First International Workshop on Agentization: Rendering Conventional Models with Agent-Based Computing

Keynote talk from W. Brian Arthur, Santa Fe Institute

Agent-based modeling (ABM) is an important computational methodology that has found wide application in the social and natural sciences. Useful for both conceptual and empirically-grounded models, ABMs can serve as a powerful technique for relaxing unrealistic model specifications, such as agent homogeneity or rationality. While ABMs are often built to study social phenomena that have been previously investigated using conventional (analytical, statistical) methods, the relation of such ABMs to conventional results is often opaque, with little effort to reproduce the standard results computationally.


By agentization we mean the explicit creation of ABMs to reproduce the results from conventional models, so as to instill confidence that well-known work can be replicated with software agents. Then, the robustness of such conventional results can be assessed by relaxing the basic specifications of the original models—number of agents, nature of the environment, behavioral specifications, attainment of equilibrium—using ABMs. Free from the limitations of analytical tractability, ABMs can expose the dimensions along which conventional results are brittle or sensitive to parameter variations. Agentizations of certain prominent models may burnish their credibility by revealing that their basic results obtain more broadly than previously understood. Alternatively, models believed to have broad applicability may be shown to have limited scope via agentization.


This workshop will present agentized versions of several well-known models in economics and political economy, including Hotelling’s model of spatial competition, Akerlof’s market for “lemons”, Gale-Shapley matching, the O-ring theory of development, and Walrasian general equilibrium theory. For each, relaxation of certain specifications will be shown to lead to different results.


We solicit additional ABMs that can be suitably parameterized to reproduce conventional mathematical or statistical results and then generalized. Formal models originating in any of the social sciences will be the main focus but we are also open to models investigating natural science phenomena. We are particularly interested in ABMs that demonstrate the main findings of conventional models to be special cases of broader patterns. (If you are unsure whether your work constitutes ‘agentization’ please contact the organizers.)

The goal of this workshop is to produce a series of papers, suitable for publication as an edited book or special issue of a journal, that illustrate the use of ABMs to generalize conventional models, leading to new results and increased understanding of the phenomena being investigated.”

Deadline for submissions: August 16, 2021; notification of acceptance immediately thereafter

Share This Event
Event Details