global consequences
• Multi-criteria decision making

5
What is MAIA?

• Modeling Agents based on Institutional Analysis
• Formal meta-model
• Institutional perspective (IAD Ostrom)
• Web based design tool
• Declarative rather than procedural
• Semi-automatic simulation generation

6
MAIA Architecture

• The MAIA meta-model

finetuning
7
Institutions

• An institution is any structure or mechanism of
  social order and cooperation governing the
  behavior of a set of individuals within a given
  human community.
• Institutions are identified with a social purpose
  and permanence, transcending individual human
  lives and intention by enforcing rules that
  govern cooperative human behavior

8
Institutions
1- Rules accepted by everyone 2- Used in
practice 3- Durability
By product of interactions
Rules created to manage activities
Individuals do activities (repetitive)
outcomes affect others too
9
Institutional frameworks

• Institutions have two sides
• Enable interactions, provide stability,
  certainty, and form the basis for trust.
• Cause power relations and may hamper reform.
• Important to understand effects of institutions

Institutional (re)design
Analyze and Understand for Design? Institutional
Frameworks
10
Institutional Analysis and Design
unit of analysis
Elinor Ostrom Nobel laureate 1933-2012
11
Institutional Analysis and Development Framework
(IAD)

1. Participants
2. Positions
3. Actions
4. Potential outcomes
5. Functions that map actions into outcomes
6. Information
7. Cost and benefits

1. Position rules
2. Boundary rules
3. Authority rules
4. Aggregation rules
5. Scope rules
6. Information rules
7. Payoff rules

Action Arena
Physical world
Action Situation
Patterns of interaction
Community
Evaluation Criteria
Participants
Rules
Outcomes
Resources, preferences, information and selection
criteria
12
Extending IAD

• Formalization of concepts
• MAIA formal model
• Robust information and consensus
• MAIA online tool supports flexible
  conceptualization through participatory
  exploration
• Supports reflection and discussion
• Outward looking
• Information collected directly reflects the
  experiences and perceptions of stakeholders
  themselves

13
MAIA Meta model
14
Collective structure set of agents
15
Constitutive Structure
16
Institutions ADICO
17
Physical components
18
Operational structure
19
MAIA Modelling Environment
http//maia.tudelft.nl/
20
Translation to Java Code

• MAIA MM is developed as an e-core model
• EMF environment in Eclipse for model-driven
  software development.
• XML specification.
• Output of MAIA web-tool is based on MAIA MM
• Explicit, fixed, rules to convert MAIA model
  (XML) to Java simulation
• Current work translator code, for automatic
  generation of code from a MAIA-based model.

21
From rules to code
22
Agent behaviour
23
MAIA Approach
declarative
24
Conclusions

• MAIA framework for agent-based simulation
• Rich enough to capture a diverse range of social
  systems
• Support developers with little/no
  programming/software engineering knowledge
• Based on Institutional analysis (Ostrom)
• Formal model
• Verification
• Model driven engineering (MDE) for semi-automatic
  code generation

25
Future work

• Extend and validate code generation
• Visualisation of simulation results
• Library of agent behaviours
• Extensive evaluation
• Transformation of MAIA models into other
  simulation environments (e.g. Netlogo or Repast)

26
Benefits of MAIA

• Four completed case studies 5 other users
• Diverse range of concepts makes you think about
  the things that you may need to consider in your
  model.
• Easy to follow by no-programmers
• Team of Developers

MAIA
MAIA
27
MAIA Architecture
More info a.ghorbani_at_tudelft.nl