Contenido

La idea del modelado basado en agentes se desarrolló como un concepto relativamente sencillo a finales de la década de 1940. No se generalizó hasta la década de 1990, ya que requiere procedimientos de cálculo intensivos.

First developments

The history of agent-based modeling dates back to the Von Neumann machine, a theoretical machine capable of reproducing itself. The device that von Neumann proposed followed precise and detailed instructions to make a copy of itself. The concept was developed by Stanislaw Ulam, a friend of von Neumann and also a mathematician, who suggested building the machine on paper, as a collection of cells on a grid. The idea intrigued von Neumann, who put it into practice by creating the first of the devices that would later be called cellular automata. Another advance was introduced by the mathematician John Conway. He built the well-known Game of Life. Unlike von Neumann's machine, Conway's Game of Life operated by simple rules in a two-dimensional virtual world similar to a checkerboard.

The Simula programming language, developed in the mid-1960s and widely implemented in the early 1970s, was the first framework for automating step-by-step agent simulations.

1970s and 1980s: the first models

One of the first agent-based models was Thomas Schelling's segregation model,[11]​ which was discussed in his 1971 article "Dynamic Models of Segregation." Although Schelling originally used coins and graph paper instead of computers, his models embodied the basic concept of agent-based models as autonomous entities interacting in a shared environment with an observed aggregate and emergent outcome.

In the late 1970s, Paulien Hogeweg and Bruce Hesper began experimenting with individual models of ecology. One of their first results was to demonstrate that the social structure of bumblebee colonies arose as a result of simple rules governing the behavior of individual bees.[12] They introduced the ToDo principle, which refers to the way in which agents "do what needs to be done" at a given time.

In the early 1980s, Robert Axelrod&action=edit&redlink=1 "Robert Axelrod (political scientist) (not yet drafted)") organized a tournament of prisoner's dilemma strategies and had them interact in an agent-based manner to determine a winner. Axelrod developed many other agent-based models in the field of political science to examine phenomena from ethnocentrism to the diffusion of culture. In the late 1980s, Craig Reynolds' work on flock models contributed to the development of some of the first agent-based biological models with social characteristics. He attempted to model the reality of living biological agents, known as artificial life, a term coined by Christopher Langton.

The first use of the word "agent" and its current meaning are difficult to trace. One candidate is John Holland and John H. Miller's 1991 article "Artificial Adaptive Agents in Economic Theory",[14] based on an earlier conference presentation of theirs. A stronger and earlier option is Allan Newell, who in the first Presidential Address of the AAAI (published as The Knowledge Level")[15]​) spoke of intelligent agents as a concept.

At the same time, during the 1980s, social scientists, mathematicians, operations researchers, and others from various disciplines developed Computational and Mathematical Organization Theory (CMOT). This field emerged as a special interest group of The Institute of Management Sciences (TIMS) and its sister society, the Operations Research Society of America (ORSA).[16]​.

1990s: Expansion

The 1990s were especially notable for the expansion of MBAs within the social sciences. A prominent example was the large-scale Sugarscape MBA, developed by Joshua M. Epstein and Robert Axtell to simulate and explore the role of social phenomena such as seasonal migrations, pollution, sexual reproduction, disease control and transmission, and even culture. Other notable developments of the 1990s include the Kathleen Carley MBA at Carnegie Mellon University, aimed at exploring the co-evolution of social networks and culture. The Santa Fe Institute (SFI) was instrumental in fostering the development of the MBA Swarm modeling platform under the leadership of Christopher Langton. The research carried out through the SFI allowed the expansion of MBA techniques to a number of fields, including the study of the social and spatial dynamics of small-scale human societies and primates. Simulation")* (JASSS). In addition to JASSS, the SpringerOpen magazine Complex Adaptive Systems Modeling") (CASM)[19] covers agent-based models from any discipline.

In the mid-1990s, the MBA's social science focus began to focus on issues such as team effectiveness, the communication necessary for organizational effectiveness, and social media behavior. CMOT—later renamed Computational Analysis of Social and Organizational Systems (CASOS)—increasingly incorporated agent-based modeling. Samuelson (2000) provides an overview of early history,[20]​ and Samuelson (2005) and Samuelson and Macal (2006) discuss more recent developments.[21][22]​.

In the late 1990s, the merger of TIMS and ORSA to form INFORMS"), and the move of INFORMS from two to one annual meeting, helped prompt the CMOT group to form a separate society, the North American Association for Computational Social and Organizational Sciences (NAACSOS). Kathleen Carley was a major contributor, especially regarding social network models, and obtained funding from the National Science Foundation") for the annual conference in addition to serving as the first President of NAACSOS. He was succeeded by David Sallach of the University of Chicago and Argonne National Laboratory, and then Michael Prietula of Emory University. Around the same time that NAACSOS was created, the European Social Simulation Association (ESSA) and the Pacific Asian Association for Agent-Based Approach in Social Systems Science (PAAA), counterpart associations of NAACSOS, were organized. In 2013, these three organizations collaborate internationally. The First World Congress on Social Simulation was held under their joint sponsorship in Kyoto in August 2006.[23] and the Second World Congress was held outside Washington, D.C., in July 2008, with George Mason University taking the lead role in the local organization.

2000s

More recently, Ron Sun developed methods for basing agent-based simulation on models of human cognition, known as social cognitive simulation. Bill McKelvey, Suzanne Lohmann, Dario Nardi, Dwight Read, and others at UCLA have also made significant contributions to organizational behavior and decision making. Since 1991, UCLA has hosted a symposium in Lake Arrowhead, California, which has become another important meeting point for organizational behavior and decision-making professionals. field.[25]​.

2020 and later

After the advent of large language models, researchers began to apply interactive language models to agent-based modeling. In a widely cited article, agentive language models were used in a simulation environment to perform activities such as planning birthday parties and holding elections.[26]​.

La mayor parte de la investigación sobre modelos computacionales describe los sistemas en estado estacionario o como movimiento entre estados de equilibrio. Sin embargo, el modelado basado en agentes, utilizando reglas simples, puede dar lugar a diferentes tipos de comportamiento complejo e interesante. Las tres ideas centrales de los modelos basados en agentes son los agentes como objetos, la emergencia y la complejidad.

Los modelos basados en agentes constan de agentes dinámicos basados en reglas que interactúan entre sí. Los sistemas en que interactúan pueden crear una complejidad similar a la del mundo real. Normalmente los agentes están situados en el espacio y el tiempo y residen en redes o en barrios reticulares. La ubicación de los agentes y su comportamiento de respuesta están codificados de forma algorítmica en programas informáticos. En algunos casos, aunque no siempre, los agentes pueden considerarse inteligentes y con propósito. En los MBA ecológicos (a menudo denominados "modelos basados en individuos" en ecología), los agentes pueden ser, por ejemplo, árboles en un bosque, y no se les consideraría inteligentes, aunque sí "con propósito" en el sentido de optimizar el acceso a un recurso (como el agua). El proceso de modelado es más bien inductivo. El modelizador hace los supuestos que considera más relevantes para la situación dada y luego observa los fenómenos que surgen de las interacciones de los agentes. En ocasiones, el resultado es un equilibrio. Otras veces es un patrón emergente. Sin embargo, a veces es una maraña ininteligible.

En cierto modo, los modelos basados en agentes complementan los métodos analíticos tradicionales. Mientras que los métodos analíticos permiten a los humanos caracterizar los equilibrios de un sistema, los modelos basados en agentes permiten generar esos equilibrios. Esta contribución generativa es el beneficio potencial más común del modelado basado en agentes. Los modelos basados en agentes pueden explicar la emergencia de patrones de orden superior: estructuras de red de organizaciones terroristas e Internet, distribuciones de la ley de potencia en los tamaños de los atascos de tráfico, guerras y caídas del mercado de valores, y segregación social que persiste a pesar de las poblaciones tolerantes. Los modelos basados en agentes también pueden utilizarse para identificar puntos de apalancamiento, definidos como momentos en el tiempo en los que las intervenciones tienen consecuencias extremas, y para distinguir entre tipos de dependencia de la trayectoria.

En lugar de centrarse en estados estables, muchos modelos consideran la robustez de un sistema: las formas en que los sistemas complejos se adaptan a las presiones internas y externas para mantener sus funcionalidades. La tarea de aprovechar esa complejidad requiere tener en cuenta a los propios agentes: su diversidad, conectividad y nivel de interacciones.

Framework

Recent work on modeling and simulation of complex adaptive systems has demonstrated the need to combine agent-based and complex network models.[27]​[28]​[29]​ describe a four-tier framework for developing models of complex adaptive systems described through several multidisciplinary case studies.

The second is the complex network modeling level, in which models are developed using interaction data from various system components. Agent-Based Exploratory Modeling Level to develop agent-based models and evaluate the feasibility of future research. This can be useful, for example, for developing proof-of-concept models or applying for funding without requiring a long learning curve for researchers. Descriptive Agent-Based Modeling (DREAM) to develop agent-based model descriptions using templates and models based on complex networks. The construction of DREAM models allows models to be compared between scientific disciplines. Validated agent-based modeling using the virtual overlay multi-agent system (VOMAS) for the development of formally verified and validated models. Other methods for describing agent-based models include code templates[30]​ and text-based methods such as the ODD (Overview, Design Concepts, and Design Details) protocol[31]​.

The role of the environment where agents live, both macro and micro,[32]​ is also becoming an important factor in agent-based modeling and simulation work. A simple environment allows simple agents, but a complex environment generates diversity of behavior.[33]​.

Multi-scale modeling

One of the main advantages of agent-based modeling is its ability to mediate the flow of information between scales. When more details about an agent are needed, a researcher can integrate them into models that describe them. If you are interested in the emergent behaviors demonstrated by the population of agents, you can combine the agent model with a continuous model that describes the dynamics of the population. For example, in a study on CD4+ T cells (a key cell type in the adaptive immune system), [34]​ researchers modeled biological phenomena that occur at different spatial (intracellular, cellular and systemic), temporal and organizational scales (signal transduction, gene regulation, metabolism, cellular behaviors and cytokine transport). In the resulting modular model, signal transduction and gene regulation are described by a logic model, metabolism by constraint-based models, cell population dynamics are described by an agent-based model, and systemic cytokine concentrations by ordinary differential equations. In this multi-scale model, the agent-based model takes center stage and orchestrates each information flow across scales.

in biology

Agent-based modeling has been widely used in biology, including analysis of the spread of epidemics,[35]​ and the threat of biological warfare, biological applications") including population dynamics,[36]​ stochastic gene expression,[37]​ plant-animal interactions,[38]​ vegetation ecology,[39]​ migratory ecology,[40]​ landscape diversity,[41]​ sociobiology,[42]​ growth and decline of ancient civilizations, evolution of ethnocentric behavior,[43]​ forced displacement/migration,[44]​ dynamics of language choice,[45]​ cognitive modeling, and biomedical applications including 3D modeling of breast tissue formation/morphogenesis,[46]​ the effects of ionizing radiation on mammary stem cell subpopulation dynamics,[47]​ inflammation,[48]​

[49]​

and the human immune system"),[50]​ and the evolution of foraging behaviors.[51]​ Agent-based models have also been used to develop decision support systems such as for breast cancer.[52]​ Agent-based models are increasingly used to model drug systems in early phase and preclinical research to aid in drug development and gain insights into biological systems that would not be possible a priori.[53]​ Military applications have also been evaluated.[54]​ In addition, agent-based models have recently been used to study biological systems at the molecular level.[55]​[56][57]​ Agent-based models have also been written to describe ecological processes acting in ancient systems, such as those in dinosaur environments and other more recent ancient systems.[58]​[59][60]​.

in epidemiology

Agent-based models now complement traditional compartmental models, which are the most common type of epidemiological models. MBAs have been shown to be superior to compartmental models in terms of prediction accuracy.[61]​[62]​ Recently, MBAs such as CovidSim") by epidemiologist Neil Ferguson&action=edit&redlink=1 "Neil Ferguson (epidemiologist) (not yet written)"), have been used to inform public health (non-pharmacological) interventions against the spread of SARS-CoV-2.[63]​ Epidemiological MBAs have been criticized for simplification. and use unrealistic assumptions.[64]​[65]​ Still, they can be useful in informing decisions regarding mitigation and suppression measures in cases where MBAs are accurately calibrated.[66]​ MBAs for such simulations are primarily based on synthetic populations"), as actual population data are not always available.[67]​.

In business, technology and network theory

Agent-based models have been used since the mid-1990s to solve various business and technology problems. Some examples of applications are marketing,[72]​ organizational behavior and cognition,[73]​ teamwork,[74][75]​ supply chain optimization") and logistics, modeling consumer behavior, including word of mouth, the effects of social networks, distributed computing, workforce management") and portfolio management").[76]​.

Recently, agent-based modeling and simulation have been applied to various areas, such as studying the impact of researchers' publication forums in the field of computing (journals versus conferences).[77]​ In addition, MBAs have been used to simulate information delivery in environmental care environments.[78]​ A November 2016 article on arXiv analyzed an agent-based simulation of the diffusion of posts on Facebook.[79]​ In the networking domain peer-to-peer, ad hoc, and other complex, self-organizing networks, the utility of agent-based modeling and simulation has been demonstrated.[80] Recently, the use of a formal computer-based specification framework has been demonstrated, along with wireless sensor networks and agent-based simulation.[81]

Agent-based search or evolutionary algorithm is a new research topic to solve complex optimization problems.[82]​.

In team science

In the field of team science, agent-based modeling has been used to evaluate the effects of team member characteristics and biases on team performance in various settings.[83]​ By simulating interactions between agents—each representing individual team members with distinct traits and biases—this modeling approach allows researchers to explore how these factors collectively influence team performance dynamics and outcomes. Consequently, agent-based modeling provides a nuanced understanding of team science, facilitating a deeper exploration of the subtleties and variabilities inherent to team-based collaborations.

In economics and social sciences

Before and after the 2008 financial crisis, interest in MBAs as potential tools for economic analysis has grown.[84][85] MBAs do not assume that the economy can reach equilibrium, and "representative agents" are replaced by agents with diverse, dynamic, and interdependent behavior, including herding behavior. MBAs take a “bottom-up approach” and can generate extremely complex and volatile simulated economies. MBAs can represent unstable systems with busts and booms that develop from non-linear (disproportionate) responses to proportionally small changes.[86] A July 2010 article in The Economist examined MBAs as alternatives to EGDE models). and other economic complexities than standard models[87] along with an essay by J. Doyne Farmer") and Duncan Foley that argued that MBAs could satisfy both Keynes's desire to represent a complex economy and Robert Lucas's desire to build models based on microfoundations. Farmer and Foley noted the progress that has been made using MBAs to model parts of an economy, but argued for the creation of a very large model that incorporates low-level models. level.[89]​ Simulating a complex system of financial analysts") based on three different behavioral profiles – imitator, anti-imitator and indifferent – financial markets were simulated with great precision. The results showed a correlation between the morphology of the network and the stock index.[90]​ However, the MBA approach has been criticized for its lack of robustness between models, where similar models can give very different results.[91]​[92]​.

MBAs have been deployed in architecture and urban planning to evaluate the design and simulate the flow of pedestrians in the urban environment[93]​ and the examination of public policy applications to land use.[94]​ There is also a growing field of socioeconomic analysis of the impact of infrastructure investment using the ability of MBAs to discern systemic impacts across a socioeconomic network.[95]​ Heterogeneity and dynamics can be easily integrated into MBA models to address wealth inequality and social mobility.[96]​.

MBAs have also been proposed as applied educational tools for diplomats in the field of international relations[97]​ and for national and international policymakers to improve their evaluation of public policies.[98]​.

In water management

MBAs have also been applied in water resources planning and management, particularly to explore, simulate and predict the performance of infrastructure design and policy decisions,[99]​ and to evaluate the value of cooperation and information sharing in large water resources systems.[100]​.

Organizational MBA: Agent-led simulation

The agent-directed simulation (ADS) metaphor distinguishes between two categories, namely, "Systems for agents" and "Agents for systems".[101] Agent systems (sometimes called agent systems) are systems that implement agents for use in engineering, human dynamics and social dynamics, military and other applications. Agent systems for systems are divided into two subcategories. Agent-supported systems deal with the use of agents as a support facility to enable computer assistance in problem solving or improving cognitive abilities. Agent-based systems focus on the use of agents for the generation of model behavior in a system evaluation (system studies and analysis).

autonomous cars

Hallerbach et al. analyzed the application of agent-based approaches to the development and validation of automated driving systems through a digital twin of the tested vehicle and microscopic traffic simulation based on independent agents.[102]​ Waymo has created a Carcraft multi-agent simulation environment to test algorithms for autonomous cars.[103]​[104]​ It simulates traffic interactions between human drivers, pedestrians, and automated vehicles. People's behavior is imitated by artificial agents based on data from real human behavior. The basic idea of ​​using agent-based modeling to understand self-driving cars was discussed as early as 2003.[105]​.

Muchos marcos de MBA") están diseñados para arquitecturas informáticas serie von Neumann, lo que limita la velocidad y la escalabilidad de los modelos implementados. Dado que el comportamiento emergente en los MBA a gran escala depende del tamaño de la población,[106]​ las restricciones de escalabilidad pueden dificultar la validación del modelo.[107]​ Estas limitaciones se han abordado principalmente mediante la computación distribuida, con marcos de trabajo como Repast HPC[108]​ dedicados específicamente a este tipo de implementaciones. Si bien estos enfoques se adaptan bien a las arquitecturas de clústeres") y superordenadores, las cuestiones relacionadas con la comunicación y la sincronización,[109]​[110]​ así como la complejidad del despliegue,[111]​ siguen siendo posibles obstáculos para su adopción generalizada.

Un desarrollo reciente es el uso de algoritmos paralelos de datos en unidades de procesamiento gráfico GPU para la simulación de MBA.[106]​[112]​[113]​ El extremo ancho de banda de memoria combinado con la enorme potencia de cálculo de las GPU de multiprocesadores ha permitido la simulación de millones de agentes a decenas de fotogramas por segundo.

Integration with other forms of modeling

Since Agent-Based Modeling is more of a modeling framework than a particular piece of software or platform, it has often been used in conjunction with other forms of modeling. For example, agent-based models have also been combined with Geographic Information Systems (GIS). This provides a useful combination in which the MBA serves as a process model and the GIS system can provide a pattern model.[114] Similarly, Social Network Analysis (SNA) tools and agent-based models are sometimes integrated, where the MBA is used to simulate the dynamics in the network, while the SNA tool models and analyzes the network of interactions.[115] Tools such as GAMA") provide a natural way to integrate system dynamics and GIS with the MBA.

Verification and validation") (V&V) of simulation models is extremely important.[116]​ Verification consists of ensuring that the implemented model matches the conceptual model, while validation ensures that the implemented model bears some relation to the real world. Facial validation, sensitivity analysis, calibration and statistical validation are different aspects of validation.[117]​ A discrete event simulation framework") approach has been proposed for the validation of agent-based systems.[118]​ Herein You can find a comprehensive resource on empirical validation of agent-based models[119].

As an example of a V&V technique, consider VOMAS (virtual overlay multi-agent system),[120]​ a software engineering-based approach, in which a virtual overlay multi-agent system is developed alongside the agent-based model. Muazi et al. They also provide an example of using VOMAS for the verification and validation of a wildfire simulation model.[121]​[122]​ Another software engineering method, i.e. Test Driven Development, has been adapted for the validation of agent-based models.[123]​ This approach has another advantage that allows for automatic validation using unit testing tools.

• - Complex system.

• - Complex adaptive system.

• - Emergency.

• - Evolutionary algorithm.

• - Flocks.

• - Multi-agent system.

• - Simulated reality.

• - Social complexity.

• - Social simulation.

• - Swarm behavior.

Articles/general information

• - Models based on social network agents, java applets.

• - Online guide for newcomers to agent-based modeling in the social sciences.

• - Introduction to agent-based modeling and simulation. Argonne National Laboratory, November 29, 2006.

• - Agent-based models in Ecology – Use of computer models as theoretical tools to analyze complex ecological systems.

• - Frequently asked questions about Web agent-based modeling for computational modeling in social and ecological sciences.

• - Multi-agent information systems – Article on the convergence of SOA, BPM and multi-agent technology in the field of business information systems. José Manuel Gómez Álvarez, Artificial Intelligence, Polytechnic University of Madrid – 2006.

• - Artificial Life Framework.

• - Article that provides methodology to transfer human behaviors from the real world to a simulation model in which the behaviors of the agents are represented.

• - Agent-Based Modeling Resources, an information center for modelers, methods and philosophy for agent-based modeling.

• - An agent-based model of the May 6, 2010 Flash Crash, with policy implications, Tommi A. Vuorenmaa (Valo Research and Trading), Liang Wang (University of Helsinki - Department of Computer Science), October 2013.

Simulation models

• - Multi-agent meeting scheduling system model by Qasim Siddique.

• - Valentino Piana multi-signature market simulation.

The main elements of an agent-based model are the following: [10]​.

Agents: They are the fundamental units of the model, which are defined by their internal and external properties, as well as their behaviors. Agents can exhibit different levels of complexity and their decision making can be based on utility rules, allowing them to “learn” and adapt to their environment.

Environment: Represents the space where agents interact. This environment can be classified as dynamic, if it changes over time, or static, if it remains constant. Furthermore, the environment can be considered as a set of interrelated agents that can also influence the behavior of individual agents.

Interactions: These are the relationships and actions that occur between agents, as well as between them and their environment. These interactions can be classified as direct actions, when agents explicitly influence other agents or the environment, or indirect influences, when an agent's actions affect other agents or the environment in a less direct way.

Observer: It is the component with which the user interacts, allowing them to manage and monitor the actions of both the agents and the environment. The observer can have a passive role (just observing) or an active role (modifying the conditions of the model).

Schedule (Temporal Scheduler): Defines the order of interactions within the model, establishing the temporal sequence of events and interactions between agents and the environment. In addition, you can specify the possible interactions of the observer with the system at certain moments in the time cycle.

Contenido

La idea del modelado basado en agentes se desarrolló como un concepto relativamente sencillo a finales de la década de 1940. No se generalizó hasta la década de 1990, ya que requiere procedimientos de cálculo intensivos.

First developments

The history of agent-based modeling dates back to the Von Neumann machine, a theoretical machine capable of reproducing itself. The device that von Neumann proposed followed precise and detailed instructions to make a copy of itself. The concept was developed by Stanislaw Ulam, a friend of von Neumann and also a mathematician, who suggested building the machine on paper, as a collection of cells on a grid. The idea intrigued von Neumann, who put it into practice by creating the first of the devices that would later be called cellular automata. Another advance was introduced by the mathematician John Conway. He built the well-known Game of Life. Unlike von Neumann's machine, Conway's Game of Life operated by simple rules in a two-dimensional virtual world similar to a checkerboard.

The Simula programming language, developed in the mid-1960s and widely implemented in the early 1970s, was the first framework for automating step-by-step agent simulations.

1970s and 1980s: the first models

One of the first agent-based models was Thomas Schelling's segregation model,[11]​ which was discussed in his 1971 article "Dynamic Models of Segregation." Although Schelling originally used coins and graph paper instead of computers, his models embodied the basic concept of agent-based models as autonomous entities interacting in a shared environment with an observed aggregate and emergent outcome.

In the late 1970s, Paulien Hogeweg and Bruce Hesper began experimenting with individual models of ecology. One of their first results was to demonstrate that the social structure of bumblebee colonies arose as a result of simple rules governing the behavior of individual bees.[12] They introduced the ToDo principle, which refers to the way in which agents "do what needs to be done" at a given time.

In the early 1980s, Robert Axelrod&action=edit&redlink=1 "Robert Axelrod (political scientist) (not yet drafted)") organized a tournament of prisoner's dilemma strategies and had them interact in an agent-based manner to determine a winner. Axelrod developed many other agent-based models in the field of political science to examine phenomena from ethnocentrism to the diffusion of culture. In the late 1980s, Craig Reynolds' work on flock models contributed to the development of some of the first agent-based biological models with social characteristics. He attempted to model the reality of living biological agents, known as artificial life, a term coined by Christopher Langton.

The first use of the word "agent" and its current meaning are difficult to trace. One candidate is John Holland and John H. Miller's 1991 article "Artificial Adaptive Agents in Economic Theory",[14] based on an earlier conference presentation of theirs. A stronger and earlier option is Allan Newell, who in the first Presidential Address of the AAAI (published as The Knowledge Level")[15]​) spoke of intelligent agents as a concept.

At the same time, during the 1980s, social scientists, mathematicians, operations researchers, and others from various disciplines developed Computational and Mathematical Organization Theory (CMOT). This field emerged as a special interest group of The Institute of Management Sciences (TIMS) and its sister society, the Operations Research Society of America (ORSA).[16]​.

1990s: Expansion

The 1990s were especially notable for the expansion of MBAs within the social sciences. A prominent example was the large-scale Sugarscape MBA, developed by Joshua M. Epstein and Robert Axtell to simulate and explore the role of social phenomena such as seasonal migrations, pollution, sexual reproduction, disease control and transmission, and even culture. Other notable developments of the 1990s include the Kathleen Carley MBA at Carnegie Mellon University, aimed at exploring the co-evolution of social networks and culture. The Santa Fe Institute (SFI) was instrumental in fostering the development of the MBA Swarm modeling platform under the leadership of Christopher Langton. The research carried out through the SFI allowed the expansion of MBA techniques to a number of fields, including the study of the social and spatial dynamics of small-scale human societies and primates. Simulation")* (JASSS). In addition to JASSS, the SpringerOpen magazine Complex Adaptive Systems Modeling") (CASM)[19] covers agent-based models from any discipline.

In the mid-1990s, the MBA's social science focus began to focus on issues such as team effectiveness, the communication necessary for organizational effectiveness, and social media behavior. CMOT—later renamed Computational Analysis of Social and Organizational Systems (CASOS)—increasingly incorporated agent-based modeling. Samuelson (2000) provides an overview of early history,[20]​ and Samuelson (2005) and Samuelson and Macal (2006) discuss more recent developments.[21][22]​.

In the late 1990s, the merger of TIMS and ORSA to form INFORMS"), and the move of INFORMS from two to one annual meeting, helped prompt the CMOT group to form a separate society, the North American Association for Computational Social and Organizational Sciences (NAACSOS). Kathleen Carley was a major contributor, especially regarding social network models, and obtained funding from the National Science Foundation") for the annual conference in addition to serving as the first President of NAACSOS. He was succeeded by David Sallach of the University of Chicago and Argonne National Laboratory, and then Michael Prietula of Emory University. Around the same time that NAACSOS was created, the European Social Simulation Association (ESSA) and the Pacific Asian Association for Agent-Based Approach in Social Systems Science (PAAA), counterpart associations of NAACSOS, were organized. In 2013, these three organizations collaborate internationally. The First World Congress on Social Simulation was held under their joint sponsorship in Kyoto in August 2006.[23] and the Second World Congress was held outside Washington, D.C., in July 2008, with George Mason University taking the lead role in the local organization.

2000s

More recently, Ron Sun developed methods for basing agent-based simulation on models of human cognition, known as social cognitive simulation. Bill McKelvey, Suzanne Lohmann, Dario Nardi, Dwight Read, and others at UCLA have also made significant contributions to organizational behavior and decision making. Since 1991, UCLA has hosted a symposium in Lake Arrowhead, California, which has become another important meeting point for organizational behavior and decision-making professionals. field.[25]​.

2020 and later

After the advent of large language models, researchers began to apply interactive language models to agent-based modeling. In a widely cited article, agentive language models were used in a simulation environment to perform activities such as planning birthday parties and holding elections.[26]​.

La mayor parte de la investigación sobre modelos computacionales describe los sistemas en estado estacionario o como movimiento entre estados de equilibrio. Sin embargo, el modelado basado en agentes, utilizando reglas simples, puede dar lugar a diferentes tipos de comportamiento complejo e interesante. Las tres ideas centrales de los modelos basados en agentes son los agentes como objetos, la emergencia y la complejidad.

Los modelos basados en agentes constan de agentes dinámicos basados en reglas que interactúan entre sí. Los sistemas en que interactúan pueden crear una complejidad similar a la del mundo real. Normalmente los agentes están situados en el espacio y el tiempo y residen en redes o en barrios reticulares. La ubicación de los agentes y su comportamiento de respuesta están codificados de forma algorítmica en programas informáticos. En algunos casos, aunque no siempre, los agentes pueden considerarse inteligentes y con propósito. En los MBA ecológicos (a menudo denominados "modelos basados en individuos" en ecología), los agentes pueden ser, por ejemplo, árboles en un bosque, y no se les consideraría inteligentes, aunque sí "con propósito" en el sentido de optimizar el acceso a un recurso (como el agua). El proceso de modelado es más bien inductivo. El modelizador hace los supuestos que considera más relevantes para la situación dada y luego observa los fenómenos que surgen de las interacciones de los agentes. En ocasiones, el resultado es un equilibrio. Otras veces es un patrón emergente. Sin embargo, a veces es una maraña ininteligible.

En cierto modo, los modelos basados en agentes complementan los métodos analíticos tradicionales. Mientras que los métodos analíticos permiten a los humanos caracterizar los equilibrios de un sistema, los modelos basados en agentes permiten generar esos equilibrios. Esta contribución generativa es el beneficio potencial más común del modelado basado en agentes. Los modelos basados en agentes pueden explicar la emergencia de patrones de orden superior: estructuras de red de organizaciones terroristas e Internet, distribuciones de la ley de potencia en los tamaños de los atascos de tráfico, guerras y caídas del mercado de valores, y segregación social que persiste a pesar de las poblaciones tolerantes. Los modelos basados en agentes también pueden utilizarse para identificar puntos de apalancamiento, definidos como momentos en el tiempo en los que las intervenciones tienen consecuencias extremas, y para distinguir entre tipos de dependencia de la trayectoria.

En lugar de centrarse en estados estables, muchos modelos consideran la robustez de un sistema: las formas en que los sistemas complejos se adaptan a las presiones internas y externas para mantener sus funcionalidades. La tarea de aprovechar esa complejidad requiere tener en cuenta a los propios agentes: su diversidad, conectividad y nivel de interacciones.

Framework

Recent work on modeling and simulation of complex adaptive systems has demonstrated the need to combine agent-based and complex network models.[27]​[28]​[29]​ describe a four-tier framework for developing models of complex adaptive systems described through several multidisciplinary case studies.

The second is the complex network modeling level, in which models are developed using interaction data from various system components. Agent-Based Exploratory Modeling Level to develop agent-based models and evaluate the feasibility of future research. This can be useful, for example, for developing proof-of-concept models or applying for funding without requiring a long learning curve for researchers. Descriptive Agent-Based Modeling (DREAM) to develop agent-based model descriptions using templates and models based on complex networks. The construction of DREAM models allows models to be compared between scientific disciplines. Validated agent-based modeling using the virtual overlay multi-agent system (VOMAS) for the development of formally verified and validated models. Other methods for describing agent-based models include code templates[30]​ and text-based methods such as the ODD (Overview, Design Concepts, and Design Details) protocol[31]​.

The role of the environment where agents live, both macro and micro,[32]​ is also becoming an important factor in agent-based modeling and simulation work. A simple environment allows simple agents, but a complex environment generates diversity of behavior.[33]​.

Multi-scale modeling

One of the main advantages of agent-based modeling is its ability to mediate the flow of information between scales. When more details about an agent are needed, a researcher can integrate them into models that describe them. If you are interested in the emergent behaviors demonstrated by the population of agents, you can combine the agent model with a continuous model that describes the dynamics of the population. For example, in a study on CD4+ T cells (a key cell type in the adaptive immune system), [34]​ researchers modeled biological phenomena that occur at different spatial (intracellular, cellular and systemic), temporal and organizational scales (signal transduction, gene regulation, metabolism, cellular behaviors and cytokine transport). In the resulting modular model, signal transduction and gene regulation are described by a logic model, metabolism by constraint-based models, cell population dynamics are described by an agent-based model, and systemic cytokine concentrations by ordinary differential equations. In this multi-scale model, the agent-based model takes center stage and orchestrates each information flow across scales.

in biology

Agent-based modeling has been widely used in biology, including analysis of the spread of epidemics,[35]​ and the threat of biological warfare, biological applications") including population dynamics,[36]​ stochastic gene expression,[37]​ plant-animal interactions,[38]​ vegetation ecology,[39]​ migratory ecology,[40]​ landscape diversity,[41]​ sociobiology,[42]​ growth and decline of ancient civilizations, evolution of ethnocentric behavior,[43]​ forced displacement/migration,[44]​ dynamics of language choice,[45]​ cognitive modeling, and biomedical applications including 3D modeling of breast tissue formation/morphogenesis,[46]​ the effects of ionizing radiation on mammary stem cell subpopulation dynamics,[47]​ inflammation,[48]​

[49]​

and the human immune system"),[50]​ and the evolution of foraging behaviors.[51]​ Agent-based models have also been used to develop decision support systems such as for breast cancer.[52]​ Agent-based models are increasingly used to model drug systems in early phase and preclinical research to aid in drug development and gain insights into biological systems that would not be possible a priori.[53]​ Military applications have also been evaluated.[54]​ In addition, agent-based models have recently been used to study biological systems at the molecular level.[55]​[56][57]​ Agent-based models have also been written to describe ecological processes acting in ancient systems, such as those in dinosaur environments and other more recent ancient systems.[58]​[59][60]​.

in epidemiology

Agent-based models now complement traditional compartmental models, which are the most common type of epidemiological models. MBAs have been shown to be superior to compartmental models in terms of prediction accuracy.[61]​[62]​ Recently, MBAs such as CovidSim") by epidemiologist Neil Ferguson&action=edit&redlink=1 "Neil Ferguson (epidemiologist) (not yet written)"), have been used to inform public health (non-pharmacological) interventions against the spread of SARS-CoV-2.[63]​ Epidemiological MBAs have been criticized for simplification. and use unrealistic assumptions.[64]​[65]​ Still, they can be useful in informing decisions regarding mitigation and suppression measures in cases where MBAs are accurately calibrated.[66]​ MBAs for such simulations are primarily based on synthetic populations"), as actual population data are not always available.[67]​.

In business, technology and network theory

Agent-based models have been used since the mid-1990s to solve various business and technology problems. Some examples of applications are marketing,[72]​ organizational behavior and cognition,[73]​ teamwork,[74][75]​ supply chain optimization") and logistics, modeling consumer behavior, including word of mouth, the effects of social networks, distributed computing, workforce management") and portfolio management").[76]​.

Recently, agent-based modeling and simulation have been applied to various areas, such as studying the impact of researchers' publication forums in the field of computing (journals versus conferences).[77]​ In addition, MBAs have been used to simulate information delivery in environmental care environments.[78]​ A November 2016 article on arXiv analyzed an agent-based simulation of the diffusion of posts on Facebook.[79]​ In the networking domain peer-to-peer, ad hoc, and other complex, self-organizing networks, the utility of agent-based modeling and simulation has been demonstrated.[80] Recently, the use of a formal computer-based specification framework has been demonstrated, along with wireless sensor networks and agent-based simulation.[81]

Agent-based search or evolutionary algorithm is a new research topic to solve complex optimization problems.[82]​.

In team science

In the field of team science, agent-based modeling has been used to evaluate the effects of team member characteristics and biases on team performance in various settings.[83]​ By simulating interactions between agents—each representing individual team members with distinct traits and biases—this modeling approach allows researchers to explore how these factors collectively influence team performance dynamics and outcomes. Consequently, agent-based modeling provides a nuanced understanding of team science, facilitating a deeper exploration of the subtleties and variabilities inherent to team-based collaborations.

In economics and social sciences

Before and after the 2008 financial crisis, interest in MBAs as potential tools for economic analysis has grown.[84][85] MBAs do not assume that the economy can reach equilibrium, and "representative agents" are replaced by agents with diverse, dynamic, and interdependent behavior, including herding behavior. MBAs take a “bottom-up approach” and can generate extremely complex and volatile simulated economies. MBAs can represent unstable systems with busts and booms that develop from non-linear (disproportionate) responses to proportionally small changes.[86] A July 2010 article in The Economist examined MBAs as alternatives to EGDE models). and other economic complexities than standard models[87] along with an essay by J. Doyne Farmer") and Duncan Foley that argued that MBAs could satisfy both Keynes's desire to represent a complex economy and Robert Lucas's desire to build models based on microfoundations. Farmer and Foley noted the progress that has been made using MBAs to model parts of an economy, but argued for the creation of a very large model that incorporates low-level models. level.[89]​ Simulating a complex system of financial analysts") based on three different behavioral profiles – imitator, anti-imitator and indifferent – financial markets were simulated with great precision. The results showed a correlation between the morphology of the network and the stock index.[90]​ However, the MBA approach has been criticized for its lack of robustness between models, where similar models can give very different results.[91]​[92]​.

MBAs have been deployed in architecture and urban planning to evaluate the design and simulate the flow of pedestrians in the urban environment[93]​ and the examination of public policy applications to land use.[94]​ There is also a growing field of socioeconomic analysis of the impact of infrastructure investment using the ability of MBAs to discern systemic impacts across a socioeconomic network.[95]​ Heterogeneity and dynamics can be easily integrated into MBA models to address wealth inequality and social mobility.[96]​.

MBAs have also been proposed as applied educational tools for diplomats in the field of international relations[97]​ and for national and international policymakers to improve their evaluation of public policies.[98]​.

In water management

MBAs have also been applied in water resources planning and management, particularly to explore, simulate and predict the performance of infrastructure design and policy decisions,[99]​ and to evaluate the value of cooperation and information sharing in large water resources systems.[100]​.

Organizational MBA: Agent-led simulation

The agent-directed simulation (ADS) metaphor distinguishes between two categories, namely, "Systems for agents" and "Agents for systems".[101] Agent systems (sometimes called agent systems) are systems that implement agents for use in engineering, human dynamics and social dynamics, military and other applications. Agent systems for systems are divided into two subcategories. Agent-supported systems deal with the use of agents as a support facility to enable computer assistance in problem solving or improving cognitive abilities. Agent-based systems focus on the use of agents for the generation of model behavior in a system evaluation (system studies and analysis).

autonomous cars

Hallerbach et al. analyzed the application of agent-based approaches to the development and validation of automated driving systems through a digital twin of the tested vehicle and microscopic traffic simulation based on independent agents.[102]​ Waymo has created a Carcraft multi-agent simulation environment to test algorithms for autonomous cars.[103]​[104]​ It simulates traffic interactions between human drivers, pedestrians, and automated vehicles. People's behavior is imitated by artificial agents based on data from real human behavior. The basic idea of ​​using agent-based modeling to understand self-driving cars was discussed as early as 2003.[105]​.

Muchos marcos de MBA") están diseñados para arquitecturas informáticas serie von Neumann, lo que limita la velocidad y la escalabilidad de los modelos implementados. Dado que el comportamiento emergente en los MBA a gran escala depende del tamaño de la población,[106]​ las restricciones de escalabilidad pueden dificultar la validación del modelo.[107]​ Estas limitaciones se han abordado principalmente mediante la computación distribuida, con marcos de trabajo como Repast HPC[108]​ dedicados específicamente a este tipo de implementaciones. Si bien estos enfoques se adaptan bien a las arquitecturas de clústeres") y superordenadores, las cuestiones relacionadas con la comunicación y la sincronización,[109]​[110]​ así como la complejidad del despliegue,[111]​ siguen siendo posibles obstáculos para su adopción generalizada.

Un desarrollo reciente es el uso de algoritmos paralelos de datos en unidades de procesamiento gráfico GPU para la simulación de MBA.[106]​[112]​[113]​ El extremo ancho de banda de memoria combinado con la enorme potencia de cálculo de las GPU de multiprocesadores ha permitido la simulación de millones de agentes a decenas de fotogramas por segundo.

Integration with other forms of modeling

Since Agent-Based Modeling is more of a modeling framework than a particular piece of software or platform, it has often been used in conjunction with other forms of modeling. For example, agent-based models have also been combined with Geographic Information Systems (GIS). This provides a useful combination in which the MBA serves as a process model and the GIS system can provide a pattern model.[114] Similarly, Social Network Analysis (SNA) tools and agent-based models are sometimes integrated, where the MBA is used to simulate the dynamics in the network, while the SNA tool models and analyzes the network of interactions.[115] Tools such as GAMA") provide a natural way to integrate system dynamics and GIS with the MBA.

Verification and validation") (V&V) of simulation models is extremely important.[116]​ Verification consists of ensuring that the implemented model matches the conceptual model, while validation ensures that the implemented model bears some relation to the real world. Facial validation, sensitivity analysis, calibration and statistical validation are different aspects of validation.[117]​ A discrete event simulation framework") approach has been proposed for the validation of agent-based systems.[118]​ Herein You can find a comprehensive resource on empirical validation of agent-based models[119].

As an example of a V&V technique, consider VOMAS (virtual overlay multi-agent system),[120]​ a software engineering-based approach, in which a virtual overlay multi-agent system is developed alongside the agent-based model. Muazi et al. They also provide an example of using VOMAS for the verification and validation of a wildfire simulation model.[121]​[122]​ Another software engineering method, i.e. Test Driven Development, has been adapted for the validation of agent-based models.[123]​ This approach has another advantage that allows for automatic validation using unit testing tools.

• - Complex system.

• - Complex adaptive system.

• - Emergency.

• - Evolutionary algorithm.

• - Flocks.

• - Multi-agent system.

• - Simulated reality.

• - Social complexity.

• - Social simulation.

• - Swarm behavior.

Articles/general information

• - Models based on social network agents, java applets.

• - Online guide for newcomers to agent-based modeling in the social sciences.

• - Introduction to agent-based modeling and simulation. Argonne National Laboratory, November 29, 2006.

• - Agent-based models in Ecology – Use of computer models as theoretical tools to analyze complex ecological systems.

• - Frequently asked questions about Web agent-based modeling for computational modeling in social and ecological sciences.

• - Multi-agent information systems – Article on the convergence of SOA, BPM and multi-agent technology in the field of business information systems. José Manuel Gómez Álvarez, Artificial Intelligence, Polytechnic University of Madrid – 2006.

• - Artificial Life Framework.

• - Article that provides methodology to transfer human behaviors from the real world to a simulation model in which the behaviors of the agents are represented.

• - Agent-Based Modeling Resources, an information center for modelers, methods and philosophy for agent-based modeling.

• - An agent-based model of the May 6, 2010 Flash Crash, with policy implications, Tommi A. Vuorenmaa (Valo Research and Trading), Liang Wang (University of Helsinki - Department of Computer Science), October 2013.

Simulation models

• - Multi-agent meeting scheduling system model by Qasim Siddique.

• - Valentino Piana multi-signature market simulation.