Description
Simulates many heterogeneous agents following behavioural rules to see how collective patterns such as adoption, congestion or polarisation emerge.
Agent-based models represent individual users (or sellers, bots, moderators) with their own attributes and decision rules, interacting with each other and an environment. Macro patterns emerge from micro rules, often in surprising ways, as Schelling showed with mild preferences producing strong segregation. ABMs suit questions where heterogeneity, networks and local interaction matter. Rigour comes from grounding rules in empirical research (including qualitative), documenting with the ODD protocol, validating against multiple observed patterns (pattern-oriented modelling) and exploring parameter space systematically rather than presenting single runs.
Attributes
- Discipline
- Systems & Complexity
- Loss type
- Unmigrated
- Stage
- Strategy
- Data
- Analytical / expert
- Effort
- Months
Steps
- Define the question and the emergent pattern to explain or predict.
- Specify agent types, attributes, decision rules and interactions, grounded in research data.
- Document the model using the ODD protocol and implement it (NetLogo, Mesa, AnyLogic).
- Calibrate and validate against several observed patterns at different levels.
- Run parameter sweeps and sensitivity analysis with many replications.
- Test interventions in simulation and design field experiments to confirm key predictions.
Sources
- Epstein & Axtell, Growing Artificial Societies (1996)
- Grimm et al., The ODD protocol for describing agent-based and other simulation models: A second update (2020), JASSS
- Wilensky & Rand, An Introduction to Agent-Based Modeling (2015)
- van der Maas, H., Complex Systems Research in Psychology (2024)
Also known as
- ABM
- Individual-based modelling
- Multi-agent simulation
Categories
- Strategy & Collaboration
- Systems & Complexity