Interacting complexities of herds and social organizations agent based modeling

Although the SD methodology is still widely used and useful, there are other ways for model building, like fuzzy logic, differential inclusions, discrete event simulation, and agent base

Evolutionary Economics and Social Complexity Science 19

Stanislaw Raczynski

Interacting

Complexities of Herds and Social Organizations

Agent Based Modeling

Evolutionary Economics and Social Complexity Science

The Japanese Association for Evolutionary Economics (JAFEE) always has adhered

to its original aim of taking an explicit "integrated" approach This path has been followed steadfastly since the Association’s establishment in 1997 and, as well, since the inauguration of our international journal in 2004 We have deployed an agenda encompassing a contemporary array of subjects including but not limited to: foundations of institutional and evolutionary economics, criticism of mainstream views in the social sciences, knowledge and learning in socio-economic life, development and innovation of technologies, transformation of industrial organizations and economic systems, experimental studies in economics, agent- based modeling of socio-economic systems, evolution of the governance structure of firms and other organizations, comparison of dynamically changing institutions of the world, and policy proposals in the transformational process of economic life In short, our starting point is an "integrative science" of evolutionary and institutional views Furthermore, we always endeavor to stay abreast of newly established methods such as agent-based modeling, socio/econo-physics, and network analysis as part of our integrative links

More fundamentally, “evolution” in social science is interpreted as an essential key word, i.e., an integrative and /or communicative link to understand and re-domain various preceding dichotomies in the sciences: ontological or epistemological, subjective or objective, homogeneous or heterogeneous, natural or artificial, selfish or altruistic, individualistic or collective, rational or irrational, axiomatic or psychological-based, causal nexus or cyclic networked, optimal or adaptive, micro- or macroscopic, deterministic or stochastic, historical or theoretical, mathematical or computational, experimental or empirical, agent-based or socio/econo-physical, institutional or evolutionary, regional or global, and so on The conventional meanings adhering to various traditional dichotomies may be more or less obsolete, to be replaced with more current ones vis-à-vis contemporary academic trends Thus we are strongly encouraged to integrate some of the conventional dichotomies

These attempts are not limited to the field of economic sciences, including management sciences, but also include social science in general In that way, understanding the social profiles of complex science may then be within our reach

In the meantime, contemporary society appears to be evolving into a newly emerging phase, chiefly characterized by an information and communication technology (ICT) mode of production and a service network system replacing the earlier established factory system with a new one that is suited to actual observations In the face of these changes we are urgently compelled to explore a set of new properties for a new socio/economic system by implementing new ideas We thus are keen to look for “integrated principles” common to the above-mentioned dichotomies throughout our serial compilation of publications We are also encouraged to create

a new, broader spectrum for establishing a specific method positively integrated in our own original way

More information about this series at http://www.springer.com/series/11930

ISSN 2198-4204 ISSN 2198-4212 (electronic)

Evolutionary Economics and Social Complexity Science

ISBN 978-981-13-9336-5 ISBN 978-981-13-9337-2 (eBook)

https://doi.org/10.1007/978-981-13-9337-2

© Springer Nature Singapore Pte Ltd 2020

This work is subject to copyright All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed.

The use of general descriptive names, registered names, trademarks, service marks, etc in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.

The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors

or omissions that may have been made The publisher remains neutral with regard to jurisdictional claims

in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Preface

According to John von Neumann, “by a model is meant a mathematical construct which, with the addition of certain verbal interpretations, describes observed phe-nomena The justification of such a mathematical construct is solely and precisely that it is expected to work — that is, correctly to describe phenomena from a reason-ably wide area.” Humans always (sometimes unconsciously) have used models cre-ated in their brains When our technical skills have grown, the models acquired the form of physical, scale models, drawings, and finally sophisticated logical and mathematical constructions The common concept of modeling is defined as a sci-entific activity, the aim of which is to make a particular part or feature of the world easier to understand

The complexity of the real world can be modeled to some extent There are many definitions of complexity, recently related to “system of systems” structures Note that a system that contains a great number of sub-systems or items or a huge number

of differential equations is not necessarily complex The complexity lies in the way the components interact with each other and the diversity of system components In such systems, the simulation results may provide information about the behavior of the whole system, which is not the sum of individual behavior patterns This is also interpreted as nonlinearity This book is focused on this kind of modeling and simu-lation experiments

Analog and digital computers gave us a powerful tool for model building and analysis At the very beginning of the computer era, the differential equations have been solved on analog machines, helping scientists and engineers to design mecha-nisms, circuits, and complex devices The field of model applications has grown over the decades, including not only the works of engineering and exact sciences but also the models of animal and human societies

At the very beginning, model builders have been looking for some kinds of braic, ordinary, or partial differential equations to describe real system behavior The most known and explored field is the System Dynamics (SD) approach that mainly uses models in the form of ordinary differential equations However, it should be noted that this is not the only way to build models A strange conviction aroused among the modelers that everything in the real world can be described by

differential equations In general, this is not true Although the SD methodology is still widely used and useful, there are other ways for model building, like fuzzy logic, differential inclusions, discrete event simulation, and agent based models, among others

The topic of this book is agent based modeling The rapid growth of the tational capacity of new computers permits us to create thousands of objects in computer memory and make them interact with each other In agent based models, the objects are equipped with certain artificial intelligence, can optimize their behavior, and take decisions Some systems can be modeled both using differential equations and agent based approach The results of these two methods are frequently quite different, for example, results of the Lotka-Volterra prey-predator model and the prey-predator agent based model Here, we will not suggest which of these mod-els is valid or not These are just different modeling methods that produce results of different kind Undoubtedly, agent based modeling is more flexible and can reflect more behavioral patterns of the individuals, providing the insight on the macro- behavior of the system In Chap 1, there are comments on some agent based model-ing tools The other chapters contain examples of applications to artificial societies and competing populations of individuals and the growth, interactions, and decay of organizations and other applications For reader’s convenience, a short recall about object- and agent-based modeling is repeated in each chapter Thus, each chapter can be read as independent unit In Chap 9, you can find a description of an experi-mental software package that uses the classic continuous system dynamics graphi-cal user interface (GUI) that is used to construct the model However, the transparent simulation engine that runs behind this GUI is discrete event simulation This way,

compu-we can compare the results of the conventional system dynamics packages with these provided by discrete event simulation The relevant differences between these two simulation paradigms are pointed out

Preface

struc-A Self-destruction game, Nonlinear Dynamics, Psychology, and Life Sciences,

2006, Vol 10, no 4, used in Chap 7 of this book,

The spontaneous rise of the herd instinct: agent-based simulation, Nonlinear Dynamics, Psychology, and Life Sciences, to appear, used in Chap 5 of this book

Simulation of the dynamic interactions between terror and anti-terror tional structures, Journal of Artificial Societies and Social Simulation, Vol 7, no

organiza-2, used in Chap 3 of this book

Influence of the gregarious instinct and individuals’ behavior patterns on macro migrations: simulation experiments, Journal of Human Behavior in the Social Environment, Vol 28, no 2, used in Chap 6 of this book Visit the journal home page at www.tandfonline.com

Stanislaw Raczynski

Contents

1 Agent-Based Models: Tools 1

1.1 General Remarks 1

1.2 Discrete Event Simulation 2

1.2.1 GPSS 4

1.2.2 Arena 4

1.2.3 SIMIO 5

1.2.4 Simula 5

1.2.5 PASION, PSM++, and BLUESSS 6

1.3 Example 12

1.4 Conclusion 16

References 16

2 Simulating Self-Organization and Interference Between Certain Hierarchical Structures 19

2.1 Introduction 19

2.2 The Model 21

2.2.1 General Concepts 21

2.2.2 Interaction Rules 23

2.3 Simulation 25

2.4 Conclusion 27

References 28

3 Interactions Between Terror and Anti- terror Organizations 31

3.1 Introduction 31

3.2 The Model 33

3.2.1 Interactions Between Structures 36

3.2.2 Simulation Tool and Model Implementation 37

3.2.3 Simulation Experiments 40

3.3 Conclusion 45

References 45

4 Organization Growth and Decay: Simulating Interactions

of Hierarchical Structures, Corruption and Gregarious Effect 47

4.1 Introduction 47

4.2 Agent-Based Modeling 49

4.3 Simulation Tool 51

4.4 The Model 52

4.4.1 The Individuals 52

4.4.2 Organizations 54

4.4.3 Auxiliary Control Process 55

4.5 Simulation Experiments 55

4.5.1 Experiment 1: Criterion Function Zero 57

4.5.2 Experiment 2: Change Criterion – Size 58

4.5.3 Experiment 3: Corruption Level 59

4.5.4 Experiment 4: Accumulated Corruption 59

4.5.5 Experiment 5: Criterion – Grow Rate (Herd Instinct) 61

4.6 Conclusion 63

References 63

5 The Spontaneous Rise of the Herd Instinct: Agent-Based Simulation 67

5.1 Introduction 67

5.2 Agent-Based Modeling 69

5.2.1 General Remarks 69

5.2.2 BLUESSS Simulation Package 70

5.3 The Model 71

5.3.1 Environment 71

5.3.2 Event: Search for Food 73

5.4 Simulations 75

5.4.1 Gregarious Factor, Search for Food 75

5.4.2 The Influence of the Threat 76

5.5 Conclusion 79

Appendix 80

References 81

6 Influence of the Gregarious Instinct and Individuals’ Behavior Patterns on Macro Migrations: Simulation Experiments 83

6.1 Introduction 83

6.2 Object- and Agent-Based Models 84

6.3 The Simulation Tool 85

6.4 The Model 86

6.5 Simulations 89

6.6 Similarity to the Real Data 94

6.7 Conclusion 95

References 96

Contents

7 Simulating Our Self-Destruction 97

7.1 Introduction 97

7.2 The Model 99

7.3 Findings 101

7.4 Conclusion 104

References 105

8 Prey-Predator Models Revisited: Uncertainty, Herd Instinct, Fear, Limited Food, Epidemics, Evolution, and Competition 107

8.1 Introduction 107

8.2 Continuous Model 109

8.2.1 Simple Simulation 109

8.2.2 Uncertainty and Differential Inclusions 110

8.3 Agent-Based Simulation 112

8.3.1 General Remarks 112

8.3.2 Simulation Tool 113

8.3.3 The Model 113

8.4 Simulation Experiments 115

8.4.1 Entity Attributes, More Detail 115

8.4.2 Results: Random Walk 116

8.4.3 Chase and Escape Direction Enabled 118

8.4.4 Food, Chase/Escape Enabled 119

8.4.5 Gregarious Instinct 120

8.4.6 Fear, Food, and Energy 121

8.4.7 Epidemics, Disaster 122

8.4.8 Evolution 124

8.4.9 Variance Analysis 125

8.5 Competition 127

8.6 Conclusion 131

References 131

9 Discrete Event Simulation vs Continuous System Dynamics 133

9.1 Introduction 133

9.2 The DESD Tool 135

9.3 Examples 136

9.3.1 A Simple Birth-Death Process 136

9.3.2 Prey-Predator Model 138

9.4 Conclusion 141

References 141

References 143

Index 149

Contents

© Springer Nature Singapore Pte Ltd 2020

S Raczynski, Interacting Complexities of Herds and Social Organizations,

Evolutionary Economics and Social Complexity Science 19,

https://doi.org/10.1007/978-981-13-9337-2_1

Chapter 1

Agent-Based Models: Tools

1.1 General Remarks

The methodological focus of this book is the object- and agent-based simulation

No state equations or system dynamics schemes are used Recall that in the discrete object-based modeling, we create objects that behave according to the user-defined rules and execute their events in discrete moments of the model time The agent- based models manage objects called agents, which are equipped with certain “intel-ligence.” They can take decisions, optimize their actions, and interact with each other and with the environment Agent-based models (ABMs) are a type of microscale models that simulate the simultaneous operations and interactions of multiple agents in an attempt to recreate and predict the appearance of global com-plex phenomena

The individuals in ABM models may be of different types Although the rules of behavior are the same for individuals of the same type, the behavior is not identical for all of them This modeling method has many applications, mainly in ecology, biology, and social sciences A key notion is that simple behavioral rules (micro model) generate complex (macro) behavior An important central tenet is that the whole is greater than the sum of the parts Individual agents are typically character-ized as rational They are presumed to be acting in what they perceive as their own interests, such as reproduction, economic benefit, or social status, using heuristics

or simple decision-making rules (Railsback et al 2006; Bandini et al 2009) Note the main difference between object-oriented and simulation package The latter, in addition to object creation, provides (or should provide) a “clock” mechanism that automatically manages the model time and event execution The ABM modeling is supported by many programming and simulation tools Let us list only some of the most popular tools: SWARM developed in 1994 by the Santa Fe Institute (Swarm Development Group, 2001), Ascape developed in 2001 (Parker 2001), Breve-2.7.2 (Klein 2002), Recursive Porous Agent Simulation Toolkit released in

2003 (Michael et al 2006), Cormas developed in 2004 by VisualWorks (Bommel

et  al 2015), MASON (Luke et  al 2005), MASS package (Tatai et  al 2005), FLAME (Coakley et al 2006; Holcombe et al 2013), MATSim of EHT Zürich (Bazzan and Klugl 2009), and SOARS developed in 2010 (Tanuma et al 2005,

2006), among others

ABMs are widely used in modeling of the organization dynamics An example of

an agent-oriented model, called the BC model, can be found in the article by Krause (2000) In that model, the agent’s attributes include “opinions,” and the interaction between agents depends on the distance between their opinions in a nonlinear way These interactions can result in an action being taken by the agent Other examples

of models of social structures based on the concept of opinion interactions can be found in Latane and Nowak (1997) and Galam  and Wonczak (2000) A similar approach is presented by Chatterjee  and Seneta (1977) and Cohen  et  al (1986) These works refer to the dynamics of forming of social groups in accordance with the existing agents’ attributes (opinions) Some quite interesting results, more closely related to the terrorism problem, are described by Deffuant et al (2002).Some more general concepts of “computational sociology” and agent-based modeling (ABM) can be found in the article of Macy and Willer (2002) Other gen-eral recommended readings in the field are Bak (1997), Cioffi-Revilla (1998), Gotts

et al (2003), Axelrod (1997), Epstein and Axtell (1996), and Holland (1998) An interesting contribution to a model of the structure of the Osama bin Laden organi-zation is included in a Vitech Corporation page (link: see Long 2002) Other (ABM)-oriented approach can be found in Crowder et al (2012) and Hughes et al (2012)

In these publications we can find discussions about the potential advantages of the ABM approach through a range of examples and through the identification of opportunities in the field of organizational psychology

Another approach is used by Lustick (2000), where the agents interact on a scape It is shown that macro-patterns emerge from micro-interactions between agents An important conclusion is that such effects are more likely when a small number of exclusivist individuals are present in the population The simulations of other mechanisms of clustering in agent-oriented models are described by Younger (2003), who deals with the creation of social structures in the process of food and material storage

land-1.2 Discrete Event Simulation

Recall that by the model time, we understand the time variable that is controlled by the simulation program during the simulation run The real time represents the time

of our (or computer) physical clock For example, simulating the movement of a galaxy, we can simulate several millions of model time years On a fast computer, his simulation may take several minutes in the real time

1 Agent-Based Models: Tools

There are many real systems, where we can define the processes named events

that consist in changing the state of the system For example, the events may describe the start or the end of a service process and a birth or death of a model entity or taking place in a waiting line In many situations such events can be considered to

be executed in a very small interval of time, compared to the total length of model simulation time The discrete event simulation means that we suppose that the model events are discrete, i.e., they are accomplished within model time interval of length zero This model simplification makes the simulations very fast

The Discrete Event Specification (DEVS) formalism is used to describe models

in discrete event simulation In the DEVS formalism, an “atomic” model M is

Atomic models can be coupled to form a coupled model The coupled models can also be coupled in hierarchical way to form more complex models The coupled DEVS model is as follows:

coupled DEVSº X self,Y self, ,D M{ } { } { }i , I i , Z j select i, ,

The subindex self-denotes the coupled model itself D is a set of unique

compo-nent references The set of compocompo-nents is:

The select component defines the order of execution for simultaneous events that may occur in the coupled model This component must be added to the model to avoid ambiguities in the simulation algorithm and to make the model implementation- independent There is a huge research done on the select algorithms because the treating of the simultaneous events is rather difficult task

To treat complex models with variable structure, the Dynamic Structure Discrete Event System Specification (DSDEVS) is used We will not discuss the DSDEVS formalism here The use of the DEVS formalism is relevant in big models, where the time of execution, hierarchical model building, and portability are important factors

By time and event management (TEM), we understand the time clock and event

queue management (inside the “simulation engine”), including the basic queuing model operations provided by the simulation package The object behavior model-ing (OBM) is a set of additional items like user-defined distributions and logical functions, nontypical operations, object attributes, and the general object behavior

1.2 Discrete Event Simulation

Let us start with GPSS (General Purpose Simulation System), omitting earlier tools like the forgotten but very nice language of the 1950s CLS (control and simu-lation language)

1.2.1 GPSS

This language, developed primarily by Geoffrey Gordon at IBM around 1960 (Gordon 1975), has contributed important concepts to every discrete event simula-tion language developed ever since This is an old tool, but it is still used and works perfectly In fact, GPSS is an object-oriented tool, although it does not fit into the modern object-oriented paradigms The objects in GPSS are called transactions These are moving items that appear, go through the fixed model facilities, and dis-appear GPSS World has been extended by PLUS, the Programming Language Under Simulation The TEM level instruction set of GPSS is simple and easy to use

It can be dominated by anyone in few hours of learning and running example ing models The OBM level mechanisms are not so easy Recall that the new ver-sions of GPSS have an embedded language PLUS. If the user wants to equip objects (transactions) with any additional properties and individual, nonstandard behavior,

queu-he must learn PLUS and dominate tqueu-he information about tqueu-he SNAs (standard numeric attributes) The PLUS manual is a whole chapter of the GPSS manual or a separate document of about 60 pages The SNA documentation occupies also sev-eral dozen pages, including great number of attributes and additional items Using all this stuff, the user can simulate more advanced models, but the created objects can hardly be considered as “intelligent.”

1.2.2 Arena

Arena modeling system from Systems Modeling Corporation is a nice and widely used simulation tool It is equipped with a graphical user interface (GUI) and ani-mation mechanisms (see Kelton et al., 2004) The TEM level of Arena permits to quickly create a queuing or manufacturing discrete event models, needs no coding, and results in clear flowcharts of the model The OBM level is somewhat more complicated Arena is built on the SIMAN (Pedgen et  al 1995) simulation lan-guage So, first of all, the user must learn SIMAN to be able to manage user-defined logics, statistics, and/or a nonstandard object behavior The Arena entities (moving objects) can be equipped with time attributes, cost attributes, entity-type variable, group member variables, and other The specification of the attributes and other Arena pre-defined variables takes about 30 pages in the Arena documentation Again, if the user wants to create and manage a little bit more complicated object behavior, he/she must learn SIMAN and dominate dozens of pages of the Arena manual

1 Agent-Based Models: Tools

1.2.3 SIMIO

This is a multi-paradigm software delivered by SIMIO LLC. SIMIO® is created by

a team of simulation software developers led by Dennis Pedgen and Sturrok (2010).Compared to Arena, SIMIO is a step forward in creating models with intelligent objects The object definition in SIMIO is more general Objects may be fixed facili-ties or moving dynamic objects named entities The user can define his/her own objects, store and reuse them, or use the objects from the standard library These may be fixed (server, machine), link (a pathway for entities), node (link intersec-tions), entity (dynamic object, like client in a shop), or transporter (it can pick up and drop entities at nodes)

The user defines the object properties They may be of different types such as strings, numbers, selections from a list, and expressions The properties are edited

in multiple edition windows There are many ways to define a SIMIO model A programmer familiar with an object-oriented language like C++ or Delphi can understand and dominate the SIMIO modeling in reasonable time and effort SIMIO creators claim that the process-based objects in SIMIO are both simpler and more powerful than the code-based objects in other modeling tools SIMIO offers both TEM and OBM facilities, although they are not clearly separated from each other

1.2.4 Simula

We must mention here Simula, its mostly known version 67 (Dahl and Nygaard

1967) Although it is a tool developed more than 50 years ago, it is still perhaps one

of the most advanced and elegant object-oriented languages In fact, Simula itself is just object-oriented and not a simulation language The modeling facilities have been added to the language as a part of its standard class library and are encapsu-lated in the Process class Any object that inherits the Process class properties can use the clock mechanism and event scheduling The object behavior management is coded directly in the language As for an old software, it originally had no GUI and other graphical facilities The language is rather difficult to learn and needs previous training in Algol

If we define the “intelligence” as the ability to make decisions due to a more sophisticated algorithms or equip the objects with some kind of artificial intelli-gence, only an advanced object-oriented algorithmic languages provide such fea-tures Simula has this capacity Perhaps this is the reason why Simula is still quite popular among the computer science researchers

Anyway, if someone wants to create an object-oriented simulation package with intelligent objects, he/she finally must create a new high-level object-oriented algo-rithmic language The question is: Isn’t it better to take a known, complete, widely known, used, and advanced language and add to it the time and queuing manage-ment layer (TEM)?

1.2 Discrete Event Simulation

1.2.5 PASION, PSM++, and BLUESSS

BLUESSS (Blues Simulation System) is the tool used in the simulations discussed

in this book The package evolved from the Delphi-based languages PASION and PSM++ The use of this particular package is not necessary The models described

in the following chapters may be simulated using other agent-based tools This package was used because of its relation to the C++ language

Important questions both in teaching and implementing computer simulation are:

• Must a simulationist be a programmer?

• Must he/she be a mathematician?

Depending on what is the role of the simulationist in the whole process of ing and using a simulation program, the response can be yes or not It seems that the commercial simulation tools are being developed in order to prevent the user from any coding and to make all the mathematics (statistical considerations) as transpar-ent and simple as possible To say that this is a correct tendency, first of all we must know who the simulationist is If he/she is a plant engineer, a sociologist, a ware-house manager, or just an amateur, the tool should be “fast and easy” (user of kind 1) However, if he/she is a professional simulationist and develops simulations in serious and professionally advanced way, he/she should be able to create and to code necessary algorithms and be aware of the model mathematics (user of kind 2) While teaching computer simulation, it is not always clear to which kind of users we should address This is an important question, because the contents of the simula-tion course and the tools we use strongly depend on the student/user kind

creat-It seems that the new discrete event and general-purpose packages are being designed mostly for the users of kind 1 The new software has always a well- designed graphical user interface (GUI) and offers a ready-to-use, encapsulated probability distributions and statistics However, despite of good manuals and addi-tional materials, the users of kind 1 frequently commit fundamental errors, like using the Poisson inter-arrival time distribution for the Poisson arrival process One could say that users of kind 2 can do their simulations in any algorithmic language and need no simulation packages Obviously, this is not true A simulation tool should provide ready-to-use mechanisms to avoid unnecessary work (but nothing more) In general, the question is if, in our simulations, we need intelligent objects (agents) at all The answer is yes Such objects are not needed in academic examples and simple simulations (users of kind 1) However, if we face the reality, for exam-ple, a real manufacturing system, it is quite sure that there will be objects that do not fit in standard blocks or facilities offered by most of the simulation packages In such case the use of intelligent objects will be inevitable

BLUESSS evolved from the PASION and PSM++ packages, related to Delphi Some applications and remarks on discrete event simulation in these packages can

be found in Raczynski (2000, 2004, 2006a, )

The package runs over the Embarcadero™ C++Builder The user can be of kind

1 or of kind 2 (programmer skills) Taking about a professional simulationist, we

1 Agent-Based Models: Tools

should rather think about users of kind 2 My point is that few really professional simulationists do not dominate C++ BLUESSS is a simple simulation language and has the BLUESSS-to-C++ translator So, the user can code model events in C++, if necessary, or use one of the BLUESSS code generators to create models without coding

The following modules (source code generators) are included in the BLUESSS package:

• Queuing Model Generator

• Flow diagrams Continuous simulation using signal flow diagrams

• Continuous simulation using bond graphs

• Continuous simulation, ordinary differential equations

After defining the model, the BLUESSS system generates the source BLUESSS code, translates into C++, and invokes the C++Builder which produces the execut-able program The event queue in BLUESSS works due to the three-phase discrete simulation strategy (see O’Keefe (1986))

Using the QMG module, the user defines the model in the Arena-like style, with

no coding at all As the process of creating exe file (stand-alone, independent cutable) passes through the C++ compilation, BLUESSS QMG module can use all the features of C++ (see Fig. 1.1)

exe-In other words, the comparison of QMG with, for example, Arena can be marized in the following table (Table 1.1)

sum-Note that the QMG graphical model editor is very simple and can be dominated

in 15 min of “training,” even without consulting any documentation The objects

Fig 1.1 BLUESSS

features

Table 1.1 Creating intelligent objects (Arena vs BLUESSS)

model editor Learn SIMAN, learn Arena manual including 30 pages of entity

attribute and expression specifications, code the necessary expressions

Use C++

1.2 Discrete Event Simulation

created in QMG can be equipped with simple abilities (logical expressions, tional attributes) or with any complicated decision-making algorithms, like fuzzy logic, iterative optimization algorithms, neural nets, and database consulting They can execute external programs or use external files The object can do everything what can be coded in C++ There are no restrictions on the type and size of its attri-butes (those can be numbers, strings, arrays, pointers, and/or C++ structures of any kind) If required, the object can communicate through the Internet, sing a song, display an OpenGL image, execute an external program, etc Obviously, objects cannot intervene in the TEM (time and event management) of QMG. There are also some restrictions on the use of pointers When the object disappears, it must execute

addi-a user-provided code to free the memory addi-allocaddi-ated to the pointed structures Otherwise memory leaks can occur

To create a QMG flowchart, the user picks up blocks (like GPSS facilities or Arena modules) and defines the basic block parameters like inter-arrival times, ser-vice times, etc Then the simulation can be invoked The entities (dynamic objects) appear, go through the blocks, and disappear As stated before, the additional entity attributes can be declared, being of any available C++ type

The relation between the entities and the C++ entity-related code is very simple Any entity which enters to any of the model blocks simply calls a global C++ func-

tion named SVOP Both calling entity and block identifiers, as well as all entity attributes, are passed to SVOP as actual parameters For the assembly operation,

SVOP is called by each entering entity and for the new (assembled) one So, in the

SVOP body, the user can identify the block/entity pair and code any required action

For example, entities can enable or disable model blocks (using the semaphore logic

variables) or execute more complicated actions (Fig. 1.2.)

Fig 1.2 Fragment of a QMG model Automatic calls to the SVOP function

1 Agent-Based Models: Tools

Suppose, for example, that we need the following actions to be taken:

• If an entity enters the queue number 59, its string attribute myname includes the string dog, and the entity age is greater than 100 model time units (the time spent

in the system), and then it invokes external program other.exe.

• If any entity enters the assembly block 42 and the sum S of the length of queue

39 and queue 41 exceeds 24, then close (disable) generators 29 and 40 If S is less

or equal to 24, then enable these generators

• If any entity waits in any queue for more than 20 time units, display a warning message

In the below code, n is the calling block number, SOURCE is the number of block where the entity has been created, TIMIN is the model time instant when the entity has been created, TIMQ is the time the entity has been waiting in a queue (if

it is actually waiting), and myname is an additional, user-defined entity attribute

TIME is a global variable representing the model time The function nr returns the queue length QUEx is the reference to the queue block number x, and SEMx is a

Boolean variable (a semaphore) that enables (if true) or disables (false) the block

number x DisplayWorning is a user-define C++ function (may show something on

the screen, emit a sound, etc.)

The SVOP procedure in this case may be as follows:

void SVOP(int n, int SOURCE,

float TIMIN, float TIMQ, String∗ myname)

This is a very simple example Inside the SVOP function, the user can insert any

C++ code to define the entity behavior and/or block operations

Queuing and manufacturing models of BLUESSS may use animation See Fig. 1.3 for an example of manufacturing animation

As stated before, BLUESSS is a  general-purpose package It contains several modules (source code generators) for queuing/manufacturing models, continuous simulation using ordinary differential equations, signal flow graphs, bond graphs, or combined models The user can create the source code or use any of the BLUESSS modules to avoid coding The final product is an independent exe file, ready to run The package structure is shown on Fig. 1.4 In BLUESSS everything (except the code taken from C++ libraries) passes through the BLUESSS source code and

1.2 Discrete Event Simulation

through the C++ code (generated automatically) The user can create his/her code or use the code generators The options are as follows (see Fig. 1.4):

Queuing models: The queuing module generates the source code which is

trans-lated to C++ and compiled

ODE (ordinary differential equations) module receives the right-hand sides of the

equations The rest is done automatically (source codes generation, compilation)

Block diagrams and signal flow module: The user defines graphically the model

structure and the necessary parameters The module generates the model tions; the rest is done as above

equa-Bond graphs: The user draws the bond graph model and gives its parameters The

rest is done automatically

Animator: 2D off-line animation of queuing models is available.

Variance analysis: Postmortem additional statistical analysis can be invoked This

includes the max-min and confidence intervals for the model trajectories, shown

as functions of time This feature, provided by few simulation packages, is very useful while simulating queuing and stochastic models In Fig. 1.5 an example of such analysis is shown This is the length of a simulated queue The gray region

is where the length of the queue is supposed to belong with probability 0.92 The curve inside the region is the average queue length in function of model time The average is taken over a series of repeated simulations If the gray region is big (big variance), then it can be seen how little informative the average value is

Fig 1.3 BLUESSS animation example Manufacturing

1 Agent-Based Models: Tools

Fig 1.4 BLUESSS package structure

Fig 1.5 Variance analysis of BLUESSS package Confidence intervals for the length of a waiting

line

1.2 Discrete Event Simulation

The user can see and modify both the BLUESSS and C++ codes For queuing models, he/she can also use the SVOP functions as described earlier Although the queuing models of BLUESSS are rather simple, the possibility of working on the generated code makes it possible to simulate any required object behavior

These are only some examples of BLUESSS features BLUESSS can use any tools available in C++ Interesting animations, both for continuous and discrete event models, can be created using the OpenGL graphics

Observe (Fig. 1.4) that both discrete event and continuous models result in the BLUESSS source code The only difference is that the continuous models are simu-lated as a sequence of events with a small time step, each event being a call to one

of the possible numerical methods for ODE. This means that at the source code level, the user can mix discrete and continuous models in the same simulation pro-gram In Fig. 1.6 you can see a screen of a continuous model simulation (multiple pendulum), animated with OpenGL graphics

1.3 Example

BLUESSS can be used not only to simulate queuing models, ODE models, or bond graphs Let us mention a somewhat nontypical application The mode we recall here was coded in PASION. As stated before, PASION was a precursor of PSM++ and BLUESSS packages The program structure and the concepts of processes and events are identical in these packages The only difference is that the event body in BLUESSS is coded in C++ instead of Delphi Pascal All other features are the same.This example model belongs to biological and medical applications It is a (sim-plified) model of our immune system We do not discuss this model in a separate chapter because it is an old research, published nearly 30  years ago Raczynski

Note that the birth-and-death equation describes the changes of the expected value for the size of the population, while the result of discrete simulation is a real-ization of the modeled stochastic process, i.e., a system trajectory and not an expected trajectory Observe that the trajectories of the expected values of the mod-eled variables do not provide sufficient information What it means, for example, to get a satisfactory average response of the immune system, if the “modeled patient” dies in 50% of the simulation runs? This information might be easily lost when using continuous, birth-and-death models The cost of the discrete simulation is, of course, rather high compared to the continuous modeling and rises considerably when more objects are generated The model described here should be treated as an

“immunological game” rather than a valid model of the HIS (recall that it is just an example of a 30-year-old research) The results are merely of qualitative type The model includes the following components:

• Macrophages These are primary defenders They consume wide amount of debris from our bloodstream as well as invading bacteria and viruses They also activate other defense mechanisms by activating the helper T-cells

• Helper T-cells These cells are activated by macrophages and stimulate the duction of other cells of the HIS. Helper T-cells also produce interleukin-2 (IL-2) and a lymphokine BCGF make other HIS cells more active

repro-• Killer T-cells The cells of this type kill the body cells which have been infected

by a virus, disrupting its replication cycle

• B-cells Activated by helper T-cells, these cells begin to replicate and to produce antibodies which neutralize the viruses

• Antibody The elements produced by the B-cells in order to disable certain type

of viruses, recognized earlier by the macrophages

• Suppressor T-cells These cells slow the defensive activities of the HIS after the infection has been conquered

• The thymus This organ generates the T-cells which enter into the bloodstream

• Virus This is a strange invader that enters the bloodstream To multiply it must slip into a body cell where it replicates quickly

1.3 Example

The rules of interaction between the model components are as follows The infecting virus enters the bloodstream and looks for a body cell in order to infect it and replicate Some of the viruses are devoured by the macrophages and those which have entered body cells replicate rapidly The infected cells die releasing new viruses ready to infect other healthy cells A macrophage that has eaten a virus dis-plays its “antigen” on its surface and couples with a helper T-cell Some of the T-cells are able to recognize the strange antigen and become active The activation

of other killer and helper T-cells and the B-cells is done by changing the level of some bodies named lymphokines in the bloodstream Thus, while detecting a strange invader, a macrophage produces a lymphokine IL-1 which activates the helper T-cells These cells begin to produce the lymphokine IL-2, which activates other helper and killer T-cells The helper T-cells also secrete a lymphokine BCGF which activates the B-cells and produce the gamma interferon (IF) which increases the activity of the B-cells and T-cells Each active B-cell becomes a factory of pro-tein molecules called antibodies which neutralize the viruses recognized earlier by the helper T-cells The joint effort of macrophages, killer  T-cells and antibodies permit to stop the replication of the viruses, deactivate, and wipe out them from the infected organism (Perelson 1988)

The model for the infection with the AIDS virus is similar The only difference

is that this virus attacks the helper T-cells, deactivating the main part of the immune system Other infecting viruses or bacteria cannot be detected and disabled effec-tively, and the consequences of any infection can be fatal The above (simplified) mechanism of the immunological response can be treated as a sequence of events and can be coded and simulated directly using any object-oriented simulation lan-guage PASION has been used because it offers all needed features (see the BLUESSS features mentioned earlier) The objects can be generated according to the process (object-type) declarations Thus, the virus, the macrophage, and each of the cells of the HIS are described as processes A process declaration specifies the object attributes, such as its lifetime and other parameters, and describes all possible events in the “life” of the object PASION has the necessary “clock mechanism” which controls the execution of the events; it is equipped with such features as inheritance, repetitive simulation, and history file and permits all Delphi structures The environment of the language supports interactive simulation, graphics, statisti-cal analyses of the resulting trajectories, etc

Consult Raczynski (1989) for more detailed description of the simulation gram Here, we only show some results provided by PASION simulation run Figure 1.7 shows the plot of the model variables in response to a viral infection The vertical axis is the relative cell number in logarithmic scale The cell population varied up to 200–500 cells each time interval equal to 10 days

pro-The figures shown here are of rather low quality; these are screenshots from simulations carried out nearly 30 years ago Figure 1.7 shows the average simulated response to a viral infection Vertical scale shows the number of simulated cells in logarithmic scale, as functions of time In Fig. 1.8 we can see the average trajectory

1 Agent-Based Models: Tools

and the confidence intervals for the number of helper T-cells The confidence level

is set equal to 0.95 Similar plot for the number of infected body cells is shown in Fig. 1.9

As stated before, this model is just an example of possible PASION and BLUESSS applications It does not pretend to be a valid model of the immunologi-cal system because it was created three decades ago, when the knowledge about the immune system was not so advanced as in recent years Undoubtedly this research should be continued

Fig 1.7 The average simulated response to a viral infection Vertical scale shows the number of

cells in logarithmic scale

Fig 1.8 The average trajectory and the confidence intervals for the number of helper T-cells

1.3 Example

1.4 Conclusion

BLUESSS is not a widely known or important commercial software This is rather

a proposal on a possible direction in simulation software development The point is that what is really needed in discrete event simulation consists in the fast clock machine (event queue management) and good model structure definition tools.The algorithmic aspect of the modeling, like intelligent object behavior, should

be handled by a high-level algorithmic language rather than complicated parameter specification and expression building incorporated in the package Using one of already developed and powerful object-oriented languages in the background gives

us much more versatile tool to handle all that is not a simple queuing and discrete event simulation Moreover, as we talk about professional level of simulation tasks, there are few users who don’t already dominate C++ or similar tools

Fig 1.9 The average trajectory and the confidence intervals for the number of infected body cells

1 Agent-Based Models: Tools

for-Cohen JE, Hajnal J, Newman CM (1986) Approaching consensus can be delicate when positions harden Stoch Process Appl 22(2):315–322

Crowder RM, Robinson MA, Hughes HPN, Sim YW (2012) The development of an agent-based modeling framework for simulating engineering team work IEEE Trans Syst Man Cybern Part

A Syst 42(6):1426–1439

Dahl O, Nygaard B (1967) Simula  – an Algol-based simulation language Commun ACM 9:671–678

Deffuant G, Amblard F, Weisbuch G, Faure T (2002) How can extremism prevail? A study based

on the relative agreement interaction model J Artif Soc Soc Simul 5(4)

Epstein JM, Axtell R (1996) Growing artificial societies: social science from the bottom up Brookings Institution Press, Washington, DC

Galam S, Wonczak S (2000) Dictatorship from majority rule voting Euro Phys J B 18(1):183–186 Gordon G (1975) The application of GPSS to discrete system simulation Prentice-Hall, Englewood Cliffs

Gotts NM, Polhill JG, Law ANR (2003) Agent-based simulation in the study of social dilemmas Artif Intell Rev 9(1):3–92

Holcombe M, Coakley S, Kiran M (2013) Large-scale modelling of economic systems Compl Syst 22(2):175–191 http://www.complex-systems.com/pdf/22-2-3.pdf

Holland JH (1998) Emergence: from chaos to order Helix Books: Addison-Wesley Publishing Company

Hughes HPN, Clegg CW, Robinson MA, Crowder RM (2012) Agent-based modelling and simulation: the potential contribution to organizational psychology J Occup Organ Psychol 85:487–502

Kelton D, Sadowski R, Sadowski D (2004) Simulation with ARENA. McGraw-Hill, New York Klein J (2002) Breve: a 3D environment for the simulation of decentralized systems and artifi- cial life Conference paper: ICAL 2003 Proceedings of the eighth international conference on Artificial life, MIT Press, Cambridge, MA. ISBN/ISSN 0-262-69281-3

Krause U (2000) A discrete nonlinear and non-autonomous model of consensus formation In: Elaydi S, Ladas G, Popenda J, Rakowski (eds) Communications in difference equations Gordon and Breach, Amsterdam

Latane B, Nowak A (1997) Self-organizing social systems: necessary and sufficient conditions for the emergence of clustering, consolidation and continuing diversity In: Barnett FJ, Boster

FJ (eds) Progress in communication sciences v.13 Ablex Publishing Corporation ISBN-13: 978-1567502770

Long JE (2002) Systems analysis: a tool to understand and predict terrorist activities Internet munication Vitech Corporation http://www.umsl.edu/~sauterv/analysis/62S-Long-INTEL.pdf

com-Luke S, Cioffi-Revilla C, Panait L, Sullivan K (2005) MASON: a multiagent simulation ment Simulation 81(7):517–527

environ-Lustick S (2000) Agent-based modeling of collective identity J Artif Soc Soc Simul 3(1) http:// jasss.soc.surrey.ac.uk/3/1/1.html

References

Macy MW, Willer R (2002) From factors to actors: computational sociology and agent-based modeling Annu Rev Sociol 28(1):143–166

Michael JN, Nicholson T, Collier JR, Vos JR (2006) Experiences creating three implementations

of the repast agent modeling toolkit ACM Trans Model Comput Simul 16(1):1–25 https://doi org/10.1145/1122012.1122013

O’Keefe RM (1986) The three-phase approach: a comment on strategy-related characteristics of discrete event languages and models Simulation 47(5):208–210

Parker MT (2001) What is ascape and why should you care? J Artif Soc Soc Simul http://jasss soc.surrey.ac.uk/4/1/5.html

Pedgen CD, Sturrok DT (2010) Introduction to Simio Conference paper: proceedings of the 2010 Winter, PA, USA

Pedgen CD, Shannon RF, Sadowski RP (1995) Introduction to simulation using SIMAN. McGraw- Hill, New York

Perelson AS (1988) Toward a realistic model of immune system In: Theoretical immunology part

II. Addison-Wesley

Raczynski S (1989) Simulating our immune system Conference paper: simulation on puters The Society for Computer Simulation Int ISBN/ISSN 0-911801-43-X

microcom-Raczynski S (2000) Alternative mathematical tools for modeling and simulation: metric space

of models, uncertainty, differential inclusions and semi-discrete events Conference paper: European Simulation Symposium ESS2000, Hamburg, Hamburg, Germany

Raczynski S (2004) Simulation of the dynamic interactions between terror and anti-terror zational structures J Artif Soc Soc Simul 7(2) ISBN/ISSN 1460-7425

organi-Raczynski S (2006a) In: Bargiela A (ed) Modeling and simulation: computer science of illusion Wiley, Chichester

Raczynski S (2006b) A self-destruction game J Nonlinear Dyn Psychol Life Sci 10(4):471–483 ISBN/ISSN 1090-0578

Railsback SF, Lytinen SL, Jackson SK (2006) Agent-based simulation platforms: review Simulation 82(9):609–623 https://doi.org/10.1177/0037549706073695

SWARM Development Group (2001) Swarm simulation system Electronic citation Electron Citation 8(1–10) http://digitalcommons.usu.edu/nrei/vol8/iss1/2

Tanuma H, Deguchi H, Shimizu T (2005) Agent-based simulation: from modeling methodologies

to real-world applications, vol 1 Springer, Tokyo

Tanuma H, Deguchi H, Shimizu T (2006) SOARS: Spot Oriented Agent Role Simulator – design and implementation In: Agent-based simulation: from modeling methodologies to real-world applications Springer, Tokyo, ISBN 9784431269250

Tatai G, Gulyas L, Laufer L, Ivanyi M (2005) Artificial agents helping to stock up on edge Conference paper: 4th International Central and Eastern European Conference on Multi- Agent System, Budapest, Hungary, ISBN:3-540-29046-X 978-3-540-29046-9 https://doi org/10.1007/11559221_3

knowl-Younger SM (2003) Discrete agent simulations of the effect of simple social structures on the benefits of resource J Artif Soc Soc Simul 6(3)

1 Agent-Based Models: Tools

© Springer Nature Singapore Pte Ltd 2020

S Raczynski, Interacting Complexities of Herds and Social Organizations,

Evolutionary Economics and Social Complexity Science 19,

The main goal of any political party is always to obtain power and nothing more Many trade union organizations have lost sight of their original goal (defending the interests of workers) and have also become power-seeking structures The social structure acts as a new agent, using its members as nothing more than a medium to achieve its goal However, in this model an organization itself is not an active pro-cess The organization macro-patterns are results of the entity activities

The interaction between different social structures is an interesting problem and can be simulated – to some extent, of course See Raczynski (2004) for the simula-tion of interactions between terrorist and anti-terrorist structures Here a similar approach and tools are used

Many existing models of social organization dynamics are of an agent-oriented type An interesting agent-oriented model, called the BC model, can be found in the

article by Krause (2000) In that model, the agent attributes include “opinions,” and the interaction between agents depends on the distance between their opinions in a nonlinear way These interactions can result in an action being taken by the agent Other examples of models of social structures based on the concept of opinion inter-actions can be found in Latane and Nowak (1997) and Galam and Wonczak (2000)

A similar approach is taken by Chatterjee and Seneta (1977) and Cohen et al (1986) The BC model and the above works refer to the dynamics of forming social groups

in accordance with the existing agents’ attributes (opinions), rather than to events such as the destruction of a part of a treelike social structure by another (adversary) structure Some quite interesting results, more closely related to the terrorism prob-lem, are described by Deffuant et al (2002)

Another agent-oriented approach is used by Lustick (2000), where the agents interact on a landscape It is shown that macro-patterns emerge from micro- interactions between agents An interesting conclusion is that such effects are more likely when a small number of exclusivist identities are present in the population The simulation of other mechanisms of clustering in agent-oriented models is described by Younger (2003) That article deals with the creation of social structures

in the process of food and material storage

Some more general concepts of “computational sociology” and agent-based modeling can be found in the article by Macy and Willer (2002) Other general rec-ommended readings in the field are Bak (1997), Cioffi-Revilla (1998), Gotts et al (2003), Axelrod (1997), Epstein and Axtell (1996), and Holland (1998) Many other sources can be found on the Internet

It should be noted that to look for a model that simulates real human behavior is utopian Nobody has ever simulated a human in its complete (mental, emotional, physical, etc.) behavior All that can be done is to choose some little part of this complex system in order to simulate its possible actions In any case, in soft system simulation and social simulation, one can hardly (or never) find any proof that the model is valid

Interesting models and simulation experiments on the survival of societies can be found in the literature Cecconi and Parisi (1998) simulate a survival problem in terms of individual or social resource storage strategies Saam and Harrer (1999) simulate the problems of social norms, social behavior, and aggression in relation to social inequality Staller and Petta (2001) discuss the emotional factor in social modeling They introduce the emotions as an essential element of modes that simu-late social norms and aggression Stocker et al (2002) examine the stability of ran-dom social network structures in which the opinions of individuals change They show that hierarchies with few layers are more likely to be unstable than deeper hierarchies The study is related to political, organizational, social, and educational contexts rather than to the self-destruction problem itself, but it is clear that an unstable social structure may be much more vulnerable to attack There are many approaches and aspects of ecological and social models, providing certain repro-duction/death formulas See, for example, Moss de Oliveira and Stauffer (1999), for

a model of aging and reproduction

2 Simulating Self-Organization and Interference Between Certain Hierarchical…

Adamic and Adar (2005) address the question of how participants in a small world experiment are able to find short paths in a social network using only local information about their immediate contacts In the e-mail network, they find that small world search strategies using a contact’s position in physical space or in an organizational hierarchy relative to the target can effectively be used to locate most individuals The authors discuss the implications of their research to social software design

From newer publications, we should mention the book of Edmonds et al (2007) The editors aimed to present a flyover of the current state of the art They divide the papers into three parts: model oriented, empirically oriented, and experimentally oriented In the other publication of Edmonds (2012), we can find an analysis of the role and effects of context on social simulation

Silverman et al (2013) present an agent-based model of a human population The model illustrates the potential synergies between demography and agent-based social simulation Elsenbroich (2012) asks what kind of knowledge can we obtain from agent-based models The author defends agent-based modeling against a recent criticism Sibertin-Blanc et al (2013) present a framework for the modeling, the simulation, and the analysis of power relationships in social organizations and more generally in systems of organized action In that article we can find a discus-sion of a model of bounded rational social actors and analytical tools for the study

of the internal properties of organizations The model may explain why, in an nizational context, people behave as they do

orga-2.2 The Model

2.2.1 General Concepts

Our model consists of three hierarchical structures interacting with each other over

a common (abstract) region Let us comment some terms used in the sequel:

Entity An individual that can be a member of a hierarchical structure.

Organization A collection of entities, with a hierarchical structure In this

simula-tion no initial structure is imposed on the organizasimula-tions They are self- organizing,

starting from the “chaos” (chaotic set of entities) Each organization has a

“cor-ruption parameter (orgcorr),” telling how corrupt or “spoiled” the organization

is The corruption level is calculated as the weighted average of the corruption parameters of all its members The weight is equal to the reciprocal of the entity level in the organization The head of the organization has level 1, its subordi-nates level 2, etc

Political map (PM) This is a one- or multidimensional region, where the entities

are placed The PM should be treated in very general terms It can be just a graphical region or a generalized space of ideas or political orientation For example, in a two-dimensional case, on axis may be the level of corruption (from

geo-2.2 The Model

honest to totally corrupt), and the other may be the political orientation (from democracy to totalitarianism)

PM corruption field (CF) One of the concepts related to the PM implemented

here is the assumption that the political and social ideas are subject to wear What was supposed to be a good idea a hundred years ago is hardly considered good now, due to the corrupted organizations that resulted from its implementation The CF is a function of the spatial variable (position on the PM) that tells how

“good” the spot is It returns zero if the spot is completely spoiled and one if it is

a good spot The value of CF is used by the entities that appear (are born, created)

on the PM. The higher the CF is, the higher is the probability that the new entity occupies the place In other words, the CF defines the probability distribution for the coordinates of new entities

Time The model time is measured in abstract time units (TU) The simulations are

run with final simulation time equal to 2000 or 5000 TU

Entity personal data This is the collection of the following parameters.

Ability This is just the ability to climb in the hierarchy of the organization Note

that such concepts as intelligence or education do not exist in this model, being irrelevant in politics

Lust for power This is the most important entity parameter In other words, the

entity may become a leader if it really wants, which occurs in the real political life

Resources The financial or other resources that help the entity to climb in the

hierarchy

Corruption level Takes values from honest to totally corrupt.

PM coordinates The place the entity takes on the PM. In general, it is the entity

political orientation As stated before, this may be a scalar or a set of nates on multidimensional PM. In this simulation experiments the PM is two

coordi-dimensional and its image on the screen is a square of dimensions MxM.

Lifetime The lifetime determines when the entity dies or just disappears from

PM (natural death) Lifetime is defined as random variable with density tion exp(70.0)

func-Superior The pointer to other entity, the “boss.” The entity is one of the

subordi-nates of the boss

Subordinates Pointers to the subordinates of the entity Each entity can have any

number of subordinates However, for the sake of clarity in the organization images, it is supposed that the entity should have four subordinates So, if the number of subordinates is less than 4, the entity attempts to catch more subordinates

Position on the PM The entity position is expressed by its position in pixels at

the image of the PM

No physical units for the ability, lust of power, resources, and the corruption

level are defined All these parameters are relative, with values in [0,1]

2 Simulating Self-Organization and Interference Between Certain Hierarchical…

2.2.2 Interaction Rules

The active components of this model are the entities An organization is just a data structure and does not take any actions of its own All what happens is the result of the actions of the organization members However, an organization has a self- organizing structure (actions of its members) and behaves as if it had a specific goal: grow and keep growing

The simulation program has been coded using the BLUESSS simulation system

Recall that main concepts of BLUESSS are processes and events A process is a

template, like a class declaration in object-oriented languages At the run time, objects (entities) are generated, being instants of the process declaration Within a process, a series of events are declared The event execution is controlled by the Bluesss system, which invokes events in discrete time instants, according to the clock mechanism and to the internal event queue For more detail, consult http://www.raczynski.com/pn/bluesss.htm

The model includes two processes: entity and monitor Note that the

“organiza-tion” is not represented by any particular process; it is just a data structure So, the organization itself has no “awareness” and does not take any actions The evolution

of organizations is the result of the actions of its members On the other hand, for an external observer, organizations behave as systems with their own goals (to grow and gain power)

Model entities are created by the monitor process After being created, the entity takes a place on the PM, due to a simple rule: the higher is the corruption level on the spot, the lower is the probability the entity will appear there The monitor also initializes three organizations, marking three (randomly chosen) entities as organi-zation heads

The interaction rules are defined by the actions taken by the entities, defined by the following events

Seek for subordinates At the very beginning, only the organization top entities

(heads) seek for subordinates This is done repeatedly, until the entity has gained four subordinates Then, the subordinates start to seek for their subordinates and

so on Any entity that has its superior and less than four subordinates does it (see Fig. 2.1)

Fig 2.1 Organization

structure Head, superiors

(sup), and subordinates

(sub)

2.2 The Model

Die This makes the entity disappear from the PM. The event occurs at the end of

the entity lifetime If the entity was a member of an organization, then one of its subordinates (say X, if any) takes its place A subordinate of X takes the place of

X and so on, iteratively

Climb The entity makes disappear his superior and takes its place A subordinate

of the entity takes its place and so on, iteratively To be able to climb, the sum of the entity lust for power, ability, and resources must be greater than the same sum

of its superior As the entity superior may change, this attempt is repeated every

30 TU, on average (exponential distribution)

Move This is a slow random walk of the entity over the PM. The entity changes

randomly its position with increment [−1,1] pixels This makes the simulation somewhat dependent on the hardware, but does not influence the results signifi-cantly The event is repeated every TU

Propagate The head of each organization propagates his own corruption level to all

members of the organization Each entity changes its corruption level as follows:

This event is repeated each time unit So, the corruption parameter within the organization becomes more uniform

Modify PM The entity changes the local value of the corruption field (CF) The

whole PM region is divided into 900 (30 × 30) square elements, each of them with its corresponding CF value In this event, the factor value is calculated using the following formula:

F = (corruption level level_ / + orgcorr)∗0 04,

where corruption_level and level are parameters in the current entity and orgcorr

is the corruption level of the organization it belongs to So, the entities with lower level value have less influence on the CF value (the head level is equal to 1; subor-dinates have one level less than the superior) The entity repeats this event each 0.5 time units

In such way, some parts of the PM become corrupted The value of the CF is truncated to [0,1] On the other hand, the CF recuperates constantly The monitor process augments the CF in each spot by 0.015, each time unit All this makes the

CF change constantly, depending on how corrupt is the organization that occupies the spot Recall that CF, after being normalized, is used as the probability density function for the appearance of new entities

2 Simulating Self-Organization and Interference Between Certain Hierarchical…

2.3 Simulation

At the beginning of the simulation run, the monitor process is activated It creates

1000 entities randomly located over the PM region For each entity its parameters

are being defined, and the events seek for subordinates, move, modify PM, and climb are invoked The entity event die is scheduled to be executed at actual model time

(when the entity was created) plus the entity lifetime If the entity has disappeared earlier, this event is ignored In the monitor process, the necessary events are initial-ized, like creating organizations (mark the heading entities) organization state dis-play, and the CF recovery The monitor process also stores the model state parameters for further analysis and trajectory plotting Then, all other events are executed auto-matically The organizations grow and entities move and execute their own events Figure 2.2 shows a typical image of the PM at the initial stage (growing organiza-tions) In Fig. 2.3 can be seen the situation after about 500 time units

Organization numbers 1, 2, and 3 are marked with circles, squares, and triangles, respectively Small gray points represent new entities, not affiliated yet The lines are links superior-subordinate The big figure is the organization head, and the size

of the figures decreases for descending entities

The monitor process shows the situation on the PM with small time steps, viding an animated image It is a nice program feature, where the entities move over the area and the “spoiled” and “good” regions change intensity and move

pro-As stated before, the experiments provide only a qualitative information about the model The model behavior is not easy to predict from the specifications of the model components and interaction rules There are some possible scenarios One

Fig 2.2 Initial simulation stage: the PM with growing organizations

2.3 Simulation

Fig 2.3 Dark PM spots, spoiled or corrupted area; white, “good” places

Fig 2.4 The relative size of the organizations as function of time

would expect that the size of the organizations as well as the other variables will change chaotically Other possibility is that one or two organizations will collapse and, after a long simulation time, only one, the strongest “winning” organization, will remain

The experiments show that none of the above occurs After a short initial tory period, the model enters in quite regular oscillations Figure 2.4 shows the size

transi-of the three organizations, in relation to the size transi-of the whole population In Fig. 2.5

we can see the relative size of an organization for longer period of time The shape

of the curves resembles interference between three signals with slightly different frequencies In our model everything is stochastic, so every simulation is different However, this oscillatory nature of the model can always be observed Recalling

concepts of stability of the control theory, the model seems to be orbitally stable

(see Chen 2004)

2.4 Conclusion

The active components of this model are the individuals called entities (the monitor

process is only an auxiliary component) The entities are “alive,” executing their events Though the decisions they take are very simple (where to appear on the PM, climb, etc.), they can be considered as agents of an agent-oriented simulation Both object- and agent-oriented models provide interesting qualitative results, which can

be used as hints while dealing with the reality The main conclusion is that no steady state is reached by the model and that the organizations are in permanent movement This movement, after sufficient simulation time, is oscillatory like the stable cycles

in nonlinear, orbitally stable dynamic systems

Fig 2.5 Relative size of organization 3 after longer simulation time

2.4 Conclusion

The important advantage of such simulations is the possibility of obtaining results that can hardly be reached by other (analytical, sociological) methods For example, how can we see from the model description, without simulating, that the organization size will oscillate with period of about 208 time units (Fig. 2.5)? Another advantage of the tool used here (BLUESSS) is the open nature of the model New events can be easily added to the entity process, reflecting a possible entity behavior and resulting in other, sometimes unexpected, behaviors of the orga-nizations This may be the topic of further research

References

Adamic L, Adar E (2005) How to search a social network Soc Netw 27(3):187–203

Axelrod R (1997) The complexity of cooperation: agent-based models of competition and ration Princeton University Press, Princeton

collabo-Bak P (1997) How nature works: the science of self-organized criticality Oxford University Press, Oxford

Cecconi F, Parisi D (1998) Individual versus social survival strategies J  Artif Soc Soc Simul 1(2):1–17

Chatterjee S, Seneta E (1977) Towards consensus: some convergence theorems on repeated aging J Appl Probab 14(1):89–97

aver-Chen G (2004) Stability of nonlinear systems In: Encyclopedia of RF and microwave engineering Wiley, New York

Cioffi-Revilla C (1998) Politics and uncertainty: theory, models and applications Cambridge University Press, Cambridge

Cohen JE, Hajnal J, Newman CM (1986) Approaching consensus can be delicate when positions harden Stoch Process Appl 22(2):315–322

Deffuant G, Amblard F, Weisbuch G, Faure T (2002) How can extremism prevail? A study based

on the relative agreement interaction model J Artif Soc Soc Simul 5(4)

Edmonds B (2012) Context in social simulation: why it can’t be wished away Computational and mathematical organization theory Comput Math Organ Theory 18(1):5–21

Edmonds B, Hernández C, Trotzsh K (2007) Social simulation: technologies, advances and new discoveries ISBN: 9781599045221

Elsenbroich C (2012) Explanation in agent-based modelling: functions, causality or mechanisms?

J Artif Soc Soc Simul 15(3):1

Epstein JM, Axtell R (1996) Growing artificial societies: social science from the bottom up Brookings Institution Press, Washington, DC

Galam S, Wonczak S (2000) Dictatorship from majority rule voting Euro Phys J B 18(1):183–186 Gotts NM, Polhill JG, Law ANR (2003) Agent-based simulation in the study of social dilemmas Artif Intell Rev 9(1):3–92

Holland JH (1998) Emergence: from chaos to order Helix Books: Addison-Wesley Publishing Company

Krause U (2000) A discrete nonlinear and non-autonomous model of consensus formation In: Elaydi S, Ladas G, Popenda J, Rakowski (eds) Communications in difference equations Gordon and Breach, Amsterdam

Latane B, Nowak A (1997) Self-organizing social systems: necessary and sufficient conditions for the emergence of clustering, consolidation and continuing diversity In: Barnett FJ, Boster

FJ (eds) Progress in communication sciences v.13 Ablex Publishing Corporation ISBN-13: 978-1567502770

2 Simulating Self-Organization and Interference Between Certain Hierarchical…