Agent-Based (AB) compared with System Dynamics (SD) Simulations

"System dynamics, which is connected to the work of Forrester (1961), grounds modelling on difference equations and impinges upon the assumption that the behaviour of individuals that are embedded within a social system can be explained by the feedback nature of causal relationships that characterises the structure of the system.

Agent-Based models, on the other hand, simulate actions and interactions of autonomous individual entities and build on the hypothesis that the behaviour of social systems can be modelled and understood as evolving out of interacting but autonomous learning agents (Epstein and Axtell 1996; Axelrod 1997; Axtell 1999). Thus, a crucial feature of agent-based models is the emergence of ordered structures independently of top-down planning.

While Agent-Based models show how the interaction among individual decision-making and learning may generate complex aggregate behaviour, the SD approach aims at reducing emerging aggregate, and often puzzling, behaviours into underlying feedback causal structures. As a consequence, SD models typically aggregate agents into a relatively small number of states assuming their perfect mixing and homogeneity (Rahmandad and Sterman 2008). On the other hand, ABM preserves heterogeneity and individual attributes at the risk of relinquishing robustness and parsimony." --
-- Edoardo Mollona,
"Computer simulation in social sciences,"
J Manage Gov (2008) 12:205-211.

Some papers that compare the two:

Last Updated 11 August, 2010
Robert Marks,