Agent-based Computational Economics (ACE) is the computational modeling of economic processes (including whole economies) as open-ended dynamic systems of interacting agents.
Below are seven basic modeling principles underlying ACE model design. These principles reflect the fundamental goal of many agent-based modelers: namely, to be able to study real-world dynamic systems as historical processes unfolding through time, driven solely by their own internal dynamics.
(MP1)Agent Definition: An agent is a software entity within a computationally constructed world capable of acting over time on the basis of its own state, i.e., its own internal data, attributes, and methods.
(MP2)Agent Scope: Agents can represent individuals, social groupings, institutions, biological entities, and/or physical entities.
(MP3)Agent Local Constructivity: The decision-making process undertaken by a decision-making agent at any given time must be entirely expressible as a function of the agent's state at that time.
(MP4)Agent Autonomy: Coordination of agent interactions cannot be externally imposed by means of free-floating restrictions, i.e., restrictions not embodied within agent states.
(MP5)System Constructivity: The state of the modeled system at any given time consists of the collection of agent states at that time.
(MP6)System Historicity: Given initial agent states, all subsequent events in the modeled system are determined solely by agent interactions.
(MP7)Modeler as Culture-Dish Experimenter: The role of the modeler is limited to the setting of initial agent states and to the non-perturbational observation of model outcomes.
Together, (MP1) through (MP7) embody the idea that ACE models are computational laboratories permitting users to explore how changes in initial conditions affect outcomes
in modeled systems over time. This exploration process is analogous to biological experimentation with cultures in petri dishes. A user sets initial conditions for the modeled system in accordance with some purpose at hand. The "cover" is then closed, and the modeled system thereafter runs forward through time as a virtual world whose dynamics are entirely determined by the interactions of its constituent agents.
From a mathematical point of view, (MP1)--(MP7) imply that ACE models are state-space models in initial value form. That is, an ACE model specifies how the state of a modeled system changes over time, starting from a given system state at an initial time. However, the critical modeling principle (MP4), which requires agent autonomy conditional on initial state conditions, differentiates ACE models from standard state-space models within economics. The latter models typically incorporate modeler-imposed equilibrium conditions that are not locally constructive; that is, these conditions could not (or would not) be met by agents acting purely on the basis of their own states.
For ACE researchers, as for economists in general, the modeling of decision-making agents is a primary concern. Here is it important to correct a major misconception still being expressed by some commentators uninformed about the powerful capabilities of modern software: namely, the misconception that ACE decision-making agents cannot be as rational (or irrational) as real people.
To the contrary, the constraints on agent decision-making implied by modeling principles (MP1)-(MP7) are constraints inherent in every real-world dynamic system. As demonstrated concretely in
the decision methods used by ACE agents can range from simple behavioral rules to sophisticated adaptive dynamic programming algorithms for the approximate achievement of intertemporal objectives. Extensive annotated pointers to reference materials on the implementation of decision methods for ACE agents can be accessed at the following site:
ACE Research Area: Learning and the Embodied Mind.
A second misconception that needs countering is the incorrect belief that modeling principles (MP1)-(MP7) rule out any consideration of stochastic processes. To the contrary, stochastic processes can easily be represented within ACE models by means of pseudo-random number generators (PRNGs) typically consisting of a few lines of source code.
These PRNGs can be included among the methods of decision-making agents, permitting them to "randomize" their behaviors and actions. For example, decision-making agents can use PRNGs to
choose among equally preferred actions or action delays, to construct mixed strategies in game situations to avoid exploitable predictability, and to
induce perturbations in current action routines in order to explore new action possibilities. PRNGs can also be included among the methods of other types of agents, e.g., biological or physical agents such as crops and weather, in order to model stochastic processes external to decision-making agents.
However, a key point with regard to stochasticity is that the modeling principles (MP1)-(MP7) require ACE models to be dynamically complete virtual worlds. Consequently, ACE modelers must explicitly identify the source of any stochastic shocks appearing within their models, not simply their impact points, because all such shocks must be generated by agents actually residing within their modeled worlds. This requirement encourages ACE modelers to think carefully about the intended empirical referents for any included stochastic shock terms.
Current ACE research divides roughly into four strands differentiated by objective.
One primary objective is empirical understanding: Why have particular observed regularities evolved and persisted? Examples of such regularities include social norms, socially accepted monies, market protocols, business cycles, persistent wealth inequality, and the common adoption of technological innovations. ACE researchers seek possible explanations grounded in the repeated interactions of agents operating in realistically rendered virtual worlds. Specifically, they try to understand whether particular types of observed regularities can be reliably generated within these worlds.
A second primary objective is normative design:
How can ACE models be used as computational laboratories to facilitate the design of institutional and regulatory policies resulting in socially desirable system performance over time? The ACE approach to normative design is akin to filling a bucket with water to determine if it leaks. A virtual world is constructed that captures the salient aspects of a system operating under a design of interest. The virtual world is then permitted to develop over time, driven solely by its own internal dynamics. One key issue is the extent to which resulting world outcomes are efficient, fair, and orderly, despite possible attempts by agents to game the design for personal advantage. A second key issue is a cautionary concern for adverse unintended consequences.
A third primary objective is qualitative insight and theory generation: How can ACE models be used to gain a better understanding of dynamic economic systems through a better understanding of their full range of potential behaviors over time? Such understanding requires consideration of the complete phase portraits for dynamic economic systems, including not only the possible existence of equilibria but also the basins of attraction for any such equilibria. A better understanding of complete phase portraits for real-world dynamic economic systems would help to clarify not only why certain types of regularities have evolved and persisted but also why others have not.
A quintessential example of this line of research is an old but still unresolved concern of economists such as Adam Smith (1723-1790), Ludwig von Mises (1881-1973), John Maynard Keynes (1883-1946), Joseph Schumpeter (1883-1950), and Friedrich von Hayek (1899-1992): namely, what are the self-organizing capabilities of decentralized market economies?
A fourth primary objective is methodological advancement: How best to provide ACE researchers with the methods and tools they need to undertake theoretical studies of dynamic economic systems through systematic sensitivity studies, and to examine the compatibility of sensitivity-generated theories with real-world data? ACE researchers are exploring a variety of ways to address this objective ranging from careful consideration of methodological principles to the practical development of programming, visualization, and empirical validation tools.
Linked below are materials of possible interest to ACE researchers as well as to researchers who wish to explore the potential usefulness of agent-based modeling for social science purposes more generally. These materials are updated on a regular basis, and suggestions for additional materials to include are welcome.
As time permits, ACE news notes are posted both to the Social Simulation (SimSoc) mailing list and to the Society for Computational Economics (SCE) mailing list to let people know which ACE website pages have been most heavily updated since the last news notes posting. If you would like to subscribe to either of these mailing lists, please visit
Please contact me at
tesfatsi AT iastate.edu
if you have ACE-related news items that you would like included at the ACE website. To keep website maintenance manageable, only items of a more persistent nature (e.g., journal articles) can be considered for the ACE website. However, you can post items of a more temporary nature (e.g., conference announcements) at the SimSoc and SCE mailing lists.
Materials Linked to Date
"Economic Systems as Constructively Rational Games: Oh, the Places We Could Go!"
On-line guide for newcomers to agent-based modeling
in the social sciences,
with links to demonstration
software (R. Axelrod and L. Tesfatsion)
Annotated pointers to ACE introductory materials
"ACE: A Constructive Approach to Economic Theory"
ACE Handbook (L. Tesfatsion and K. L. Judd), Elsevier/North Holland, 2006