A DISTRIBUTED AGENCY METHODOLOGY APPLIED

TO COMPLEX SOCIAL SYSTEMS

A Multi-Dimensional Approach

Bogart Yail Márquez, Manuel Castanon-Puga, Juan Ramón Castro

Baja California Autonomous University, Chemistry and Engineering Faculty

Calzada Universidad 14418, Tijuana, 22390, Baja California, México

Eugenio D. Suarez

Trinity University, Department of Business Administration, One Trinity Place, San Antonio, 7821, TX, U.S.A.

Keywords: Social simulation, Complex systems, Distributed agency, Sustainable systems theory.

Abstract: The methodology refers to the forms in which reality and knowledge can be studied; it does not question

knowledge that has been accepted as true by the scientific community but instead concentrates on strategies

to expand the knowledge. This work is motivated by need to establish a methodology for the study of

complex social systems in situations where conventional analysis is insufficient in describing the intricacies

of realistic social phenomena and social actors. The proposed general methodology we describe requires the

use of all available computational techniques and interdisciplinary theories. This growing consensus must

be able to describe all aspects of social life as well as serve as a common language in which different

theories can be contrasted.

1 INTRODUCTION

The objective of our study is to develop a

methodology and corresponding computational

platform that incorporates available mathematical

and computational theories that have not been

appropriately considered in models of complex

social phenomena. Even though applications of

Multi-Agent Systems (MAS) have been developed

for the social sciences have been widely considered

in some areas such as Artificial Intelligence (AI)

(Gilbert, 2007). The modelling of a realistic social

system cannot be achieved by resorting to only one

particular type of architecture or methodology. The

growing methodology of Distributed Agency (DA)

represents a promising research avenue with

promising generalized attributes, with potentially

ground breaking applications in engineering and in

the social sciences—areas in which it minimizes the

natural distances between physical and sociological

nonlinear systems. In this work we thus lay the

foundations for a DA description of socioeconomic

realities, in a process that weaves different available

computational techniques in the context of DA to

represent social and individual behaviour in a

contextualized fashion, accommodating agents with

limited rationality and complex interactions.

We consider a disentangled agent that is formed

by multiple and relatively independent components.

Part of the resulting agent’s task is to present

alternatives, or ‘fields of action’ to its components.

Correspondingly, the composed agent is itself

constrained by a field of action that the

superstructure to which it belongs presents. We

therefore drop customary assumptions made in

traditional social disciplines and MAS about what is

considered a decision-making unit. To arrive at this,

we redefine what a unit of decision is

by unscrambling behavioural influences to the point

of not being able to clearly delineate what the

individual is, who is part of a group and who is not,

or where a realm of influence ends; the boundary

between an individual self and its social coordinates

is dissolved.

The systems are complex entities that represent

a whole that cannot be understood by looking at its

parts independently (Yolles, 2006). The proposed

204

Márquez B., Castanon-Puga M., Castro J. and Suarez E..

A DISTRIBUTED AGENCY METHODOLOGY APPLIED TO COMPLEX SOCIAL SYSTEMS - A Multi-Dimensional Approach.

DOI: 10.5220/0003514402040209

In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 204-209

ISBN: 978-989-8425-54-6

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

intermediate agent can be thought of as a person, a

family, a social class, a political party, a country at

war, a species as a whole, or a simple member of a

species trying to survive. The archetype of the

agents we attempt to describe can be summarized as

a group of colluded oligopolists, such as the oil-

producing countries of OPEC. As a whole, they

share the common interest of jointly behaving like a

monopoly and restricting their production, but they

cannot avoid having an incentive to deviate and

produce above their quota.

Reductionist linear science has concentrated on

the study of entities that are clearly delineated,

where one could separate what belongs to an agent's

nature against the backdrop of what does not. The

relevant agent is taken to be exogenous, and

therefore disconnected from the system to which it

belongs. At their core, these traditional disciplines

are based on a selfish and unitary agent, or atom of

description. Implicitly or explicitly, these paradigms

claim that all aggregate complexity can be traced

back to the lower level of the system: the strategies

and actions of the selfish agent. In other words, these

represent research agendas that purposely de-

emphasize the existence of any level other than that

of the individual.

2 METHODOLOGY

The methodology we are aiming to create represents

a novel approach to simulation architectures,

creating a language that links the social sciences to

programmable terminology and that can thus be

broadly applied. The methodology of DA represents

a general theory of collective behaviour and

structure formation (Suarez, Rodríguez-Díaz et al.,

2007), which intends to redefine agency and reflect

it in multiple layers of information and interaction,

as opposed to the traditional approach in which

agency is only reflected in individual, atomized and

isolated agents (Márquez, Castañon-Puga et al.,

2011). The methodology proposed consists of eight

steps:

2.1 Determining the Levels of Agency

and their Implicit Relationships

We analyse the social system and the existing

relationships to determine the necessary number and

topology of the necessary levels of description. To

this end, we first identify the problems to solve, so

that we can describe their operation within a

physical framework. This process will in turn allow

us to identify the input and output variables of the

system as a whole as well as all necessary

subsystems, whether these variables are decision

variables or other measurable parameters.

In this process, it is imperative that we adopt a

holistic approach that does not attempt to reduce the

system to its individual components. In the real

world, no phenomenon exists in isolation

(Heylighen, 2008). We must therefore in this step of

our methodology establish the objective functions of

all levels of agency that are considered, as well as

the interactions that are prevalent in the

corresponding networks.

2.2 Data Mining

Studies within the social sciences and in particular in

economics are normally performed on a large data

platforms, in situations in which there is too much

rather tan to few data points. The most problematic

aspect of this stage of the modelling process is to

define data sets that match the desired architecture

(Marquez, Castañon-Puga et al., 2010). The

continuous expansion of available information for

social simulation makes the use of data mining

unavoidable. In our case study, for example, we

have access to many different sources of quantitative

and qualitative data describing both economic as

well as sociological aspects of reality. Many of these

data sets are readily available from governmental

sources. Data mining provides us with the process

for extracting implicit information, such as social

patterns, that reveal ingrained knowledge.

2.3 Generating a Rule-Set

We propose the use of a Neuro-Fuzzy system for the

automatic generation of the necessary rule-set of our

simulation. This phase of the data mining process

can become complicated and computationally

intensive, as the Fuzzy system must determine the

necessary number of layers to describe the norms

and variables to keep track of in the simulation. We

further propose using the Nelder-Mead (NM) search

method, which is more efficient than other available

methodologies such as Genetic Algorithms, as has

been shown in multiple investigations (Rantala and

Koivisto, 2002). The NM method is widely used,

mostly because, in general, this optimization

algorithm tends to produce more precise models

with fewer rules (Stefanescu, 2007).

2.4 The Modelling Based on

Distributed Agency

Modelling based on DA allows us to better

A DISTRIBUTED AGENCY METHODOLOGY APPLIED TO COMPLEX SOCIAL SYSTEMS - A Multi-Dimensional

Approach

205

understand the structures and relevant processes of

social systems (Gilbert, 2007). MAS models and

artificial societies are currently built on common

themes, generally using techniques that stem from

dynamical systems modelling, but also using tools

from cellular automaton platforms, genetic

algorithms and DA systems. The difference in

available approaches are normally concentrated in

the idiosyncrasies of the particular model and the

design of the research methodology (Drennan,

2005). The modelling process, the researcher must

build a basic model of the system to be analysed,

where the most important aspects to be represented

are stressed. This can normally be achieved using an

approximate dynamical representation in terms of

stocks and flows in the system, focusing on points

where information a decision making can be

transformed into decision rules. The process of rule

generation that will stem from this original

framework refers back to the previous step (defined

in step 3), and it is focused on the behaviour of the

agents that are influenced by the decision rules in a

probabilistic fashion.

2.5 Implementation

The implementation of the simulation can be done in

vast array of platforms, but the social scientist that

does not want to spend a large amount of time

working on code may simply choose to base this

step in the NetLogo simulation platform, which is

free, readily available, easy to understand and

widely used. Because of its voluminous library, this

platform is ideal for modelling social phenomena. It

is capable of modelling complex systems which can

independently provide instructions to thousands of

interdependent agents operating in a holistic

environment (Wilensky, 1999).

The NetLogo platform also allows us to easily

assign geographical information, that is, by creating

simulation data that represents vectors of

information that include a special component (Vidal,

2007). It is an appropriate platform for the modelling

of a wide range of complex systems that have

temporal dynamics, allowing the researcher to assign

independent instructions to different agents at any

given moment. This relevant characteristic of

NetLogo can provide the researcher with

opportunities for finding the connection between the

micro level of behaviour of a multiplicity of

individuals and the macro patterns of behaviour that

emerge from the interactions of the individuals

(Wilensky, 1999).

2.6 Validating the Model

Real-world simulations must include some form of

validation (Drennan, 2005). This validation will

ultimately reflect the consistency between the real

world and the artificial model. Based on the results

obtained during the validation, the process must then

go back to the beginning, so that the problems found

can be addressed and the model refined.

2.7 A Simulation and Optimization

Experiment

In this phase of the process the researcher must

provide a statistical evaluation of the models outputs

to determine the quality of the simulation based

upon some pre-established evaluation criteria of

performance measurement. As part of this process, it

is important to verify whether the object of interest

reflects seasonal aspects, in which case the data must

be transformed so as to analysed the transformed

stationary data. Finally, in this stage a methodology

for experiment design must be adopted, and it must

be based on repetitions of the simulation performed

with the exogenous variables set at significantly

different levels.

2.8 Analysing the Outputs

In this last stage, the results of the model are

analysed, so that the researcher can understand the

aspects of interest in the behaviour of the system. It

is these outputs and their ultimate understanding that

can then be used to make sense of the social system

in study.

3 MODELING COMPLEX

SYSTEMS IN THE COMPUTER

Computer Simulations can aide in the understanding

of social phenomena, by explaining and predicting

many aspects of its chaotic nature. This

methodology is academically young, but it has been

consistently growing in the scientific field and has

already been used successfully in a number of

research projects (Becker, Niehaves et al., 2005).

Furthermore, available computational techniques can

facilitate the selection of relevant data as well as

aide the processing of information, in processes that

involve high performance computing. It is through

this process that social simulation is developed and

potentially the most efficient way of making sense

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

206

of the vast amounts of information available today

(Rossiter, Noble et al., 2010).

This growing field is intrinsically

interdisciplinary, naturally linked to the sciences of

complexity and to systems theory (Miler and Page

2007). To carry out a useful simulation of a social

system the methodology must be holistic. The

intention is to create a representation that can make

reference to different levels within a given reality

within a general methodology; taking into account

that each level is separated from others in ways that

cannot be described in a reductionist fashion, that

are to some extent in different dimensions and thus

following different rules and temporal granularities

(Suarez, Rodríguez-Díaz et al., 2007). One of the

corollaries of this approach is that an entity that is

represented as a multiplicity of agents in one level

may be considered a unitary agent in another level of

description.

A complex system is composed of subsystems

that may be simple and complex, linear and

nonlinear. Simple, linear systems are in turn

composed of particles and the system. On the other

hand, complex systems require at least three

hierarchical levels: particles, agents and the system

(Halloy, 1999). In a complex system—such as that

of a group, an organization, a growing population or

a market economy, where the self-created

organization comes about from of the interaction of

many component parts—the macro patterns are not

easily discernible or understood from the

understanding of the behaviour of the individual

parts, whether these are simple components, autarkic

agents or rational consumers (Mitchell and Newman,

2002; Ashby, 2004). One of the main challenges of

our approach is to provide a methodology to analyze

the many different levels associated within a social

reality.

The proposition implies that the researcher

observe behaviour, and then use backwards

induction to portray the forces at play that could

have given rise to the decisions taken, as well as

patterns and structures that emerged. Traditionally,

we have begun with a clearly defined agent and tried

to understand its actions as a maximization of

objectives given constraints. In the proposed

paradigm, we assume maximization occurs, and then

work towards the delineation of the benefited entity.

As such, this proposition is not a theory or

hypothesis, but rather a language in which different

models can be expressed. The complexities of the

proposed architecture can be endless. This

notwithstanding, the paradigm for a new pandemic

and inter-disciplinary science built in a distributed

agency architecture would accept the

intercommunion by means of a parsimonious model

that is broad enough to accept the nature of realistic

agents, but at the same time tractable enough for the

capabilities of an appropriate MAS simulation,

expressed at a minimum desired level of realism.

The methodology therefore intends to advance the

development of a common language in which novel

ideas can be transmitted across disciplines.

Such a language allows us to compare a model in

which disentangled humans in a given culture have

some degree of independent agency, but are also to

some degree objects of their social circumstances—

to another one in which countries are trying to

position themselves in the evolving global arena, but

are nonetheless fighting with their internal political

differences, as well as with established international

norms and existing treaties. In sociology, for

example, the individual is ascribed little agency

when compared to the group or social structure;

classical economics, on the other hand, grants zero

agency to upper level creatures, as the selfish actions

of individuals are carried by an invisible hand to an

efficient allocation. As it applies to evolutionary

biology, this distinction represents the core of the

controversy between individual selection theory and

group selection theory. The language of distributed

agency can also serve as a common ground in which

individual vs. group selection theories can discuss

their visions of evolution. Just as the process of

evolution perfects individuals, it must as well have

the same effect in groups and societies. The

surviving members of a cooperating group, however,

will not be ‘fittest’ at an individual level; their

individual traits and natures, for example, only make

sense within the context of the cooperating group.

4 OUR CASE STUDY:

MODELING THE CITY

OF TIJUANA

The principal difference between MAS and our

proposed approach is that in our methodology the

space includes transformations performed by a

higher level of agency. This upper-level agent is

composed of lower-level subcomponents the may

enjoy agency in their own right. It is the

responsibility of this intermediate agent to present its

subcomponents with individual phase-spaces that are

tailored to induce the desired behaviour from the

lower-level agents which inhabit it, when it chooses

according to its own objective function. Therefore,

for our proposed work-in-progress case study, if we

consider a municipality an agent, this upper-level

agent is composed by subcomponents, which in our

case study of the city of Tijuana, Mexico, will be

represented by the AGEBS that compose this city.

A DISTRIBUTED AGENCY METHODOLOGY APPLIED TO COMPLEX SOCIAL SYSTEMS - A Multi-Dimensional

Approach

207

AGEBS is the terminology used to describe the

different areas of the city that are in turn are

composed of neighbourhoods.

In our study we use three levels of agency: the

upper-level agent is represented by the whole city of

Tijuana, the intermediate agents are represented by

the AGEBS and the lower level agents are the

individual inhabitants of the city.

The purpose of this approach is to define poverty in

each level of agency. Originally, when it began to be

measured, poverty was defined in static and one-

dimensional, and was mostly studied in economic

terms, only referring to what where incomes lower

than was considered to be a minimum level

acceptable for society (Akindola, 2009). It is for this

reason that we consider necessary to propose a

concept defined in terms that are dynamical, diffuse

and multidimensional, as recent studies demonstrate

that poverty is not only determined by income levels

but also by the lack of certain non-monetary

resources and opportunities for improvement, such

as education and access to appropriate living

conditions. It is through this light that we want to

analyse poverty—as a multidimensional concept that

cannot be reduced to its individual causes (Akindola,

2009).

Figure 1: Levels of agents represented on the social

system.

In the particular case of the city of Tijuana, the

data set used came from the Instituto Nacional de

Estadística y Geografía (INEGI), the Mexican

governmental organization in charge of gathering

data at a federal level including aspects that are

geographical, socio-demographic and economical.

The data set of the city of Tijuana is divided into 363

areas, known as AGEBS. “The urban AGEB

encompass a part or the totality of a community with

a population of 2500 inhabitants or more in sets that

generally are distributed in 25 to 50 blocks.” (INEGI

2006) For each AGEB we determine the degree of

poverty taking into account 10 income and

employment variables, 23 variables dealing with

education and 15 variables related to the resources

available in the household, such as a television, a

telephone or a refrigerator, among other articles. The

resulting matrix has a total information size of

48x363 variables.

The data sets for this case study were originally

compiled in an information system that is

intrinsically geographical. These systems helped in

the generation, classification and formatting of the

required data—a fact which facilitates the edition of

the different thematic layers of information, in

which one can quantify the spatial structure to

visualize and interpret the areas and different spatial

patterns in Tijuana.

Using the Neuro-Fuzzy system for the automatic

generation of rules, this phase of the data extraction

from the data may become complicated, as the

process needs to appropriately establish the number

of sufficient norms and variables that the study

needs to take into account. Using this grouping

algorithm, we obtain the appropriate rule-set

assigned to each agent representing an AGEB or a

inhabitant of it, the agent receives inputs from its

geographical environment and in turn much choose

an action in an autonomous and flexible fashion

(Wooldridge and Jennings, 1995; Drennan, 2005;

Gilbert, 2007). The purpose of this structure without

central control is to garner agents that are created

with the least amount of exogenous rules and to

observe the behaviour of the global system through

the interactions of its existing interactions, such that

the system, by itself, generates an intelligent

behaviour that is not necessarily planned in advance

or defined within the agents themselves; in other

words, creating a system with truly emergent

behaviour (Botti and Julián, 2003; Russell and

Norvig, 2004). Distributed agents do not necessarily

define agents in lower-levels of description, but

rather consider all levels of agency that are

interconnected in a type of organism that spreads

throughout the system (Suarez, Rodríguez-Díaz et

al., 2007; Suarez, Castañón-Puga et al., 2010).

5 CONCLUSIONS

The methodology we are proposing is developed in a

holistic manner, originally focusing on the

description and interconnection of different levels of

reality, whether these refer to either different

dimensions or different time granularities. The

applications of the approach are ultimately very

general, but they are particularly useful for

interdisciplinary analysis, where different disciplines

overlap or interact in their description of natural or

social phenomena. This general language links

together the developments in computational science

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

208

with those in the social sciences, as they pertain to

the nascent paradigm of complexity. The resulting

methodology represents a powerful alternative for

complementing, substituting or augmenting

traditional approaches in the social sciences. The

study of interdisciplinary connections, of

consilience, and of modelling several levels of

reality jointly remains an area of research with vast

fields of unexplored territory. The growing

disciplines of Computational Social Science and

Social Simulation should be trail blazers in this

effort. (Márquez, Castañon-Puga et al., 2011)

ACKNOWLEDGEMENTS

This work was supported in part by the Mexican

National Council for Science and Technology,

(CONACYT).

REFERENCES

Akindola, R. B. (2009). "Towards a Definition of Poverty

Poor People’s Perspectives and Implications for

Poverty Reduction." Journal of Developing Societies

25: 121-150.

Ashby, R. (2004). "Principles of the self-organizing

system." E:CO Special Double Issue 6: 102-126.

Becker, J., B. Niehaves, et al. (2005). "A Framework for

Epistemological Perspectives on Simulation." Journal

of Artificial Societies and Social Simulation 8.

Botti, V. and V. Julián (2003). Estudio de métodos de

desarrollo de sistemas multiagente. Inteligencia

Artificial, Revista Iberoamericana de Inteligencia

Artificial. 18.

Drennan, M. (2005). "The Human Science of Simulation:

a Robust Hermeneutics for Artificial Societies."

Journal of Artificial Societies and Social Simulation

8(1).

Gilbert, N. (2007). Agent-Based Models. Los Angeles,

Sage Publications Inc.

Halloy, S. (1999) "A Theoretical Framework for

Abundance Distributions in Complex Systems." 16.

INEGI (2006). “II Conteo de Población y Vivienda 2005”.

Instituto Nacional de Estadística Geografía e

Informática.

Márquez, B. Y., M. Castañon-Puga, et al. (2011).

"Methodology for the Modeling of Complex Social

System Using Neuro-Fuzzy and Distributed

Agencies." Journal of Selected Areas in Software

Engineering (JSSE).

Marquez, B. Y., M. Castañon-Puga, et al. (2010). On the

Modeling of a Sustainable System for Urban

Development Simulation Using Data Mining and

Distributed Agencies. 2nd International Conference

on Software Engineering and Data Mining Chengdu,

China, IEEE.

Miler, J. and S. Page (2007). Complex Adptative Systems,

An introduction to computational models of social life.

Mitchell, M. and M. Newman (2002). "Complex Systems

Theory and Evolution." In Encyclopedia of Evolution

(M. Pagel, editor).

Rantala, J. and H. Koivisto (2002). Optimised Subtractive

Clustering for Neuro-Fuzzy Models. 3rd WSEAS

International Conference on Fuzzy Sets and Fuzzy

Systems. Interlaken, Switzerland.

Rossiter, S., J. Noble, et al. (2010). "Social Simulations:

Improving Interdisciplinary Understanding of

Scientific Positioning and Validity." Journal of

Artificial Societies and Social Simulation

Russell, S. and P. Norvig (2004). Inteligencia Artificial.

Un Enfoque Moderno, Pearson Prentice Hall.

Stefanescu, S. (2007). "Applying Nelder Mead’s

Optimization Algorithm for Multiple Global Minima."

Romanian Journal of Economic Forecasting.: 97-103

Suarez, E. D., M. Castañón-Puga, et al. (2010). A Multi-

Layered Agency Analysis of Voting Models. 3rd

World Congress on Social Simulation Kassel,

Germany.

Suarez, E. D., A. Rodríguez-Díaz, et al. (2007). Fuzzy

Agents. Soft Computing for Hybrid Intelligent Systems.

O. Castillo, P. Melin, J. Kacprzyk and W. Pedrycz.

Berlin, Springer. 154: 269-293.

Suarez, E. D., A. Rodríguez-Díaz, et al. (2007). Fuzzy

Agents. Soft Computing for Hybrid Intelligent Systems.

O. Castillo, P. Melin, J. Kacprzyk and W. Pedrycz.

Berlin / Heidelberg, Springer. 154.

Vidal, J. M. (2007). Fundamentals of Multiagent Systems

with NetLogo Examples.

Wilensky, U. (1999). "NetLogo Software." from

http://ccl.northwestern.edu/netlogo.

Wooldridge, M. and N. Jennings (1995). "Intelligent

Agents: Theory and Practice." Knowledge Engineering

Review.

Yolles, M. (2006). Organizations as Complex Systems: An

Introduction to Knowledge Cybernetics. Managing the

Complex. Greenwich, Connecticut, USA, Information

Age Publishing. 2: 866.

A DISTRIBUTED AGENCY METHODOLOGY APPLIED TO COMPLEX SOCIAL SYSTEMS - A Multi-Dimensional

Approach

209