Agent-Based Models and Microsimulation

: This project presents an effort to establish theories and techniques of modeling, analyzing and controlling information dynamics and emergent behavior of heterogeneous-agent systems and demonstrate the application of these theories and techniques to heterogeneous-agent systems making up information supply networks. A heterogeneous-agent system is a Complex Adaptive System (CAS) involving a large collection of semi-autonomous agents. According to Complexity Theory for CAS, the aggregate system behavior of heterogeneous agents and their inter-connections emerges from the evolving local interactions of agents in a dynamically changing environment. A CAS best organizes from the bottom up through self-organization. We base the modeling of heterogeneous-agent systems on the innovative integration of three modeling paradigms: agent-based modeling, control-theoretic modeling, and stochastic discrete-event modeling. The model represents the physical, information and knowledge elements of each agent, which includes: fitness function, tagging, building blocks, internal model, dynamic resources and processes under the control of the agent, as well as the interaction of the agent with other agents and the environment. The analysis of heterogeneous-agent systems and their emergent behavior is based on the innovative application of non-linear time-series analysis techniques from chaos theory and nonlinear dynamical systems theory as well as multivariate statistical techniques to detect emergent states and temporal patterns. We base the control of heterogeneous-agent systems on the innovative bottom-up self-organization approach to coping with desirable and undesirable emergent states of heterogeneous-agent systems for system stability and robustness.

Agent-based models (ABM) and their simulations have been used to study complex systems with interacting entities and to model multi-agent systems. Simulations are used to explore the dynamic consequences of these models. In many cases, the behaviours that are of interest are emergent ones that arise as a result of interactions between agents rather than the actions of any individual agent. In this paper, we propose a formalism for describing emergent behaviours at any level of abstraction based on the idea that event types can be defined that characterise sets of behavioural 'motifs'. This provides the basis for a method for studying the associations between multi-level behaviours in simulations. There are two categories of hypotheses that we seek to address with respect to an ABM and its simulations: - Hypotheses concerned with associations between emergent behaviours defined at various levels of abstraction. - Hypotheses concerned with the links between parameter sensitivity / initial conditions and emergent behaviours e.g. the ABM is sensitive to a parameter x because x predisposes the system or part of the system to exhibit a particular (emergent) behaviour.

Agent-Based Modeling has been recently recognized as a method for in-silico multi-scale modeling of biological cell systems. Agent-Based Models (ABMs) allow results from experimental studies of individual cell behaviors to be scaled into the macro-behavior of interacting cells in complex cell systems or tissues. Current generation ABM simulation toolkits are designed to work on serial von-Neumann architectures, which have poor scalability. The best systems can barely handle tens of thousands of agents in real-time. Considering that there are models for which mega-scale populations have significantly different emergent behaviors than smaller population sizes, it is important to have the ability to model such large scale models in real-time. In this paper we present a new framework for simulating ABMs on programmable graphics processing units (GPUs). Novel algorithms and data-structures have been developed for agent-state representation, agent motion, and replication. As a test case, we have implemented an abstracted version of the Systematic Inflammatory Response System (SIRS) ABM. Compared to the original implementation on the NetLogo system, our implementation can handle an agent population that is over three orders of magnitude larger with close to 40 updates/sec. We believe that our system is the only one of its kind that is capable of efficiently handling realistic problem sizes in biological simulations.

This chapter highlights the importance of studying sustainability management and social responsibility issues from a systems perspective and argues that agent-based computational modelling is a suitable methodology for studying sustainability issues from an interdisciplinary, multilevel systems viewpoint. The chapter presents relevant computational-oriented works in the field of sustainability management, organized according to key research themes derived from the extant literature. The chapter also illustrates representative computational modelling works that focus on studying the adoption of corporate social responsibility practices. It argues that agent-based modelling can be used to capture not only the heterogeneous incentives of stakeholders, as more traditional methodologies have done, but also the intricate relations among such incentives that, additionally, may evolve over time. The chapter concludes by arguing that agent-based modelling can be fully leveraged if it takes into account considerations of intraorganizational dynamics, joint effects of first-order and second-order behavioural change, multilevel aggregation, systems resilience, interdependencies and temporal dynamics of stakeholders’ incentives, organizational structure, and complex forms of the firm–environment relations.

Agent-based modeling and simulation have evolved as a powerful tool for modeling complex systems, offering insights into emergent behaviors and interactions among diverse agents. Recently, integrating large language models into agent-based modeling and simulation presents a promising avenue for enhancing simulation capabilities. This paper surveys the landscape of utilizing large language models in agent-based modeling and simulation, discussing their challenges and promising future directions. In this survey, since this is an interdisciplinary field, we first introduce the background of agent-based modeling and simulation and large language model-empowered agents. We then discuss the motivation for applying large language models to agent-based simulation and systematically analyze the challenges in environment perception, human alignment, action generation, and evaluation. Most importantly, we provide a comprehensive overview of the recent works of large language model-empowered agent-based modeling and simulation in multiple scenarios, which can be divided into four domains: cyber, physical, social, and hybrid, covering simulation of both real-world and virtual environments, and how these works address the above challenges. Finally, since this area is new and quickly evolving, we discuss the open problems and promising future directions. We summarize the representative papers along with their code repositories in https://github.com/tsinghua-fib-lab/LLM-Agent-Based-Modeling-and-Simulation.

Chapter 6 - Agent-Based Modeling and High Performance Computing

It is plain that the Austrian revival that began in the 1970s has yet to succeed in convincing the mainstream of the academy to jettison their physics-based mathematical models in favor of the sort of models and forms of argumentation that contemporary Austrians advocate. Agent-based computational modeling is still in its relative infancy but is beginning to gain recognition among economists disenchanted with the neoclassical paradigm. The purpose of this paper is to assuage concerns that readers might have regarding methodological consistency between agent-based modeling and Austrian economics and to advocate its adoption as a means to convey Austrian ideas to a wider audience. I examine models developed and published by other researchers and ultimately provide an outline of how one might develop a research agenda that leverages this technique. I argue that agent-based modeling can be used to enhance Austrian theorizing and offers a viable alternative to the neoclassical paradigm.

The current economic crisis has provoked an active response from the interdisciplinary scientific community. As a result many papers suggesting what can be improved in understanding of the complex socio-economics systems were published. Some of the most prominent papers on the topic include (Bouchaud, 2009; Farmer and Foley, 2009; Farmer et al, 2012; Helbing, 2010; Pietronero, 2008). These papers share the idea that agent-based modeling is essential for the better understanding of the complex socio-economic systems and consequently better policy making. Yet in order for an agent-based model to be useful it should also be analytically tractable, possess a macroscopic treatment (Cristelli et al, 2012). In this work we shed a new light on our research group's contributions towards understanding of the correspondence between the inter-individual interactions and collective behavior. We also provide some new insights into the implications of the global and local interactions, the leadership and the predator-prey interactions in the complex socio-economic systems.

The basic concepts that form the basis of integrated modeling of the behavior of antagonistic agents in cybersecurity systems are identified. It is shown that the emphasis is largely on modeling the behavior of one of the cyber conflict parties only. In the case when the interaction of all parties to the conflict is considered, the approaches used are focused on solving particular problems, or they model a simplified situation. A methodology for modeling the interaction of antagonistic agents in cybersecurity systems, focused on the use of a multi-model complex with elements of cognitive modeling, is proposed. For this objective, the main components of cyber conflict are highlighted, the models of which must be developed. Modeling the interaction of antagonistic agents is proposed to be implemented as a simulation of situations. The concept of a situation is formulated and its components are presented. In the proposed methodology, traditional methods and modeling tools are not opposed, but are considered together, thus forming a unified methodological basis for modeling the antagonistic agents’ behavior. In the proposed multi-model complexes, the individual elements and functions of the entities under study are described using various classes of models at a certain level of detail. Coordinated use of various models allows improving the quality of modeling by compensating for the shortcomings of some models by the advantages of others, in particular, reflecting the dynamics of interaction in system-dynamic and agent-based models, which is difficult in classical models of game theory. Multi-model complexes allow stating the concept of «virtual modeling». This concept allows simulation using models of various classes. The choice of a class of models should correspond to the goals and objectives of modeling, the nature and structure of the source data. As a result of research, a methodology is proposed for modeling the interaction of antagonistic agents in cybersecurity systems using methods based on the proposed models of the reflective behavior of antagonistic agents under modern hybrid threats conditions

In recent years, agent-based models (ABMs) have become a prevalent approach for modelling complex urban systems. As a class of bottom-up method, ABMs are capable of simulating the decision-making as well as the multiple interactions of autonomous agents and between agents and the environment. The definition of agents' behaviour is a vital issue in implementing ABMs to simulate urban dynamics. Urban economic theory has provided effective ways to cope with this problem. This theory argues that the formation of urban spatial structure is an endogenous process resulting from the interactions among individual actors that are spatially distributed. However, this theory is used to explain urban phenomena regardless of spatial heterogeneity in most cases. This study combines GIS, ABM and urban economic models to simulate complex urban residential dynamics. The time-extended model is incorporated into an ABM so as to define agents' behaviour on a solid theoretical basis. A spatial variable is defined to address the neighbourhood effect by considering spatial heterogeneity. The proposed model is first verified by the simulation of three scenarios using hypothetical data: (1) single dominated preference; (2) varying preferences on the basis of income level; and (3) spatially heterogeneous environment. Then the model is implemented by simulating the residential dynamics in Guangzhou, China.

Chapter 28 focuses on agent-based modelling. Agent-based models are computer programs composed of autonomous agents that are diverse and interact with one another and their environment. The program is simulated over time, exploring how the microlevel actions and interactions of many agents give rise to macrolevel patterns and dynamics. The chapter discusses the application of agent-based models to exploring emergent social-ecological systems (SES) outcomes and dynamics, policy assessment, the assessment of response of SES to environmental or social change, explaining emergent SES phenomena, and as boundary objects in participatory processes. It goes on to discuss the types of SES problems and research questions commonly addressed by this set of methods, as well as their limitations, resource implications and new emerging research directions. The chapter also includes an in-depth case study showcasing the application of agent-based modelling, and suggested further readings.

Understanding and modeling behavior of multi-agent systems is a central step for artificial intelligence. Here we present a deep generative model which captures behavior generating process of multi-agent systems, supports accurate predictions and inference, infers how agents interact in a complex system, as well as identifies agent groups and interaction types. Built upon advances in deep generative models and a novel attention mechanism, our model can learn interactions in highly heterogeneous systems with linear complexity in the number of agents. We apply this model to three multi-agent systems in different domains and evaluate performance on a diverse set of tasks including behavior prediction, interaction analysis and system identification. Experimental results demonstrate its ability to model multi-agent systems, yielding improved performance over competitive baselines. We also show the model can successfully identify agent groups and interaction types in these systems. Our model offers new opportunities to predict complex multi-agent behaviors and takes a step forward in understanding interactions in multi-agent systems.

Luciano Floridi’s Information Ethics (IE) is a new theoretical foundation of Ethics. According to Floridi, ICT with all informational structures and processes generates our new informational habitat, the Infosphere. For IE, moral action is an information processing pattern. IE addresses the fundamentally informational character of our interaction with the world, including interactions with other agents. Information Ethics is macro-ethics as it focuses on systems/networks of agents and their behavior. The IE’s capacity to study ethical phenomena on the basic level of underlying information patterns and processes makes it unique among ethical theories in providing a conceptual framework for fundamental level analysis of present globalised ICT-based world. It allows computational modeling – a powerful tool for study which increases our understanding of informational mechanisms of ethics. Computational models help capturing behaviors invisible to unaided mind which relies exclusively on shared intuitions. The article presents an analysis of the application of IE as interpreted within the framework of Info-Computationalism. The focus is on responsibility/accountability distribution and similar phenomena of information communication in networks of agents. Agent-based modeling enables studying the increasing complexity of behavior in multi-agent systems when agents (actors) are ranging from cellular automata to softbots, robots and humans. Autonomous, learning artificial intelligent systems technologies are developing rapidly, resulting in a new division of tasks between humans and robots/softbots. The biggest present-day concern about autonomous intelligent systems is the fear of human loss of control and robots acting inappropriately and causing harm. Among inappropriate kinds of behavior is the ethically unacceptable one. In order to assure ethically adequate behavior of autonomous intelligent systems, artifactual ethical responsibility/accountability should be one of the built-in features of intelligent artifacts. Adding the requirement for artifactual ethical behavior to a robot/softbot does not by any means take responsibility from humans designing, producing and controlling autonomous intelligent systems. On the contrary, it will make explicit the necessity for all involved with such intelligent technology to assure its ethical conduct. Today’s robots are used mainly as complex electromechanical tools and do not have any capability of taking moral responsibility. But technology progress is remarkable; robots are quickly improving their sensory and motor competencies, and the development of artifactual (synthetic) emotions adds new dimensions to robotics. Artifactual reasoning and other information processing skills are advancing – all of which is causing significant progress in the field of Social Robotics. We have thus strong reasons to try to analyze future technological development where robots/softbots are so intelligent and responsive that they possess artifactual morality alongside with artifactual intelligence. Technological artifacts are always part of a broader socio-technological system with distributed responsibilities. The development of autonomous, learning, morally responsible intelligent agents relies consequently on several responsibility feedback loops; the awareness and preparedness for handling risks on the side of designers, producers, implementers, users and maintenance personnel as well as the support of the society at large which will provide a response on the consequences of the use of technology. This complex system of shared responsibilities should secure a safe functioning of hybrid systems of humans and intelligent machines. Information Ethics provides a conceptual framework for computational modeling of such socio-technological systems. Apart from examples of specific applications of IE, interpretation of several widely debated questions, such as the role of Levels of Abstraction, naturalism and complexity/diversity in Information Ethics, is offered through Info-Computationalist analysis.

Agent-Based Models (ABMs) assist with studying emergent collective behavior of individual entities in social, biological, economic, network, and physical systems. Data provenance can support ABM by explaining individual agent behavior. However, there is no provenance support for ABMs in a distributed setting. The Multi-Agent Spatial Simulation (MASS) library provides a framework for simulating ABMs at fine granularity, where agents and spatial data are shared application resources in a distributed memory. We introduce a novel approach to capture ABM provenance in a distributed memory, called ProvMASS. We evaluate our technique with traditional data provenance queries and performance measures. Our results indicate that a configurable approach can capture provenance that explains coordination of distributed shared resources, simulation logic, and agent behavior while limiting performance overhead. We also show the ability to support practical analyses (e.g., agent tracking) and storage requirements for different capture configurations.

The practical utility of agent-based models in decision-making relies on their capacity to accurately replicate populations while seamlessly integrating real-world data streams. Yet, the incorporation of such data poses significant challenges due to privacy concerns. To address this issue, we introduce a paradigm for private agent-based modeling wherein the simulation, calibration, and analysis of agent-based models can be achieved without centralizing the agents' attributes or interactions. The key insight is to leverage techniques from secure multi-party computation to design protocols for decentralized computation in agent-based models. This ensures the confidentiality of the simulated agents without compromising on simulation accuracy. We showcase our protocols on a case study with an epidemiological simulation comprising over 150,000 agents. We believe this is a critical step towards deploying agent-based models to real-world applications.