Skip to main content

Table 7 Characteristics of agent-based and system dynamics models, and a hybrid model (Martin and Schlüter 2015)

From: Business ecosystem modeling- the hybrid of system modeling and ecological modeling: an application of the smart grid

 Agent-based modelSystem dynamics modelHybrid model
Characteristic questionHow do emergent system-level interaction (e.g. spatially, between individuals?)How do stocks change or stabilize? (given that rates are constant) Which process/feedback is dominating?How do changing process rates (impacted by decisions) affect dynamics?
How do changing stocks affect agent states/the distribution of traits?
Purposes in general for all: improve system understanding rather than prediction or forecastingTo identify mechanisms (specific interactions) that are responsible for emerging system-level patterns (disaggregated)
Generate hypotheses, exploration of micro-level behavior.
Investigate system-level dynamics (aggregated), stability properties of the system, loop dominance, explaining temporal dynamics, projection into the future.Investigating different micro- or system-level mechanisms that drive certain dynamics. Generate hypotheses of system state-change (when does dominance of feedbacks change?) or structural development over time (when does an average trait of agents change?)
FocusMicro-level interactions between entities, network structure (heterogeneous characteristics of individuals/actors, temporal discrete behavior), transient dynamics.Processes driving accumulation in stocks at (sub-)system level, stable-states, feedbacks (balancing, amplifying), non-linearities.Process of restructuring in a system which can focus either on a structure affecting the processes, or processes affecting the structure.
Tests for model calibrationStatistical pattern matching-can the model grow patterns that are found in reality?Stability analysis-under which parameter setting can fixed points/equilibria occur? How stable are they?Separate sub-system test (paradigm specific) and qualitative check for the coupled version.
Suitable and traditional analysis tools, typical experimentsOnly through simulations, often with multiple repetitions because of stochastic elements: plotting group/system-level characteristics over time (average), evaluation a limited parameter range, describing transient dynamics.Simple models through analytical tools (basins of attraction, bifurcation analysis, overall stability), and more complex through simulations (state space plots from simulations, evaluating stable-states, equilibria.)Through simulations with a focus on either
1. change in structure/ parameters: how does it affect the dynamics?
2. Change in dynamics: how does it affect the structure?
Type of outcomeEmerging spatial/agent patterns, scenario comparison between structurally different model versions, system properties such as the average state of a population.Aggregated system properties in terms of stability, loop dominance.Time series of merging state-transitions.