Agent-based computational economics (ACE) is the area ofcomputational economics that studies economic processes, including wholeeconomies, asdynamic systems of interactingagents. As such, it falls in theparadigm ofcomplex adaptive systems.^[1] In correspondingagent-based models, the "agents" are "computational objects modeled as interacting according to rules" over space and time, not real people. The rules are formulated to model behavior and social interactions based on incentives and information.^[2] Such rules could also be the result of optimization, realized through use of AI methods (such asQ-learning and other reinforcement learning techniques).^[3]

As part ofnon-equilibrium economics,^[4] the theoretical assumption ofmathematical optimization by agents inequilibrium is replaced by the less restrictive postulate of agents withbounded rationalityadapting to market forces.^[5] ACE models applynumerical methods of analysis tocomputer-based simulations of complex dynamic problems for which more conventional methods, such as theorem formulation, may not find ready use.^[6] Starting from initial conditions specified by the modeler, the computational economy evolves over time as its constituent agents repeatedly interact with each other, including learning from interactions. In these respects, ACE has been characterized as a bottom-up culture-dish approach to the study ofeconomic systems.^[7]

ACE has a similarity to, and overlap with,game theory as an agent-based method for modeling social interactions.^[8] But practitioners have also noted differences from standard methods, for example in ACE events modeled being driven solely by initial conditions, whether or not equilibria exist or are computationally tractable, and in the modeling facilitation of agent autonomy and learning.^[9]

The method has benefited from continuing improvements in modeling techniques ofcomputer science and increased computer capabilities. The ultimate scientific objective of the method is to "test theoretical findings against real-world data in ways that permit empirically supported theories to cumulate over time, with each researcher's work building appropriately on the work that has gone before."^[10] The subject has been applied to research areas likeasset pricing,^[11]energy systems,^[12]competition andcollaboration,^[13]transaction costs,^[14]market structure andindustrial organization and dynamics,^[15]welfare economics,^[16] andmechanism design,^[17]information and uncertainty,^[18]macroeconomics,^[19] andMarxist economics.^[20][21]

Recent integrations of reinforcement learning and deep learning architectures have enabled simulation of AI-driven agents in complex multi-agent economic models, enhancing realism and emergent behaviour forecasting.^[22]

The "agents" in ACE models can represent individuals (e.g. people), social groupings (e.g. firms), biological entities (e.g. growing crops), and/or physical systems (e.g. transport systems). The ACE modeler provides the initial configuration of a computational economic system comprising multiple interacting agents. The modeler then steps back to observe the development of the system over time without further intervention. In particular, system events should be driven by agent interactions without external imposition of equilibrium conditions.^[23] Issues include those common toexperimental economics in general^[24] and development of a common framework for empirical validation^[25] and resolving open questions in agent-based modeling.^[26]

ACE is an officially designated special interest group (SIG) of the Society for Computational Economics.^[27] Researchers at theSanta Fe Institute have contributed to the development of ACE.

One area where ACE methodology has frequently been applied is asset pricing.W. Brian Arthur, Eric Baum,William Brock, Cars Hommes, and Blake LeBaron, among others, have developed computational models in which many agents choose from a set of possible forecasting strategies in order to predict stock prices, which affects their asset demands and thus affects stock prices. These models assume that agents are more likely to choose forecasting strategies which have recently been successful. The success of any strategy will depend on market conditions and also on the set of strategies that are currently being used. These models frequently find that large booms and busts in asset prices may occur as agents switch across forecasting strategies.^[11][28][29] More recently, Brock, Hommes, and Wagener (2009) have used a model of this type to argue that the introduction of new hedging instruments may destabilize the market,^[30] and some papers have suggested that ACE might be a useful methodology for understanding the 2008financial crisis.^[31][32][33]See also discussion underFinancial economics § Financial markets and§ Departures from rationality.

Agent-Based Macroeconomics (ABM) builds understanding of the economy from the very bottom up. Instead of broad averages, ABM explicitly simulates many diverse, heterogeneous agents. Each agent operates with specific rules for behavior and learning. They also form expectations in unique ways. This "bottom-up" philosophy traces back to Herbert Simon's early work. Economic events emerge organically from the agents' interactions. They are not simply caused by external shocks. ABM incorporates empirically-informed behavioral rules. It also represents persistent heterogeneity among agents. Explicit institutional structures and market frictions are included too.^[34]

• Agent-based social simulation
• Artificial financial market
• Computational economics
• Econophysics
• Macroeconomic model
• Multi-agent system
• Statistical finance

• Computational economics
• Monte Carlo methods in finance
• Computational fields of study
• Agent-based model

• Webarchive template wayback links
• All articles with dead external links
• Articles with dead external links from September 2025
• Articles with short description
• Short description matches Wikidata
• Use dmy dates from February 2020
• Articles with excerpts