A computational modeling paradigm that represents a system as a collection of autonomous, heterogeneous agents that follow simple rules for perception and action; the global behavior is obtained by simulating many agents and their local interactions to observe emergent system-level phenomena.

Principle

Bottom-up generation of macroscopic patterns from micro-level agent rules and local interactions; heterogeneity and interaction topology determine emergent outcomes rather than analytic closed-form aggregates.

Demonstration

A city pedestrian simulation in which individual agents choose routes and speeds according to local density and goal priorities; from these local rules, crowding, lane formation, and bottlenecks emerge without imposing global constraints.

Misapplication

Treating the model as a calibrated black box for prediction without sensitivity analysis, or aggregating agent behavior into mean-field statistics and assuming those capture phenomena driven by discrete heterogeneity and spatial structure.

Consequence

When used properly, ABMs reveal how individual variability, local rules, and interaction networks produce non-linear emergent phenomena, enabling exploration of interventions, scenario testing, and explanation of system-level surprises.

Reversal

A top-down aggregate model such as a mean-field differential-equation system that prescribes macroscopic state evolution without explicit individual agents or micro-level interactions.

Boundary

Appropriate when agents have autonomous decision rules, local interactions, and heterogeneity; not a good fit for strictly linear, globally coupled systems solvable by closed-form equations or when analytic tractability is required.

Semantic Tension

Tension with equation-based models and system-dynamics approaches that prioritize tractable aggregate descriptions; trade-offs exist between fidelity to individual behavior and analytical simplicity.

Synthesis

An Agent-Based Model is a computational instantiation of micro-level decision rules and interactions used to simulate and study how heterogeneous agents collectively produce emergent, system-level behavior.