Towards a Data-driven Approach for Agent-Based Modelling: - PowerPoint PPT Presentation

Towards a Data-driven Approach for Agent-Based Modelling: Simulating Spanish Postmodernisation Defended by Samer Hassan Collado Directed by Juan Pavón Mestras Millán Arroyo Menéndez

Contents  Frame the problem  Define a methodology  Design an agent framework  Develop a case study  Validate through its results  Apply AI technologies 2 Towards a Data-driven Approach for Agent-Based Modelling

Introduction 3 Towards a Data-driven Approach for Agent-Based Modelling

Agent-Based Social Simulation  Simulation of Complex Systems  Nonlinearity: equations?  Self-organisation  Emergence: Total > ∑ parts  Non-deterministic  Social Simulation  Simulation of social phenomena  Agent-Based Social Simulation  Agent ~ Individual  Different from classical MAS  Simple micro rules => Complex macro behaviour Artificial Intelligence, Engineering, Physics, Mathematics, Philosophy,  Ecology, Economics, Sociology, Anthropology, Political Science, Psychology, Cognitive Science 4 Towards a Data-driven Approach for Agent-Based Modelling

Agent-Based Social Simulation 5 Towards a Data-driven Approach for Agent-Based Modelling

Agent-Based Social Simulation 6 Towards a Data-driven Approach for Agent-Based Modelling

Objectives  Build a methodology for DDABM  Design an agent framework  Develop a data-driven case study  Validates framework & methodology  Supports theoretical hypotheses  Applications of AI in ABM  Agents + Fuzzy Logic + NLP + DM 7 Towards a Data-driven Approach for Agent-Based Modelling

Methodological Approach 8 Towards a Data-driven Approach for Agent-Based Modelling

Methodological Approach  Review  Research Aim  Theoretical  Data-driven  Proposed Data-driven Approach  Stress on Data  Deepening KISS  Data-driven Cycle 9 Towards a Data-driven Approach for Agent-Based Modelling

Classical Logic of Simulation Gilbert (1999) 10 Towards a Data-driven Approach for Agent-Based Modelling

Methodological Approaches Positivism Goldspink (2002) McKelvey (1999) 11

Methodological Approaches Positivism Social Sciences Goldspink (2002) McKelvey (1999) 12

Methodological Approaches Positivism Social Sciences Social Simulation Goldspink (2002) McKelvey (1999) 13

Research Aim  Theoretical  KISS  Structural Validation  Abstract  General 14 Towards a Data-driven Approach for Agent-Based Modelling

Research Aim  Data-driven  Theoretical  Non-KISS  KISS  Empirical Validation  Structural Validation  Specific (case study)  Abstract  Expressive  General 15 Towards a Data-driven Approach for Agent-Based Modelling

Research Aim  Data-driven  Theoretical  Non-KISS  KISS  Empirical Validation  Structural Validation  Specific (case study)  Abstract  Expressive  General 16 Towards a Data-driven Approach for Agent-Based Modelling

Classical Logic of Simulation 17 Towards a Data-driven Approach for Agent-Based Modelling

Proposed Data-Driven Logic 18 Towards a Data-driven Approach for Agent-Based Modelling

Data-driven Approach  Empowering Empirical Data  Data Collection => Exhaustive  Random Initialisation => Empirical Initialisation  Abstraction => Data-driven Design  Structural Validation => Empirical Validation 19 Towards a Data-driven Approach for Agent-Based Modelling

Data-driven Approach  Deepening KISS  Gradually increasing complexity  Exploration of the model space Antunes et al (2006) 20 Towards a Data-driven Approach for Agent-Based Modelling

Proposed Data-driven Cycle 21

Proposed Data-driven Cycle 22

Data Collection 23

Proposed Data-driven Cycle 24

Data-driven Design 25

Methodology Highlights  Stress on Data  How to handle it  Deepening KISS  Guides agent framework architecture  Guides model extension  Data-driven Cycle  Guides ABM building 26 Towards a Data-driven Approach for Agent-Based Modelling

The Sociological Problem 27 Towards a Data-driven Approach for Agent-Based Modelling

The Sociological Problem 28 Towards a Data-driven Approach for Agent-Based Modelling

Social Phenomena 29 Towards a Data-driven Approach for Agent-Based Modelling

Case Study  Sociological problem  Change in social values • Mainly: political & religious  Specific context • Time: period 1980-2000 • Space: Spanish society  Validating  Theory from Sociology  The Data-driven methodology  The Agent framework Arroyo (2004, 2005) 30 Towards a Data-driven Approach for Agent-Based Modelling

Sociological Theory 31 Towards a Data-driven Approach for Agent-Based Modelling

The Change of Values  Cultural Modernisation  Individualisation • Discourages Authorities • Encourages Self-expression  Post-Materialism • Based on Maslow hierarchy • Material welfare => Post-Mat. Priorities  Dynamics  Inter-generational  Intra-generational Inglehart (1997) Halman (1994) Maslow (1987) 32 Towards a Data-driven Approach for Agent-Based Modelling

The Change of Values  Inter -generational dynamics  Change across generations  Demography  Intra -generational dynamics  Change within a generation  Internal evolution in a person course of life  Inglehart  Socialisation in youth  Stability over course of life  Inter-gen >> Intra-gen Inglehart (1997) Halman (1994) Maslow (1987) 33 Towards a Data-driven Approach for Agent-Based Modelling

Inglehart’s hypothesis  Inter -generational dynamics  Change across generations  Demography  Intra -generational dynamics  Change within a generation  Internal evolution in a person course of life  Inglehart  Socialisation in youth  Stability over course of life  Inter-gen >> Intra-gen 34 Towards a Data-driven Approach for Agent-Based Modelling

Axiomatic Base  Model for the change of values  Focus: political & religious values  Demography • Intergenerational dynamics  Socialisation in youth • Stability over course of life 35 Towards a Data-driven Approach for Agent-Based Modelling

Modelling Decisions 36 Towards a Data-driven Approach for Agent-Based Modelling

Theoretical Modelling Decisions  Demography  Static: • Population pyramids  Dynamics: • Reproduction <-- Partner • Partner <-- Friendship • Friendship <-- Social Network  Socialisation  Inheritance of crossed values  Stability in course of life 37 Towards a Data-driven Approach for Agent-Based Modelling

Methodological Modelling Decisions  Empowering Empirical Data  Exhaustive Data Collection  Empirical Initialisation  Data-driven Design  Empirical Validation 38 Towards a Data-driven Approach for Agent-Based Modelling

Methodological Modelling Decisions  Empowering Empirical Data  Exhaustive Data Collection • Multiple data sources & waves  Empirical Initialisation • European Values Study • Representative sample: EVS-1980  Data-driven Design • Demography • Equations (e.g. birth rate) • Qualitative info from domain expert • Life cycle, Social relations  Empirical Validation • EVS-1990, EVS-1999 39 Towards a Data-driven Approach for Agent-Based Modelling

Model Insight 40 Towards a Data-driven Approach for Agent-Based Modelling

Mentat: Architecture 41 Towards a Data-driven Approach for Agent-Based Modelling

Mentat: Architecture  Methodological implications  Agent Framework design  Global pattern: MVC  Agent pattern: Layers  Model description  World  Agent 42 Towards a Data-driven Approach for Agent-Based Modelling

Mentat: Architecture  ‘Deepening KISS’ approach  Gradually increasing complexity • Modularity & Flexibility  Exploration of the model space • Activating independent modules • Facilitate extension • Architectural patterns  Architectural pattern  Fundamental structural organization schema  Subsystems, responsibilities, interrelations 43 Towards a Data-driven Approach for Agent-Based Modelling

Mentat Architecture: MVC 44

Agent Architecture: Layers 45 Towards a Data-driven Approach for Agent-Based Modelling

Agent Architecture: Layers 46 Towards a Data-driven Approach for Agent-Based Modelling

Agent Architecture: Layers 47 Towards a Data-driven Approach for Agent-Based Modelling

Agent Architecture: Layers 48 Towards a Data-driven Approach for Agent-Based Modelling

Agent Architecture: Layers 49

Mentat Model  World:  3000 agents  Grid 100x100  6 indep. parameters  Social Network:  Communication with Neighbourhood  Friend Network  Family Network 50 Towards a Data-driven Approach for Agent-Based Modelling

Mentat Agent  General characteristics  Gender, Age  Education, Economy  Religiosity, Political Ideology  Tolerance levels  Euthanasia, Suicide, Homosexuality, Abortion, Divorce, etc  Life cycle  Child, Adult, Elder 51 Towards a Data-driven Approach for Agent-Based Modelling

Mentat Agent: Life Cycle 52 Towards a Data-driven Approach for Agent-Based Modelling