A computational simulation approach that represents individual entities as autonomous agents with attributes and rules for behavior and interaction; system-level patterns emerge from local agent interactions rather than being imposed top-down.

Principle

Macro-level phenomena can be generated by simple micro-level rules; explicit modeling of heterogeneity, local interactions, and stochasticity allows exploration of emergent dynamics, spatial structure, and policy interventions.

Demonstration

An ABM of a fishery might model each fisher as an agent deciding where to fish based on local catch rates and social information; collective fishing pressure and collapse can emerge from individual movement and effort rules without prespecifying population trends.

Misapplication

Treating an ABM as automatically explanatory without validating agent rules or testing sensitivity to rule specification risks creating models that reproduce patterns for wrong reasons or that are overfitted to observed outcomes.

Consequence

Properly constructed and validated ABMs reveal mechanisms linking individual behavior to system-level outcomes, test alternative behavioral rules, and inform management by simulating counterfactual scenarios under uncertainty.

Reversal

Replacing agent heterogeneity and local interactions with aggregate equations (e.g., mean-field compartmental models) removes fine-grained emergence and enforces predefined population-level dynamics, losing potential behaviors produced by individual variance.

Boundary

Appropriate for systems where individuals, space, and local interactions matter; not optimal for well-mixed systems with negligible individual differences where simpler analytic models may suffice; results depend on rule choices and parameterization.

Semantic Tension

Tension exists with equation-based models that prioritize analytical tractability and generality; ABMs prioritize realism and heterogeneity at the cost of analytic solutions and sometimes reproducibility without rigorous documentation.

Synthesis

Agent-based models link explicit micro-level rules to emergent macro-level phenomena by simulating interacting autonomous agents, serving as mechanistic laboratories for hypothesis generation and policy experiments while requiring careful validation and sensitivity analysis.