Methodological Approach to Agent-Based Modeling of Social Networks Olga Vasylieva1, Borys Butvin2 and Yuriy Shtyfurak3 1 Foreign Intelligence Service of Ukraine, 24/1, Nahirna Street, Kyiv, 04107, Ukraine 2 Central Research Institute of the Armed Forces of Ukraine, 28b Povitroflotskyi Avenue, Kyiv, 03049, Ukraine 3 National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, 37, Prosp. Peremohy, Kyiv, 03056, Ukraine Abstract The article considers social networks as an environment with a complex structure, which is dynamically changing and difficult to analyze. Here is presented the basic methodological approaches for application in the theoretical analysis of social networks. Further, it is proposed to apply the method of agent-based modeling, which today, in the authors’ opinion, can be the most adaptive for modeling internal processes of such dynamic social systems as social networks. The most popular software tools for building agent-based models are described and the software AnyLogic developed by XJTeknologies, which has a number of competitive advantages, is highlighted as the most versatile and multifunctional. Keywords 1 Agent-based modeling, social networks, information influence 1. Introduction linguistics, psychology and other knowledge can become a tool to manipulate human communities. The most acceptable instrument for Aim of the study. Modeling of social processes informational influence in terms of its own such as information influence in the social structure and the availability of appropriate target networks by applying agent-based modeling and audiences are social networks today. The term selecting the most appropriate software tool. "virtual (network) community" was firstly Today, the information influence on human introduced in 1993 by G. Rheingold, who defined resource has a special place in the system of it this way: "Virtual communities are social management decision-making, the political associations, growing out of the Web, when a component, the development of business group of people maintain an open discussion long processes, etc. This has become possible due, enough and humanly enough to form a network of firstly, to the rapid development of digital personal relations in cyberspace" [1]. technologies, including those used for data Social networks today are an important dissemination and information exchange, the element of the structure of modern society, and creation and development of new types of their influence extends to various spheres of information resources, increased access to human activity: production, daily life, culture, information for all segments of the population. politics, etc. They perform communication, Second, the large masses of information that informational, entertaining, socializing functions circulates in cyberspace, is open, easily accessible in the society; they provide opportunities for self- and such that with the help of technology, expression, exchange of information and experience, without any age, professional or any ISIT 2021: II International Scientific and Practical Conference «Intellectual Systems and Information Technologies», September 13–19, 2021, Odesa, Ukraine EMAIL: olga.vasiljeva37@gmail.com (A. 1); butvin_bl@ukr.net (A. 2); yura.shtyfurak@gmail.com (A. 3) ORCID: 0000-0001-8263-782X (A. 1); 0000-0001-6086-6592 (A. 2); 0000-0001-7863-8862 (A. 3) ©️ 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CEUR Workshop Proceedings (CEUR-WS.org) other restrictions [2]. The statistics are provided 2. Dynamic – attention is focused on below demonstrate significant penetration of changes in the network structure over time. social networks in the life of modern human The reasons of disappearance and appearance society. of edges of a network; changes of structure of For example, statistics of social networks in a network in case of external influences; 2021 showed that 42% of the world population - stationary configurations of a social network that is a colossal 3.2 billion people - use social are studied [6]. networks. The number of social network users has 3. Normative – studies the level of trust increased by more than 13% over the last year, between actors, as well as norms, rules and with almost half a billion new users registered sanctions that affect the behavior of actors in before 2021. On average, more than 1.3 million the social network and the processes of their new accounts appeared every day in 2020, that's interaction. In this case, the social roles about 15.5 new users per second, according to the associated with a particular edge of the Digital Global 2021 report [3]. network are analyzed [6]. A social network is a set of agents (vertices) 4. Resource – considers the actors' ability to that can interact with each other, and the attract individual and network resources to connections between them are social. From a achieve certain goals and differentiates actors formal point of view, such networks are dynamic being in identical structural positions of the social systems, it is convenient to present them in social network, according to their resources. the form of graphs and apply the appropriate Individual resources can be knowledge, mathematical apparatus for their analysis. prestige, wealth, ethnicity, gender (gender This is what determines the relevance of identity). Under network resources understand applying methodological approaches for influence, status, information, capital [6]. modeling such systems in order to further predict In order to involve all four mentioned various social processes in networks, including approaches of social network analysis, taking into information influence. account both the structure of dynamic system and Since formalized models of information its individual actors, defining their interaction, dissemination process in social networks are certain rules of the game, as well as giving them actually absent, taking into account their inherent certain characteristics, the authors proposed to subjective uncertainty, modeling the processes consider such method of simulation modeling as that occur in them, as well as modeling social Agent-Based Modeling [7]. network itself turns into a weakly formalized Agent-based modeling is a simulation method problem [4]. that examines the behavior of decentralized As a first step, we should consider the autonomous agents, their interaction (both methodological approaches that are used to individual agents and collective, such as analyze social networks. They differ from organizations or groups) and how such behavior traditional approaches in sociological sciences. determines the behavior of the entire system as a The postulate that the attributes of individual whole. It combines elements of game theory, actors are less important than the relationships and complex systems, emergence, computational connections with other actors in the network sociology, multi-agent systems, and evolutionary comes to the fore. That is, the attributes of programming. It’s used in noncomputational individual actors - friendliness or unfriendliness, scientific fields including biology, ecology, and level of intelligence - begin to play no major social sciences. role [5]. Agent-based modeling is used to analyze Currently there are four main methodological decentralized systems that are quite large, have approaches to the analysis of social networks: heterogeneous structures, and are dynamically 1. Structural – emphasizes the geometric changing: old connections die off and new ones shape and intensity of interactions (weight of appear. That is why this method is effective to edges). All actors are viewed as vertices of a study the process of spreading of information graph, which influence on the configuration of influence in social networks. edges and other actors in the network. Special To create an agent-based model, all actors are attention is paid to the mutual arrangement of viewed as separate agents. According to the vertices, centrality, and transitivity of structural approach, the social network can be interactions [6]. viewed as vertices with certain connections between them. The network structure of the model is dynamic, it is a kind of system that is self- of personal characteristics, which determines the created, the elements of which appear and die. In resource approach. such a system, the rules of behavior of each of the Considering the paradigms of system agents and their social roles are also defined. simulation modeling (Figure 1) Finally, each of the agents is given a certain pool Figure 1: Three paradigms of system simulation modeling it can be concluded that agent-based modeling possibility to simulate communication and is the most multiple-purpose one. In contrast to information exchange [8]. discrete event modeling, which is consonant with During simulation experiments, computational the low and medium abstraction level, and the complexities can arise because agent-based system dynamics approach with a high abstraction models on average require more hardware and level, agent-based models can be both very software power to run simulations than system detailed, when agents represent physical objects, dynamics or discrete event simulations. Agent- and extremely abstract, when competing based simulations can be implemented on small companies or state governments are modeled desktop computers, or using large clusters of using agents. computers, or any variation between the first two. The main difference between the agent-based Desktop agent-based models can be simple approach and the first two is the bottom-up and used mostly to teach how to model using construction of the model. Dependencies between agents, test agent-based model development aggregated quantities are not set on the basis of concepts, and analyze the results. Desktop utilities knowledge about the real world, but are obtained include spreadsheets, particularly Excel, and in the process of modeling individual behavior of mathematical computing systems such as tens, hundreds or thousands of agents, their Mathematica and MATLAB. interaction with each other and with objects, Large-scale agent-based models extends the which are modeling the environment. capabilities of simple agent-based desktop models The advantages of the agent-based approach and allows a larger number of agents (thousands include: the absence of determinacy in the to millions) to participate in complex interactions. behavior of the system at the global level that can Large-scale agent-based modeling is usually lead to new hypotheses about its functioning performed using dedicated modeling during model simulation; realism and flexibility environments that include a time-based scheduler, in describing the system, the ability to model the communication mechanisms, flexible agent most complex nonlinear feedbacks and to use any interaction topologies, a wide variety of devices required level of detail and abstraction. In agent- for storing and displaying agent state [9]. based modeling, there are no restrictions on the Due to the fact that the agent-based approach heterogeneity of model elements; but there is emerged in the 1990s in the U.S. university environment, so far most of the tools are intended for academic and educational purposes, and are and Java. Unlike Repast, the Swarm scheduler not commercial products in full. only supports time progression at fixed intervals. One of the most popular developments of this Swarm supports a full set of communication type is the Swarm environment. It’s a collection mechanisms and can simulate all major of C language libraries created at the Santa Fe topologies. Swarm includes a good set of utilities Institute. The most famous commercial tools are for storing and displaying agent states. Since RePast, AnyLogic, NetLogo and MASON. Swarm is based on a combination of Java and MASO is a fast multi-agent modeling toolkit Objective-C, it is object-oriented. But this mix of in Java that was developed as a framework for a languages causes some difficulties with wide range of multi-agent modeling tasks, from integration into some large-scale development swarm robotics to machine learning and socially environments, such as Eclipse [10]. complex environments. MASON makes a careful NetLogo is another cross-platform agent- distinction between models and visualization, based simulation environment that is widely used allowing models to be dynamically separated and supported. Originally based on the StarLogo from or attached to visualizers, and platforms to system, NetLogo adapts agent-based systems be changed at runtime. MASON is a collaborative consisting of a combination of live and software effort between the Computer Science Department agents. It is ideal for modeling complex systems at George Mason University and the Center for containing hundreds or thousands of agents Social Complexity at George Mason University. interacting simultaneously. It allows users to One of interest sources is social and biological explore the relationship between micro-level models, particularly models of economics, land agents and behavior at macro-level. The language use, politics, and population dynamics [10]. has been developed heavily influenced by Logo The REcursive Porous Agent Simulation and is intended for users from many disciplines - Toolkit (Repast) is the open and free source of economists, anthropologists, physicists and many libraries for large-scale agent-based modeling. others. The interface allows users to interact with Repast supports the development of extremely variables within a simulation and visualize results flexible agent-based models and is used in social without having to look at the code itself. The process modeling. Users build their model by language is similar to English, which makes it incorporating components from the Repast library easy for a non-specialist to understand the into their programs or by using visual Repast for functionality of each line of code. In addition, the Python Scripting environment. NetLogo contains an extensive library of models Repast has a sophisticated built-in scheduler that includes example programs from a wide that supports discrete-event modeling and allows variety of disciplines, which is very useful for using a large set of communication mechanisms teaching and learning purposes [11]. with a variety of interaction topologies and AnyLogic is a development of XJTeknologies, includes a full set of utilities for storing and which has found wide application among users. displaying agent states. The system also includes The competitive advantage of AnyLogic is the utilities for automatic integration with both support of all three simulation paradigms and the commercial and freely available geographic ability to use them within a single model. information systems (GIS). Integration with AnyLogic also features a powerful productive commercial GIS includes automatic connection to kernel that can simulate the behavior of millions widely used geographic information systems such of agents; rich animation and graphical model as ESRI and ArcGIS. Moreover, since Repast is description capabilities; support for various types based on the Java language, the Microsoft .NET of experiments, including sensitivity analysis, platform and Python scripts, it is fully object- Monte Carlo method, built-in OptQuest oriented [9]. optimizer; integration capabilities with databases, Swarm was the first agent-based development ERP and CRM systems; a set of library objects environment. First launched in 1994 by Chris from logistics, business processes, and pedestrian Langton at the Santa Fe Institute Swarm is an dynamics areas. open source and free set of libraries and is During developing an agent-based model in currently maintained by the Swarm Development AnyLogic, the user inputs agent parameters Group (SDG). The Swarm modeling system (people, companies, assets, projects, vehicles, consists of two core components. The kernel cities, animals etc.), defines their behavior, places components run simulation code written in them in an environment, establishes possible general-purpose language Objective-C, Tcl/Tk, connections and then runs the simulation. The individual behavior of each agent forms the global Figure 2: Structure of agent-based modeling the behavior of the simulated system [12]. Covid_19 infection spread There are also some "templates" that simplify model creation and are included in AnyLogic: Figure 3 shows the dynamics of epidemic - standard architecture; spread over time. - agent-based synchronization ("steps"); - state (continuous or discrete); - mobility and animation; - agent-based connections (networks, e.g., social networks) and communication; - dynamic creation and destruction of agents. AnyLogic provides a graphical language that greatly simplifies the creation of agent-based models:  statecharts for specifying agent behavior;  the action diagram for describing complex algorithms;  the "Environment" element is used to describe the "world", in which agents "live" and to collect various statistics;  the "Event" element is used to describe random or periodically occurring events. It should be noted that this software was used in scientific studies regarding the prediction of the spread of Covid_19 infection in dynamic social Figure 3: The changing dynamics of epidemic groups, which are immanently identical to social networks. These studies showed high reliability of spread results over time. the obtained results: the experimental results coincided with the real ones. Fig.2 shows an It should also be noted that these tools allow example of the agent-based modeling application describing almost all behavioral features of - model structure of the Covid_19 infection spread agents. Moreover Java language allows study on a local multiplicity (N = 10,000) of simulating any special behavior or logic. Also the agents. specific character of AnyLogic is possibility to combine agent-based models with discrete-event and system-dynamic models. That is why the authors of the article propose to use AnyLogic software for modeling social networks. 2. Conclusions Therefore, agent-based modeling allows creating a model of a social network, where, for example, social processes such as distribution of certain information take place, that is information influence is carried out, and all basic methodological approaches to the analysis of social networks are applied. Also, by analyzing the comparative characteristics of the most popular agent-based modeling platforms, it was determined that the most adaptive and multiple- purpose, as well as supporting a pool of platforms for optimization, is the AnyLogic platform, which among other things is designed to model such complex dynamic systems as social networks and mentalnye_sredstva_agentnogo_modelirova their internal processes, such as the distribution of niya information influences. [10] Замятина Е.Б. Современные теории имитационного моделирования. 3. References Специальный курс/ Пермский гос. ун-т. Уч. пособие, 2007. - 119 с. [11] Dante Sblendorio Agent-based Modeling in [1] Патаракин Е.Д. Сетевые сообщества и NetLogo [Електронне джерело] – Режим обучение [Текст] / Е.Д. Патаракин. – М.: доступу: https://sweetcode.io/agent-based- «ПЕР СЭ». – 2006. – 112 с. modeling-netlogo/ (дата звернення [2] Ажмухамедов И.М., Мачуева Д.А., 21.07.2021) Жолобов Д.А. Моделирование процесса [12] М.В. Киселева Имитационное распространения информации в моделирование систем в среде AnyLogic: социальных сетях // Фундаментальные учебно- методическое пособие/ М.В. исследования. – 2017. – № 5. – С. 9-14. Киселёва. Екатеринбург : УГТУ - УПИ, [3] 3. Digital Global 2021 [Електронне 2009. 88 с. джерело] – Режим доступу:https://wearesocial.com/digital- 2020. [4] Chi-Jen Lin. Assignment Problem for Team Performance Promotion under Fuzzy Environment // Hindawi Publishing Corporation. Mathematical Problems in Engineering. 2013. 10 p. [Електронне джерело] – Режим доступу: http://dx.doi.org/10.1155/2013/791415. [5] Федонюк А.А., Деякі аспекти моделювання соціальних мереж, 2014 [Електронне джерело] – Режим доступу: http://science.lp.edu.ua/sites/default/files/Pa pers/plugin-56_33.pdf. [6] Hanneman, Robert A. and Mark Riddle. Introduction to social network methods. - Riverside, CA: University of California, Riverside, 2005. - 322 p. [Електронне джерело] – Режим доступу: http://faculty.ucr.edu/~hanneman/. [7] Бутвін Б. Л., Штифурак Ю. М. Методичні засади дослідження складних динамічних систем на основі агентного моделювання // Матеріали VII міжнародної наукової конференції «Актуальні проблеми забезпечення кібербезпеки та захисту інформації» – 2021 – С. 8-10. [8] Сидоренко В.Н. Имитационное моделирование в науке и бизнесе: подходы, инструменты, применение Моделирование и анализ бизнес- процессов Бизнес-информатика №2(08)- 2009г. ст.52-56 [9] Инструментальные средства агентного моделирования Портал студенческих и научных материалов Ozlib.com. [Електронне джерело] – Режим доступу: https://ozlib.com/847655/informatika/instru