CUE-BASED DECISION-MAKING IN CONSTRUCTION WORK CREWS:
AN AGENT-BASED MODELING APPROACH
By
Ali Lahouti

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Construction Management – Doctor of Philosophy
2017

ABSTRACT
CUE-BASED DECISION-MAKING IN CONSTRUCTION WORK CREWS:
AN AGENT-BASED MODELING APPROACH
By
Ali Lahouti
Construction work crews often deal with ambiguities (e.g., mismatch between provided directives
and existing situations) at the work face about ‘what-to do’. In such a circumstance, the work crew
is forced to interpret cues present at the work face surroundings to make a decision. This research
argues that dynamics of a construction work crew, as a collection of two or more skilled trade
workers who are continuously engaged in face-to-face interactions influence its interpretation of
work face cues. Dynamics of a construction work crew may be adapted to the group Input-ProcessOutput (I-P-O) framework where inputs such as characteristics of each member enable and/or
mediate a group’s process like communication, and interaction to generate outputs – e.g., a
decision.

A population of one hundred and seventy six (176) construction skilled trade workers were
qualitatively interviewed: (i) to learn their most-immediate actions in a situation where directives
did not match the work face condition and while physical hints were present; and (ii) to identify
input factors they experienced to be influential on the process of decision-making and their
interactions to determine ‘what-to do’. Contents of these interviews were clustered in categories
of implied descriptions, 69 percent of which suggested that these work crews were more eager to
autonomously resolve the ambiguity and most-likely by discussing the issue with their crew-mates
instead of contacting their supervisors (25%), or customers (5%). Contents of these interviews also
revealed that knowledge and skill of crew members, their confidence and trust in others crew-

mates, and their open-mindedness were discussed most frequently as impactful factors on decision
making and interactions between members.

In order to understand impacts of these input factors on interactions between work crew members
and their interpretations of cues, an Agent-based Modeling Simulation (ABMS) was developed.
The abstract ABM (in NetLogo environment) simulated a work crew of two members with
uniformly-distributed levels of skillfulness, confidence, and open-mindedness who were to
complete an activity by making decisions on initiating a series of tasks. The crew operation time,
amount of rework it conducted, and frequency of its call-for-assistance were measured for 199,948
cycles and were analyzed with respect to input factors. Skillfulness was identified as the most
influential factor on operation time. It was also realized that all three outputs measures were most
sensitive to input factor skillfulness; and that each output measure was almost equally influenced
by variation of the factors confidence, and open-mindedness.

The research dissertation concluded that variations (or rather heterogeneity) of skillfulness,
confidence, and open-mindedness in work crews would increase intensity of interactions between
members, and would, thus, decrease the likelihood of expected optimum operation time, no
rework, and less frequency of call-for-assistance.

The core of this research dissertation offers a sound basis for construction work crew recruitment
by identifying knowledge and skills as an intuitive requirement while screening for additional traits
of confidence, and open-mindedness. It also provides an explanation to better understand ‘why’
the Last PlannerÂŽ System (LPSÂŽ) of Project Planning and Production Control is instrumental in
maintaining a reliable, and predictable workflow for a construction operation.

Copyright by
ALI LAHOUTI
2017

To my late father, Hassan Lahouti,
who taught me education is not as much about learning facts as is about training minds to think;
and
To my loving, and caring mother, Fatemeh Lahouti,
the cornerstone of my life, who by example taught me it is not what one gathers but scatters in
life that will tell the story of life he or she lived.

v

ACKNOWLEDGEMENTS

Teachers enable their students to learn by enlightening them to understand principles, and by
indulging them to experience facts. And I do believe that those principles, and facts are not only
lessons one learns in classrooms but also the morals he or she takes away from each encounter.

I am very fortunate to have been taught by a thoughtful teacher, a careful scholar, and on top of all
an honorable friend: Dr. Tariq S. Abdelhamid, Associate Professor of Lean Construction. As words
simply will fall short in articulation, thanking him for his academic advisement, his professional
leadership, and his friendly mentorship is merely acknowledging the value of standing,
unconditional support this gentle man has offered me – specifically during inevitable, and
sometimes unfortunate events of life. I hope one day I can pay it forward.

I wish to offer a sincere thank you to my advising committee faculty: to Dr. Patricia L. Crawford,
Associate Professor of Landscape Architecture, who taught me that each of our endeavors has its
own merits. Whether it was a classroom discussion, a research meeting, or a professional
conversation, her actions, and reactions always encouraged the ‘worthiness’ in me; to Dr. Arika
Ligmann-Zielinska, Associate Professor of Geography, Environment, and Spatial Sciences, who
by example taught me that teaching is not filling others with information but releasing your own
knowledge; and to Professor Timothy L. Mrozowski, Professor of Construction Management who
taught me patience in our long, spirited debates.

I am very thankful to many of my colleagues in Infrastructure Planning and Facilities at Michigan
State University; in particular to my supervisor Mr. Dean D. Geisenhaver who educated me with
the practical knowledge of construction operations; and to Alteration and Improvement (A&I)
vi

crew who enthusiastically shared with me their years of experience to shape the platform for this
research dissertation.

And last, I wish to offer my most special thank you to Dr. Mahmood M. Moallemian, who has, in
difficult moments of my life, stood by me as a true friend; supported me like a parent; and advised
me as a mentor. I am, for life, in his debt.

vii

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................................ x
LIST OF FIGURES .................................................................................................................... xii
KEY TO SYMBOLS.................................................................................................................. xiv
(1)
Introduction ....................................................................................................................... 1
(1.1) Overview ......................................................................................................................... 2
(1.2) Significance of problem .................................................................................................. 2
(1.3) Problem Statement .......................................................................................................... 4
(1.4) Goal and Objectives ........................................................................................................ 7
(1.5) Dissertation Research Outline ......................................................................................... 8
(2)
Background and Literature Review .............................................................................. 10
(2.1) Definition of a Group .................................................................................................... 12
(2.2) Conceptual Reference Framework for a Group ............................................................ 14
(2.3) Variables in a Group ..................................................................................................... 17
Data Collection for Variables in a Group .............................................................. 19
Structured and Qualitative Interviews .................................................................... 20
(2.4) Construction Operation and Variation in its Resources ................................................ 21
Incompleteness of Instruction and Implicitness of Information .............................. 24
Incompleteness/Implicitness of Information/Instruction and Cue Interpretation ... 28
(2.5) Agent-based Modeling Simulation (ABMS)................................................................. 31
(2.6) Summary ....................................................................................................................... 33
(3)
Research Method ............................................................................................................ 35
(3.1) Objective One – Research Approach ............................................................................ 36
(3.2) Objective Two – Research Approach ............................................................................ 38
(3.3) Objective Three – Research Approach .......................................................................... 39
Agent-based Model Simulation ............................................................................... 40
Agent-based Model Analysis ................................................................................... 43
(3.4) Summary ....................................................................................................................... 46
(4)
Data Collection and Analysis ......................................................................................... 47
(4.1) Construction Work Crew and Group ............................................................................ 48
Construction Work Crew ........................................................................................ 48
Construction Work Crew as a Group ..................................................................... 49
(4.2) Conceptual Reference Model for Dissertation Research .............................................. 50
viii

(4.3)
(4.4)

(4.5)

Research Findings: Analysis and Discussion ................................................................ 52
Autonomous Problem-Solving ................................................................................ 56
Statistical Analysis ........................................................................................................ 63
Examination for Association ................................................................................... 64
Analysis of Differences ........................................................................................... 66
Summary ....................................................................................................................... 72

(5)
Agent-Based Modeling Simulation (ABMS) ................................................................. 74
(5.1) Agent-based Model Information ................................................................................... 75
(5.2) Agent-based Model Presentation .................................................................................. 76
Agent-based Modeling Environment....................................................................... 79
(5.3) Agent-based Model Simulation ..................................................................................... 80
(5.4) Agent-based Model Result ............................................................................................ 85
(5.5) Agent-based Model Analysis ........................................................................................ 88
Exploratory Analysis ............................................................................................... 88
Sensitivity Analysis (SA) ......................................................................................... 99
(5.6) Summary ..................................................................................................................... 108
(6)
Conclusion ..................................................................................................................... 109
(6.1) Discussion ................................................................................................................... 112
(6.2) Contribution ................................................................................................................ 119
(6.3) Limitation .................................................................................................................... 122
(6.4) Future Research ........................................................................................................... 123
APPENDICES ........................................................................................................................... 125
APPENDIX A: Framework for Qualitative Interview ............................................................ 126
APPENDIX B: Agent-based Model ........................................................................................ 128
BIBLIOGRAPHY ..................................................................................................................... 134

ix

LIST OF TABLES

Table 2.1 – Foci of Survey in Existing Literature ........................................................................ 12
Table 2.2 – A Sample of group Definitions Focused on Concept ................................................ 12
Table 2.3 – A Sample of Characteristics Focused in Group Definitions ...................................... 13
Table 2.4 – Strengths & Weaknesses in Sources of Evidence ...................................................... 20
Table 2.5 – Literatures Recognized Incomplete Kit Effect on Construction Project Performance
....................................................................................................................................................... 25
Table 3.1 – Summary of Dissertation Research Plan ................................................................... 36
Table 4.1 – Population Characteristics of Interviewee Construction Crews ................................ 53
Table 4.2 – Work Crews Most-Likely Immediate Actions in an Ambiguous Situation............... 54
Table 4.3 – Influencing Factors on Construction Work Crews Interactions ................................ 60
Table 4.4 – Influencing Factors on Construction Work Crews Interactions with Respect to Trade
....................................................................................................................................................... 65
Table 4.5 – Influencing Factors on Construction Work Crews Interactions with Respect to Crew
Type .............................................................................................................................................. 66
Table 4.6 – Fisher's Exact Statistical Significance Test Result: Influential Factors of Greater
Frequencies for Self-Performing Work Crews ............................................................................. 68
Table 4.7 – Fisher's Exact Statistical Significance Test Result: Most-likely Immediate Actions 70
Table 4.8 – Fisher's Exact Statistical Significance Test Result: Influential Factors of Greater
Frequencies for Specialty Sub-Contractor Work Crews ............................................................... 72
Table 5.1 – Essential Components of Developed ABMS ............................................................. 75
Table 5.2 – Presentation of Essential Components of Developed ABMS .................................... 76
Table 5.3 – Probability Distribution Function Introduced for Influential Factors ........................ 78
Table 5.4 – Frequencies of Influential Factors with Respect to Crew Type................................. 83
Table 5.5 – Apportioned Frequencies of Influential Factors with Respect to Trade and Crew Type
....................................................................................................................................................... 84

x

Table 5.6 – Statistical Descriptive Measures for ABMS Outputs Variables ................................ 87
Table 5.7 – Frequency Analysis of ABMS Output: Number of Neighbors Visited by Two Agents
....................................................................................................................................................... 93
Table 5.8 – Sensitivity Indexes for Variation of Output: Number of Neighbors Visited ........... 100
Table 5.9 – Guideline for Interpretation of SimLab Sensitivity Analysis Output ...................... 101
Table 5.10 – Sensitivity Indexes for Variation of Output: Rework ............................................ 104
Table 5.11 – Sensitivity Indexes for Variation of Output: Call for Assistance .......................... 105
Table 6.1 – Frequency Analyses of Influencing Factors on Work Crew Members Interactions 119

xi

LIST OF FIGURES

Figure 1.1 – Category of Factors Influencing Interactions between Group Members ................... 5
Figure 1.2 – Conceptual Representation of Group Decision-Making ............................................ 6
Figure 1.3 – Schematic Process Map for Cue Interpretation in a Construction Work Crew .......... 7
Figure 2.1 – Conceptual Presentation for Mutually Interacting Components of a Work Crew ... 15
Figure 2.2 – Input-Process-Output (I-P-O) Conceptual Framework ............................................ 16
Figure 2.3 – Conceptual Reference Framework for Improvisation .............................................. 26
Figure 2.4 – Conceptual Reference Framework for Cue Interpretation ....................................... 29
Figure 3.1 – Schematic Presentation of Dissertation Research Plan ............................................ 37
Figure 3.2 – Graphical Presentation of Developed ABMS........................................................... 42
Figure 3.3 – Graphical Presentation for Analysis Reference Value ............................................. 44
Figure 4.1 – Construction Work Crew Input-Process-Output (I-P-O) Conceptual Framework ... 51
Figure 4.2 – Frequency Analyses of Work Crews Immediate Actions ........................................ 55
Figure 4.3 – Conceptual Framework for Work Crews Autonomous Problem-Solving ............... 57
Figure 4.4 – Frequency Analyses of Work Crews Autonomous Problem-Solving Heuristics ..... 58
Figure 4.5 – Conceptual Heuristic of Cue Interpretation in Construction Work Crews............... 59
Figure 4.6 – Frequency Analyses of Influencing Factors on Work Crew Members Interactions 61
Figure 4.7 – Comparison of Influential Factors with Greater Frequencies for Owner Selfperforming Work Crew ................................................................................................................. 67
Figure 4.8 – Comparison of Immediate Actions to Ambiguous Situations with Respect to
Construction Work Crews Types .................................................................................................. 70
Figure 4.9 – Comparison of Influential Factors with Greater Frequencies for Specialty Subcontractor Work Crews ................................................................................................................. 71
Figure 5.1 – Graphical Presentation of Developed ABMS........................................................... 79
Figure 5.2 – Conceptual Framework for Development of an Abstract Agent-Based Model ....... 80
xii

Figure 5.3 – Decision-Making Process for Agent i in Group of n Agents ................................... 81
Figure 5.4 – Linear Correlation between Influential Factors Confidence & Trust and Openmindedness .................................................................................................................................... 85
Figure 5.5 – Variation in Number of Neighbors Visited with Respect to Group Size ................. 90
Figure 5.6 – Graphical Presentation for Analysis Reference Value ............................................. 92
Figure 5.7 – Skillfulness Level for Two Agents with Respect to Simulation Cycles Outputs ..... 94
Figure 5.8 – Number of Neighbors Visited with Respect to Skillfulness Level for Two Agents 95
Figure 5.9 – Number of Neighbors Visited for Skillfulness Level ‘D’ with Respect to Confidence,
& Open-mindedness...................................................................................................................... 97
Figure 5.10 – Number of Neighbors Visited for Skillfulness Level ‘C’ with Respect to Confidence,
& Open-mindedness...................................................................................................................... 98
Figure 5.11 – Independent Contribution of Input Factors to Outputs Variations: Number of
Neighbors Visited ....................................................................................................................... 102
Figure 5.12 – Contribution of Input Factors Interactions to Outputs Variations: Number of
Neighbors Visited ....................................................................................................................... 102
Figure 5.13 – Comparison of Interactions Intensity of Input Factors with Respect to Outputs
Variables ..................................................................................................................................... 105
Figure 5.14 – Variations in Sensitivity Indexes with Respect to Number of Simulation Cycles 107
Figure 6.1 – Average Operation Time and Variation Performance with Respect to Crew Size 113
Figure 6.2 – Comparison of Independent Contributions of Input Factors with Respect to Outputs
Variables ..................................................................................................................................... 116
Figure 6.3 – Schematic Presentation of Iterative Process of Cue Interpretation ........................ 118
Figure 6.4 – Overall-Frequency Analyses of Work Crews Immediate Actions in Ambiguous
Situations..................................................................................................................................... 120

xiii

KEY TO SYMBOLS

Symbol

Description and Representation

Data

Information that enters (i.e., input) to be processed in a
system; result that leaves (i.e., output) as consequences of a
system conduct, and process

Decision

A conditional step; an emerging point for inputs

Process

An action; a task is performed

Start / End

SubProcess

A system initiation; a process termination

A sequence of steps – that itself is a system

Transfer of temporary control from a unit to next; passage
of state between consecutive units

xiv

(1)

Introduction

1

(1.1)

Overview

Construction work crews often face ambiguities while completing an activity. These ambiguities
may be caused by incomplete construction drawings, missing/incorrect/misinterpreted
information, an implicit verbal instruction, etc., which do not provide an ideal state that this
dissertation research refers to as “perfect information”. When a work crew does not receive
complete instruction and perfect information, that work crew tends to gather, look for, and/or
leverage physical hints (which this thesis will specifically refer to as cues) existing at the location
of work in order to make sense of and augment the incomplete instruction and imperfect
information.

This investigation will focus on the situation when a construction crew encounters ambiguities
during the execution of work on the job site. The conjecture of the research dissertation is that the
construction work crew will make use of the existing physical hints (cues) to augment the missing
information. The cue can be deceiving and may present a challenging process for the work crew
members. This challenge intensifies the interaction among crew members to proceed with the
work; and the interaction in turn impacts the interpretation. How all these dynamics eventually
impact work execution, from the view-point of time and cost, is not well understood and
(definitely) is not quantified with respect to resource utilization.

(1.2)

Significance of problem

Resources (e.g., work crew; material; etc.) are utilized in delivering any construction activity.
When interpretation of a cue(s) causes deviation from expectation, and therefore the anticipated
outcome is not delivered due to an error(s) or rather misinterpretation, then more resources – at

2

least twice as much – will have to be employed to accomplish the same activity correctly. In the
general sense, this over-use/re-use of resources can at the least include:

(a)

removing wrong work installed right is a waste of resources. It is a necessary task to be
done and, yet, of no value as it was not planned to exist;

(b)

rework can be recognized as another consequence. A biased interpretation of an
instruction and making a decision based on a readily-available cue may result in an
undesired – i.e., wrong or incomplete – outcome which will require a do over. This
exceeding quantity of resource utilization can at least introduce as three times budgetover run and two times task duration delay for what is initially planned for this task –
let alone its impact on over-all schedule performance;

(c)

delay may be another consequential waste. When a cue from the surroundings is used
in decision-making and a wrong thing is conducted (i.e., an undesired outcome), delay
can be experienced either due to unavailability of required work to the successor task
or due to some schedule slippage as a result of mandated re-work.

When this concept is associated with construction project management productivity performance,
the importance of sufficient information and the impact of cue interpretation are very critical as
the cost of re-work due to absence of explicit instructions is magnified.

3

(1.3)

Problem Statement

One: When a work crew encounters a physical hint or a cue, that work crew practically interprets
the cue to fill in for the missing instruction and then makes a decision about ‘what-to’ do (Lahouti
& Abdelhamid, 2012). This decision on ‘what-to’ do may not align with the original intent.
Therefore, the work crew may ‘do the wrong thing right’: (i) it is the ‘wrong thing’ because cueinterpretation can interrupt the alignment between the ‘what-to’ that the crew ends up completing
and that the original instruction intended; and yet (ii) it will be done ‘right’ because the work crew
is (expected to be) appropriately skilled to deliver the ‘what-to’ that the complete original
instruction intended.

On the note of doing the wrong thing right, it is important to state that the problem is not about
technicality of the ‘what-to’; it is not about how to execute; it is not about application of rules and
laws; it is not about the craftsmanship; it is not about equipment break-down and/or availability;
it is not about planning. In fact, the problem is about the final description of a ‘what-to’ do; it is in
the search by a crew to define the task at the location of work immediately before it is executed; it
is the fine-tuned description of ‘what-to’ do based on what makes sense on the job site.

The point of departure from previous researches (Lahouti & Abdelhamid, 2012; Lahouti, 2013) is
the consideration of work crew size. Lahouti and Abdelhamid (2012), and Lahouti (2013)
discussed interpretation of cues and physical hints on construction a job site with the assumption
of a one-member work crew. This research dissertation releases this assumption and considers a
work crew of more than one members.

Two: In a construction project, activities/tasks are performed in small group settings. Therefore,
each work crew may be considered a representation of a small group. Group consultation, or rather
4

interaction, is a common characteristic of a small group. This common, instinctive behavior can
be prompted in a small group by variety of factors. As illustrated in Figure 1.1, interactions in a
group may be influenced in some fashion by families of individual (e.g., experience; skills), group
(e.g., hierarchy; size), activity (e.g., complexity; difficulty), and environment (e.g., strategic
location) factors.

individual
characteristics

environment
characteristics

Group
Interaction

group
characteristics

activity
characteristics

Figure 1.1 – Category of Factors Influencing Interactions between Group Members
Adopted from Source: (Forsyth, 2014, p. 408)

A decision is an example of an immediate output of a small group interaction mediated by these
categories of factors. As conceptually presented in Figure 1.2, members of a construction work
crew, an example of a small group, interact to plan the tasks required for completion of their
assigned activity, reach consensus on tasks to conducted, and make a decision(s) as to initial action.

5

Work Crew
Member
(n)

Concensus

Decision

Interaction

Work Crew
Member
(n-1)
Interaction

Figure 1.2 – Conceptual Representation of Group Decision-Making

Combination of claims One and Two can be stated as:
In a construction project, when a work crew receives incomplete instructions about ‘what-to’ do,
it prompts the interaction(s) between members of that work crew for two apparent reasons: (i)
there are pieces of information missing in the instruction; and (ii) there is (almost always) a
physical cue(s) present at the job site – which may substitute for the incompleteness of instruction.
As Figure 1.3 depicts, the output of this interaction(s) is interpretation of the cue(s). A decision,
which is based on cue-interpretation, will then define the ‘what-to’ that will be executed on the job
site. Such may define the relationship between group and its task. That is, how a group responds
to its activity following the interactions between its members. It should be noted that interaction
refers to processes such as information acquisition, information processing, decision making, etc.;
and that since interactions between members of a work crew are influenced by different factors, it
is rational to also claim those factors influential on interpretation of cues and physical hints.

6

Work Crew
Process

Work Crew
Member
(n-1)
Missing
Information
/ Cue

Interpretation
of cues/hints

Interaction

Decision

Work Crew
Member
(n)
n≥1

Figure 1.3 – Schematic Process Map for Cue Interpretation in a Construction Work Crew

(1.4)

Goal and Objectives

The overall goal of this research is to better understand the impact an incomplete instruction and
an imperfect piece of information will impose on production performance on a construction job
site. More specifically, it aims to understand how interactions between work crew members may
influence interpretation of cues and physical hints on a construction job site; and to investigate
whether a relationship exists between interactions of work crew members and interpretation of a
cue when the work crew receives an incomplete instruction and imperfect information. To
approach this goal, the following achievable objectives were followed:

(1)

a questionnaire was developed to identify factors that influence interactions between
work crew members, and its outputs;

7

(2)

a strategy was developed to introduce the factors identified in Objective (1) into a
computer simulation package and in order to form a model and simulate interaction
among construction work crew members;

(3)

an Agent-based Modeling Simulation (ABMS) was developed, and simulation results
were analyzed to determine whether a correlation existed between work crew
interaction and interpretation of a cue: key outcomes (i.e., amount of time spent; rework
generated) were measured.

(1.5)

Dissertation Research Outline

This dissertation research is divided into six (6) chapters:

In Chapter One the research problem was stated, its significance was discussed, and objectives
were proposed in order to approach the investigation goal.

In Chapter Two literature on topics of Small Groups, Group Dynamics, and Group Performance
are surveyed: the term Group is defined, its properties discussed, and a reference conceptual
framework for How Group Works is reported. The literature on construction project research is
also reviewed on information and instruction – which were concluded to be necessary resources
for construction activities in project operations. It is further discussed that lack of information and
instruction would encourage construction work crews to improvise and utilize available resources
in their surroundings, and to interpret cues and physical hints in order to fill the missing pieces of
information; to make a decision and proceed; and technically to make-do.

8

In Chapter Three methods by which this dissertation research was conducted are explained: how
factors influential on work crew members’ interactions when interpreting incomplete instructions
were identified; and how such interactions and decision-making were simulated using Agent-based
Modeling Simulation (ABMS).

Chapter Four begins with a discussion on whether characteristics of a construction work crew
would satisfy group’s criteria; and how group’s Input-Process-Output framework would fit a work
crew. Then, contents of qualitative interviews with construction work crews are analyzed, and
influential factors are reported as a portion of dissertation research findings.

In Chapter Five outputs of the Agent-based Modeling Simulation (ABMS), developed based on
findings reported in Chapter Four, are analyzed and reported as a portion of dissertation research
findings.

Chapter Six concludes this investigation by re-stating the problem of interest, and its objectives,
reviewing the need for this research with respect to existing academic literatures, and summarizing
and comprehensively discussing the findings of research methods. Limitation of research conduct,
contributions of findings to body of construction-industry (academic and professional) knowledge,
and opportunities for further development of this work are also deliberated.

9

(2)

Background and Literature Review

10

As stated in section (1.3), this dissertation research suggested that a construction work crew, when
faced with ambiguity, would make use of cues and physical hints available and/or existing in its
surroundings to augment the missing information. The main concepts in this statement – as Table
2.1 summarizes in two sections – are (1) a construction work crew, and (2) decision-making in a
construction work crew.

The speculations about these two main concepts are: (i) that a construction work crew is a group;
(ii) that such challenges (of ambiguity) intensify interactions between members of this work crew;
and (iii) that these interactions, in turn, impact interpretation of the situation (i.e., decisionmaking). The following research questions were considered to investigate these claims:

(1)

Does a construction work crew meet required criteria and qualifications to be a group?

(2)

How does a construction work crew function from the perspective of a group?

(3)

What independent factors (i.e., initial states and mediators) do influence a construction
work crew performance? And which of those influencing factors are critical on a work
crew’s performance in presence of cues and physical hints?

(4)

How does completeness of information and explicitness of an instruction, or its lack
thereof, play a role in executing an activity in a construction project operation at its
crew level?

(5)

How may correlation between interpretation of a cue and interaction of a construction
work crew members be investigated?

11

Table 2.1 – Foci of Survey in Existing Literature
Section One:
Construction Crew as a Group

Section Two:
Decision-Making in a Construction Crew

Question (1)  Group Characteristics

Question (4)  Group Performance

Question (2)  Group Functions

Question (5)  Group Analysis

Question (3)  Group Variables

Section One: Construction Crew as a Group
Question One: Does a construction work crew meet the required criteria/qualifications to be
addressed as a group?
Answering this very question requires a general understanding of a group – and what a group is.

(2.1)

Definition of a Group

Social Scientists – those who study groups – have not agreed on a comprehensive description of
the term group. However, they have defined it with respect to varying interests of their specific
research. That is, no unique approach has been employed to describe the concept group. In fact,
some have defined a group based on its characteristics as a unit. Table 2.2 presents two of such
definitions.

Table 2.2 – A Sample of group Definitions Focused on Concept
Definition

Source

“A dynamic whole based on interdependence”

(Lewin, 1948, p. 184)

“An intact social system … interdependence for some shared purpose”

(Hackman & Katz, 2010, p. 1210)

12

While some researchers have expressed a group “at a great length”, others have found it more
appropriate to define a group by emphasizing “one (or a few) of the many characteristics” it possess
(Shaw, 1981, p. 4). In other words, researchers have chosen certain properties of interest and have
studied groups accordingly. In this latter approach, in contrast with the former one, focus is on
constituent elements of a group as the unit of analysis. As summarized in Table 2.3, for example,
commonality of a goal, a norm, and perception among members of a group are the main
characteristic which has been emphasized by some researchers.

Table 2.3 – A Sample of Characteristics Focused in Group Definitions
Focus of Definition

Commonality of goals, norms, and perceptions

Interaction, and interdependence

Relationship

Source
(Smith, 1945, p. 227); (Newcomb, 1951, p. 3);
(Mills, 1967, p. 2); (Cartwright & Zander, 1968, p.
46); (Mabry & Barnes, 1980, p. 4); (Foley &
MacMillan, 2005, p. 21); (Forsyth, 2014, p. 400)
(Bales R. F., 1950, p. 33); (Cartwright & Zander,
1968, p. 46); (Merton, 1968, p. 339); (Crosbie,
1975, p. 2); (Shaw, 1981, p. 8); (Shaw, 1981, p. 8);
(Jackson, 2010, p. 317); (VandenBos, 2015, p. 471)
(Brodbeck, 1958, p. 2); (McDavis & Harari, 1968,
p. 237); (Davis, 1969, p. 4); (McGrath, 1984, p. 8);
(Syer & Connolly, 1996, p. 7); (Forsyth, 2014, p. 4)

In another instance, several researches have selected interaction – which is a form of
interdependence (Cartwright & Zander, 1968; Shaw, 1981) – as the criterion of investigation; and
have stressed that in their descriptions of a group. Or many social scientists have focused attention
upon relationship as an essential characteristic to describe a group. Although size is a single
characteristic that is unanimously emphasized, these definitions each signifies one (or a few) of

13

many different characteristics in a group which has attracted social scientists, and explicitly asserts
that a group is not a collection of arbitrary individuals.

Taken together, these characteristics provide essential attributes for a group in alignment with the
interest of this dissertation research. It is therefore proposed that the term group refers to:

(a)

a collection of two or more individuals;

(b)

who are functionally related;

(c)

who continuously engage in a direct (i.e., face-to-face) interaction(s); and

(d)

who purposively perform towards a common objective(s);

Note that members of a group are “functionally” related for they are members of a specific
organization which is supposed to last for a certain period of time – e.g., co-workers on a particular
construction job site and for a specific project.

Question Two: How does a construction work crew function from the perspective of a group?

(2.2)

Conceptual Reference Framework for a Group

In aforementioned diverse definitions for term group alongside with its characteristics, individuals,
their relationships, their interactions, their objectives, etc. are [constitutional] elements of a group;
and their varying properties exist at the time a group is formed. These elements can be referred to
as initial states of a group (Hackman & Morris, 1975), and in part or as a whole facilitate
interaction between its members.

While influenced by initial states, the interaction between group members imposes impact on
initial states, and a cycle takes place. This cycle emerges to a product (e.g., an action; a completed
14

activity; a decision; etc.) and represents performance of the group. While the performance is a
variable depending on initial states and interactions, it itself will also affect both. These as a whole
shape an interconnected system whose components are [classes of] variables which are everevolving, ever-influencing, interdependent, and interrelated. What adds more to this complexity is
the simultaneous involvement of these variables in each cycle while a group functions. This
“complex of mutually interacting components” (Arrow, McGrath, & Berdahl, 2000; Forsyth, 2014,
p. 407) is conceptually presented in Figure 2.1.

Individuals

Work Crew
[Structure]

Activity
/ Task

Interaction

Environment

Performance

Figure 2.1 – Conceptual Presentation for Mutually Interacting Components of a Work Crew
Adapted from Source: (McGrath, 1964, p. 70)

In order to understand such a complex system and explore such an “ambiguous set of concepts”
(McGrath, 1984, p. 12) it indeed is useful to create a model which serves as a guiding reference: a
layout which depicts its various components; and a framework which represents underlying logics
of this phenomena. A widely known reference framework for a group in Organizational Behavior
literature is illustrated in Figure 2.2.
15

Group
Characteristics

Individual
Characteristics

Mediator

Input

Activity/Task
Characteristics

Process

Output

Environment
Characteristics

Figure 2.2 – Input-Process-Output (I-P-O) Conceptual Framework
Adapted from Source: (McGrath, 1964, p. 70; Hackman & Morris, 1975, p. 50; McGrath, 1984, p. 13)
and (Stock, 2004, p. 278; Forsyth, 2014, p. 408)

In accordance with this conceptual model, a group may be represented as a linear, casual chain
where:

(a)

independent initial states are aggregated into Input – these factors feed into a group, and
set the stage for members’ collective work. Those independent initial states which
indirectly affect the interrelated Input-Process cycle and either strengthens or weakens it
are presented as Mediators;

(b)

interdependent actions among members of the group are introduced as Process – it is
technically what takes place in a group while members utilize resources and collectively
perform an activity/a task to deliver a meaningful output; it is the relationship between
Input and Output; and

(c)

emergences from those two constructs are categorized in a dependent Output – that is a
tangible result of a group’s activities.

16

It is important to note that a group functions in manner much complicated than a simple causeand-effect perspective may be able to capture. And for that matter, there are several critiques on
this system. For example, it may not seem realistic to introduce components of such a complex,
dynamic system as variables since what takes place in a group is mostly an emergent state(s); it is
not sufficient enough to present a dynamic, complex system as a series of sequential status where
a successor starts only when its predecessor completes its duties; there is an absolute need for a
feedback cycle; etc. (Ilgen, Hollenback, Johnson, & Jundt, 2005; Forsyth, 2014).

Question Three: What independent factors (i.e., initial states and mediators) do influence a
construction work crew performance? And which of those influencing factors are critical on a
work crew’s performance in presence of a cue and a physical hint?

The conceptual Input-Process-Output (I-P-O) framework of Figure 2.2 is to systematically present
how a group functions: when at least two individuals form a group to deliver an output. It is
expected that in this I-P-O paradigm inputs influence the outputs through process. In other words,
the initial states of a group enable the group to process and deliver an output. Social Science
literature focused on groups dynamics introduced different families of factors as such enablers.

(2.3)

Variables in a Group

The conceptual Input-Process-Output (I-P-O) framework of Figure 2.2 is to systematically present
how a group functions: when at least two individuals form a group to deliver an output. It is
expected that in this I-P-O paradigm inputs influence the outputs through process. In other words,
the initial states of a group enable the group to process and deliver an output:

17

Morris (1966), Hackman (1968), and Hackman & Morris (1975) stated that characteristics of a
task have a significant influence on group process. These authors found that type of a task
contributes to group member involvement and offered a strong impact on group output; and that
difficulty level of a task influenced the output of a group to a lesser degree.

Kent & McGrath (1969) found that task type and group gender composition were among strong
determiners of a group performance characteristics.

Katzell et al. (1970) focused on a group process and its output; and identified three (3) input
parameters leadership, difficulty of a task, and compatibility of group members to be of great
importance. These authors argued that presence of leadership increased interaction among
members of a group and decreased demands for information both of which improve output of
group process. They also claimed that difficulty of a task contributes to agreement rate among
group members which results in a better output. These authors also stated that compatibility of
members improved productivity of group process and consequently impacted group output.

Stock (2004) focused on group processes; and found that leadership, characteristics of group
members, and group interdependence were among input parameters which positively influenced
collaboration, communication, and cooperation [i.e., process] in a group.

There are numerous initial states or inputs which have been introduced in social science literature
and have been identified of important effects on a group process and, in turn, its output(s); and
depending on focus of the investigation and what aspect(s) of a group was at stake in the study,
some initial states were found to be more useful than others. It is therefore crucial to understand
whether a relationship exists between a particular input factor variable and an output of a specific
18

situation. Arguably, input factors and initial states which were collected during previous researches
may not be rightly applicable to a different study on group performance because each researcher
collected data and introduced input factors in accord with her/his own specific purpose for the
study, and consequently such data would need to be verified for the new research environment.

Deploying existing input factors in this investigation would have offered major challenges: (i)
choosing from input factors for such widely studied phenomenon requires empirical verification –
which would mean correlating output to input factors; or deriving input factors from output instead
of input factors emerging outputs – and (ii) empirically verifying factors demand experimenting
different what-if scenarios – neither of which would have been effective nor efficient; and could
have potentially offered a false presentation of originality. These signify the importance of
identifying influential input factors on a work crew interactions and consequently its outputs for a
specific scenario – presence of cues and physical hints in case of this research dissertation.

Question: How data can be collected about real-life experience of construction work crews?

Data Collection for Variables in a Group
One source of data can be direct observation; however, it may not be efficient as it is not known
when the instance under investigation will occur. Another source of data can be interviews with a
group (e.g., a work crew) members and asking them about their experiences. Any commonlyknown English dictionary defines ‘Interview’ as ‘a meeting of people face to face especially for
consultation’ in its verb form and as ‘question (someone) to discover their opinions or experience’
in its noun form. In academic research, however, interview is defined more specifically as ‘verbal
and body language’ data gathered about a specific matter – e.g., another person’s explanation of

19

some behavior or action; a recollection (Yin, 2011). As Table 2.4 summarizes common strengths
and weaknesses of an interview, it may be concluded that no source of data collection offers a set
of complete advantages over any other one(s):

Table 2.4 – Strengths & Weaknesses in Sources of Evidence
Adapted from Case Study Research: Design & Methods (Yin, 1994, p. 80)
Source of Evidence

Strengths

Weaknesses

Biased induced due to poorly
Targeted: it focuses directly on
constructed questions may
the topic of investigation.
lead to errors.
Insightful: it provides perceived
Response bias
causal inferences.
Interview
Inaccuracies resulting from
poor recall
Reflexivity: interviewee gives
what interviewer wants to
hear.

Structured and Qualitative Interviews
An interview can be conducted in two fashions (Yin, 2011): (i) a structured interview in which
group (e.g., a work crew) members will be given a questionnaire or a list of specific questions.
The questionnaire of a structured interview – also known as a survey – will force the participants
to ‘respond’ to a uniform set of questions. This will represent a close-ended discussion.

A structured interview is favorable in an investigation whose intent is to limit the responses to a
set of pre-defined categories, and basically to ask a close-ended question (Murphy, 1980). In such
a procedure, the pre-defined responses [for a question] are often offered for the purpose of
corroboration. For the matter of such distinctive nature, structured interviews are rarely able to
20

deal with the context of social life; and are the least common in evaluating ‘attitude, orientations,
circumstances, and experiences’ (Babbie, 2010).

(ii) a qualitative interview in which the conversation and discussion will be based on a conceptual
framework of the topic instead of a set of specific and uniform questions. In such an approach, an
interviewee is encouraged to not only be a ‘respondent’ but also an ‘informant’ about the matter
of investigation. In this open-ended question-and-answer the participant(s) is considered an
‘informant’ since the relative pieces of information will emerge from the discussion and in
interviewees’ own words: how he/she lived a specific situation; and what her/his key take-aways
were.

A qualitative interview may be considered an exploratory tool which is to depict a complex social
world from a participant’s perspective; and to examine the reasons for complex events. It focuses
on understanding the lived experience of other people and the meaning they make of that
experience – that is, ‘what’ actually occurred, ‘why’ it happened, and ‘how’ the process was
operated; the ‘decision’ that was made; and the ‘impact’ such an experience and decision poses
(Murphy, 1980; Yin, 2011).

Section Two: Decision-Making in a Construction Crew
Question Four: How does completeness of information and explicitness of an instruction, or its
lack thereof, play a role in a construction project operation at its crew level?

(2.4)

Construction Operation and Variation in its Resources

The construction project performance literature has typically focused on productivity performance
measures in a construction operation and reported many factors of crucial influence on productivity
21

performance when they deviated from the plan; and because they deviated from what was expected
from them. Whether due to a random and undesired event (e.g. equipment breakdown) or a nonrandom and controllable situation (e.g., unclear objectives) that deviation caused disruption, then
resulted in work flow variation, and thus hindered productivity – (Koskela, 1992); (Halligan,
Demsetz, & Brown, 1994); (Howell & Ballard, 1997); (Kaming, Olomolaiye, Holt, & Harris,
1997); (Chua, Kog, & Loh, 1999); (Radosavljević & Horner, 2002); (Thomas, Horman, de Souza,
& Zavrski, 2002); (Liberda, Ruwanpura, & Jergeas, 2003); (Wambeke, Hsiang, & Liu, 2011);
(Menches & Chen, 2014).

Variation has many dimensions (e.g., output; resource; time; etc.) and is a common phenomenon
in almost any construction project – even those which are planned and managed well. In the
Construction Industry, variation generally refers to deviation from an expectation – (Rilett, 1998);
(Ballard & Howell, 1998; Howell, Ballard, Tommelein, & Koskela, 2004; Wambeke, Hsiang, &
Liu, 2011); (Tommelein, Riley, & Howell, 1999); (Koskela, 1992); (Abdelhamid, 2011);
(Wambeke, Hsiang, & Liu, 2011). In manufacturing, operation variation is defined as a random
deviation, as a consequence of an event which is not under immediate control (Hopp & Spearman,
2008).

Lean Thinking in a broad scope, and Lean Production and Lean Construction as specific domains
recognize the devastating influence of variation on work flow (Womack & Jones, 2003; Hopp &
Spearman, 2008), and focus on reducing, if possible eliminating, variation in a construction
operation work flow by ensuring that required resources for an activity are available at the right
time. The Last PlannerÂŽ System (LPSÂŽ) of Project Planning and Production Control is one of the
Lean Construction techniques which emphasizes that forming a quality assignment for a work

22

crew can reduce variation in work flow in an operation (Ballard & Howell, 1998; Ballard, 2000;
Pikas, Sacks, & Priven, 2012).
The Last PlannerÂŽ System (Ballard, 2000) suggested that a construction task be shaped based on
four (4) quality criteria:

(i)

Definition – an activity should be explicitly described so that beginning and end of the
work can be identified; the appropriate amounts of resources can be considered; etc.

(ii)

Soundness – an activity requires its pre-requisite activities/work to be completed; every
required resource is available; etc.

(iii)

Sequence – an activity should to be scheduled to start according to project delivery
commitments; consistent with the logic of work execution; etc. and

(iv)

Size – an activity should create only amount of work aligned with production capability of
a work crew; achievable in plan available time; etc.

In manufacturing operations, Ronen (1992) referred to adequate and complete resources as the
“Complete Kit” – a set of every requirement that an operation needs in order to be completed. The
concept of Complete Kit suggests that an operation should not start until each and every
requirement for its completion is available. The Last PlannerÂŽ System soundness criteria can be
inferred as an equivalent of the Complete Kit since it is to ensure that equipment, information,
material, personnel, etc. necessary for an upcoming activity are available (Ballard & Howell, 1998;
Koskela, 1999; Ballard, 2000; Koskela, 2004; Formoso, Sommer, Koskela, & Isatto, 2011).

23

Incompleteness of Instruction and Implicitness of Information
A construction project is a collection of activities which focus on a specific output (Koskela, 2000);
a set of directive-driven activities which are often interdependent and interrelated and are designed
to form certain physical outcome (Howell & Ballard, 1997; Ballard & Howell, 1998). An activity
requires a set of resources to start; it consumes these resources in order to progress; and to produce
a physical outcome at the right time. External conditions, directives and instructions, materials,
personnel, etc. are among such resources.
A directive refers to “the information or instructions required for the construction crew to start the
activity” – e.g., submittals; construction specifications; shop drawings; request-for-information
(Abdelhamid, 2011). Prior research suggested that insufficient (i.e., variation in) “information”
and implicit “instruction” adversely impact productivity performance of a construction project.
For example, Ballard & Howell (1998), Chua et al. (1999), Dai et al. (2009), and Formoso et al.
(2011) argued that errors in shop drawings, inadequacy of construction specifications, or
inadequate information about project design, plans, and procedures created waste, resulted in
rework, and ultimately hindered project productivity performance. Kaming et al. (1997), Koskela
(2004), Lahouti & Abdelhamid (2012), and Arashpour et al. (2013) discussed that missing and/or
unclear instruction brought about rework; that work redone was a waste and resulted in a longer
execution time and cost. Ronen (1992), and Koskela (2004) referred to lack of information and
need for more information as an Incomplete-Kit syndrome which may create rework and reduce
productivity, and can eventually increase cost in an activity cycle and ultimately preclude
completion of the planned activity (Handa & Rivers, 1983; Dozzi & Abourizk, 1993; Chua, Kog,
& Loh, 1999).

24

Some other literature which recognized information and instruction as influential factors on
construction project performance (e.g., cost; productivity; rework; safety; waste; etc.) are reported
in Table 2.5.

Table 2.5 – Literatures Recognized Incomplete Kit Effect on Construction Project Performance
Factor

Reference

Insufficient information

(Mohsini & Davidson, 1992)

Information [variation]

(Thomas & Sakarcan, 1994)

Poor instruction

(Kaming, Olomolaiye, Holt, & Harris, 1997)

Missing/unstable information and instruction

(Koskela, 2000; Koskela, 2004)

Information [variation]

(Thomas, Horman, de Souza, & Zavrski, 2002)

Non-availability of information

(Liberda, Ruwanpura, & Jergeas, 2003)

Shortage of information

(Shohet & Frydman, 2003)

Lack of information/inadequate instruction

(Dai, Goodrum, & Maloney, 2009)

Unavailable/unclear/incomplete information

(Formoso, Sommer, Koskela, & Isatto, 2011)

Lack of information

(Hamzeh, Morshed, Jalwan, & Saab, 2012)

Incomplete instruction/imperfect information

(Lahouti & Abdelhamid, 2012; Lahouti, 2013)

Lack of direction & information

(Menches & Chen, 2014)

Starting an activity with incomplete resources is technically starting an activity with whatever is
available and to improvise and to proceed with the means at hand. According to Ciborra (1999)
improvisation is a natural human instinct and plays an “important role” when an element in a
process falls short; it is a greatly contextual and absolutely situational action which cannot be
planned before it actually happens. As cited by Cunha et al. (1999), improvisation is the ability to
effectively adapt to the unforeseen; it is efficiently planning with resources to match the
unexpected situation; and it is “making do with materials at hand” (p. 307). Cunha (2005) also
defined improvisation as using whatever is readily available to make the best out of a situation

25

when there is a problem; to make the best out of limited available resources in a moment when an
unanticipated event happens. Hamzeh et al. (2012) defined improvisation as the ability to come up
with a resolution for an unexpected issue and only utilizing resources that are available.
Koskela (2004) referred to this improvisation as a “making-do”; a situation in which a construction
activity starts or continues with either unavailable or non-optimal and non-standard inputs. A
making-do is indeed a response to the disruption which was caused by inadequacy of resources
intended to accommodate and overcome the interruption induced by insufficiency of resources.
Ronen (1992) argued that senses of obligation to improvise and to start an activity (or a portion of
it) even with an incomplete kit – as if the activity will be completed earlier – were due to an urge
of maximum resources utilization assuming it increases the overall productivity and “the fallacious
notion that a worker should be busy all the time causes managers to have their people working
using incomplete kits just so that they should not be idle” (p. 2459).

As Figure 2.3 conceptualized, the acts of improvisation and making-do are (examples of) how the
construction work overcomes the disruption, and determines what-to do:

Missing
Information

Incomplete
Kit

Improvisation
& Making-Do

Figure 2.3 – Conceptual Reference Framework for Improvisation

For example, Formoso et al. (2011) introduced “lack of information” among the main causes of
improvisation and making-do in a construction project operation; and stated that when a
construction work crew faced an ambiguous situation, and when there was an uncertain and/or
26

unexpected situation at the work face, the work crew had a tendency to employ any resources [e.g.,
information] which was readily available in its surrounding to reach the desired objectives.

Pikas et al. (2012) stated that when a planned construction activity was not sound with accord to
the Last PlannerÂŽ System (LPSÂŽ), it technically meant that the construction work crew was in need
for more information. Therefore, the work crew evaluated its surroundings, gathered readily
available information, and improvised and did make-do in order to fulfill its objective.

Hamzeh et al. (2012) discussed that when an activity could not start due to its unavailable
prerequisite resources such as perfect information and a complete directive, the construction work
crew would improvise and would utilize available information in order to execute that activity.
The authors further argued that even employing the Last PlannerÂŽ System as a Lean Construction
tool might not eliminate improvisation, or interpretation, or making-do because focus of the LPSÂŽ
would be on availability of “inputs” and “prerequisites”, it could not anticipate unexpected events
– thus improvisation and making-do would help construction work crew to manage the situation.
Menches and Chen (2013), and Menches and Chen (2014) introduced “lack of information and
direction” as an unexpected event which would disrupt construction project work flow; and as a
trigger for improvisation and in-situ decision-making in a construction project operation. The
researchers argued that “disruption” and “improvisational decision-making” were naturally built
in any construction project because unexpectedness was an element of a construction project; and
that because a construction project was a series of “plan-driven” operations, any disruption
regardless of its type invited the work crew to improvise and to make a decision in order to
overcome the situation and to resume execution of the plan.

27

Although recognizing that improvisation is itself a disruption, and that it can hinder the progress
of an operation and/or a project, Eden et al. (2000) argued that improvisation must be done so the
crew can “work-around” the disruption; otherwise further and perhaps other types of disruption
would occur. For example, Pikas et al. (2012) reasonably claimed when a construction work crew
decided to exercise making-do and continued the progress: (i) the flow of work might eventually
stop because the work could not be completed which would add no value and could even cause
further disruptions; (ii) the work would be done and progress would be made but under sub-optimal
process which would deliver partial value and product might not be as optimal as expected and
might require rework; and (iii) work could be completed under an optimal process and full value
would be delivered.

Incompleteness/Implicitness of Information/Instruction and Cue Interpretation
Interpretation of cues differs from the act of improvisation. With reference to the reviewed
literature in previous section (2.4.1), improvisation in construction operation is to make the best
out of incomplete, limited, available resources and to come up with a solution in the spur of the
moment. Cue interpretation as conceptually illustrated in Figure 2.4 is, by contrast, to take a
potentially deceptive solution that is already present at the work face, and fill the incomplete piece
of information.

28

Physical
Hint / Cue

Missing
Information

Incomplete
Kit

Interpret Cue
/ Physical Hint

Figure 2.4 – Conceptual Reference Framework for Cue Interpretation

Cues are physical hints; they are existing elements of a construction job site – which can potentially
misguide the work crew under specific circumstances; and they are NOT generated solutions. In
other words, cues and physical hints appear to offer complementary information to the work crew
members, yet fool their users. As Lahouti and Abdelhamid (2012), and Lahouti (2013) argued in
order to restore the continuity of operation the construction work crew directs its efforts to deploy
the most likely resolution by taking advantage of existing cues; and by interpreting the most
relevant physical hints at work face surroundings to fill in for the missing information and
instruction.

Question: What is the gap in existing literatures which this research dissertation uncovers?

The aforementioned Construction-Industry literatures addressed improvisation, and make-do as
approaches work crews would take to deal with an incomplete kit. Employing existing cues and
interpreting physical hints at the work face, however, was not discussed as other approaches for a
construction work crew to autonomously deal with ambiguity of the incomplete kit. Neither was it
considered that dynamics of work crews might influence improvisations, and make-do’s as well
their outputs. This research dissertation, thus, aims to fill in this gap by uncovering how

29

interactions between members of a construction work crew correlates with usage and interpretation
of cues when it faces incomplete directives at the work face.
With reference to Section (2.2) – Figure 2.1 and Figure 2.2 – the acts of improvising and makingdo, gathering available pieces of information and interpreting of cues as outputs of work crew
function may be well influenced by dynamics of their conductors – i.e., member of the work crew;
yet no literature to the extent surveyed by this author – addressed such correlation.

The focus of discussion in this exploration is on the operation level of a construction project, and
it is important to note that improvisation, making-do, and/or interpretation of a cue occur in all
stages of a construction project (Formoso, Sommer, Koskela, & Isatto, 2011). For example, Hsu
and Liu (2000) claimed that in the early design phase of a construction project, the design team
would have to use a large amount of incomplete and imprecise pieces of information in its decision
making process. And as such information could lead to different and perhaps conflicting
improvisations and interpretations. Gomes et al. (2014) argued that the design team would spend
much time to reason through different interpretations, to reach a common understanding of the
matter, and to make-do.

Question Five: How may correlation between interpretation of a cue and interaction between
construction work crew members be investigated?
As was discussed in Section (2.2) – and with reference to Figure 2.1 – a group is a complex,
interdependent, and interrelated system with constituents which are heterogeneous entities
carrying sets of unique characteristics and distinctive behaviors. Interactions between these
constituents are dynamic processes which cannot be truly perceived in advance and before they

30

occur – for it will be a product of several interactions within this complex system (Ahn, 2014). As
pertains to research of this dissertation, for example, the influences that such interactions may
impose on interpretation of cues; and that whether correlation exist between interpretation of
physical hint and work crew interactions would not be apparent in advance as outputs of these
ever-evolving interactions are emerging results, cannot be drawn from expectations, and ought to
be observed per each scenario as a bottom-up process.

Observation is not always feasible. On the one hand, deliberately altering values and manipulating
attributes of the process will be – if possible at all – a challenging task especially when it comes
to observing human behavior. On the other hand, as Watkins et al. (2008) claimed, “direct
observation” of events were often time consuming, and cost-inefficient; and as Sterman (1992, p.
4) added, in situations where such constraints did not apply people “even experts, have great
difficulty inferring accurately the behavior of complex dynamic systems.” After all, as Simon
(1957, p. 198) argued, “the capacity of the human mind for formulation and solving complex
problems is very small compared with the size of the problem whose solution is required for
objectively rational behavior in the real world or even for a reasonable approximation of such
objective rationality.” Computer modeling and simulation approaches are helpful tools in
overcoming some of these issues. Oreskes et al. (1994) argued that “the reason for modeling is a
lack of full access, either in time or space, to the phenomena of interest”.

(2.5)

Agent-based Modeling Simulation (ABMS)

Among computer simulation and modeling tools, Agent-based Modeling Simulation (ABMS) can
represent dynamics of a construction project. Being a collection of interacting agents (Castiglione,
2006) in which each entity and its interactions may be directly presented (Ligmann-Zielinska,
31

Agent-Based Models, 2010), an ABMS would be able to represent complex behavior and to reveal
information about the dynamics of the system it represents (Bonabeau, 2002; Fortino & North,
2013).

An ABMS is constructed of two key components: a set of agents and an environment. Agents are
heterogeneous entities which share a common environment, and autonomously interact with one
another and also their underlying environment in order to represent a real-world system. Each
agent owns unique characteristics, is bounded by a set of rules that govern its decision-making,
and follows a specific set of actions each time it autonomously makes a decision (Bonabeau, 2002;
Sawhney, Walsh, & Mulky, 2003; Ligmann-Zielinska, 2010; Du, 2012).

ABMS has become popular in Construction Industry research (Sawhney, Walsh, & Mulky, 2003;
Desai & Abdelhamid, 2012; Du, 2012; Lahouti & Abdelhamid, 2012; Moore, 2013; Ahn, 2014)
for perhaps advantages it offers over other computer modeling and simulation approaches. For
example, Streman (1992), Banks et al. (2002), Bonabeau (2002), Srbljinovic & Skunca (2003),
and GalĂĄn et al. (2009) discussed that ABMS captures emergent phenomena: an emergent
phenomenon is an unpredictable and often ever-perpetuating resultant of a complex, dynamic
system. The unique, nonlinear, and unpredictable interactions between members of a construction
work crew that generate outputs (e.g., a decision) are instances of such phenomena.

ABMS provides a flexible framework for simulation: this research dissertation at large aimed
exploring impacts of interactions between members of a construction work crew on its decisionmaking. This decision-making may be driven by a series of what-if scenarios in a crew operation.
Simulating these what-if’s mandates deliberate alteration, and manipulation of key components of

32

the ABMS by changing, for instance, population sizes for agents, adjusting rules for agents
interactions, modifying agents attributes and characteristics, etc.
ABMS is explicit, and creates a natural description of a system – i.e., it represents a system from
view-points of its constituents (i.e., bottom-up) instead of the processes that system follows (i.e.,
top-down). In example of a construction work crew, a more natural and realistic model defines
operations of this work crew based on behaviors, and characteristics of its members and
interactions between them (i.e., a bottom-up process), and therefore allows for outputs to emerge.
In contract, a top-down model defines behaviors, and characteristics for members of this work
crew based on their outputs.

In conclusion, ABMS seems an appropriate framework for modeling and a simulating method in
understanding complex dynamics of construction projects, and specifically in crew operation level.

(2.6)

Summary

In this chapter exiting literature was surveyed with the purpose of answering research questions:
First, the definitions for a group, its characteristics, and how it functions were reviewed. It was
discussed that variety of factors and initial states would play roles in conceptual Input-ProcessOutput (I-P-O) framework of a group; that these factors would often be case-specific; and that
deploying previously-identified factors in any other disciplines (e.g., the Construction Industry
research) might not serve the purpose the investigation.

Second, literatures was reviewed to learn about studies on improvisation in construction project
operations and in presence of imperfect information and incomplete instruction, and to solidify the
topic of this dissertation research by determining the existing gap.
33

Last, agent-based modeling was discussed to as an appropriate simulation tool for ConstructionIndustry research for it is an explicit, flexible framework which can capture nonlinear,
unpredictable, and emergent phenomena.

34

(3)

Research Method

35

This dissertation research, as stated in Section (1.3), postulates that interactions between members
of a construction work crew influence the interpretation of cues on a construction project job site
where directives and/or pieces of information are missing and/or incomplete. In order to
investigate this problem and to verify its impacts, three objectives were proposed in Section (1.4).
This Chapter outlines a plan to satisfy these objectives in Five (5) phases.

As summarized in Table 3.1, in Phase One existing relevant literatures were surveyed. Actual and
real-life data was collected in Phase Two and its analyses were presented. In Phase Three a
conceptual model of the Problem Statement was developed to provide a reference framework for
Phases Four and Five in which a (computer) simulation was created, a series of experiments were
designed, and corresponding outputs were analyzed.

Table 3.1 – Summary of Dissertation Research Plan
Phase Task

(3.1)

Product

One: Review & summarize literature



Objectives One & Two

Two: Collect data & conduct statistical analyses



Objective One

Three: Model a conceptual framework



Objective Two

Four: Simulate the conceptual model



Objective Three

Five: Conduct (simulation) experiment & analyze outputs



Objective Three

Objective One – Research Approach

Stated in Section (1.4) and presented in Figure 3.1, the first objective of this dissertation research
was developing a framework to identify factors which influence interaction among members of a
construction work crew and its outputs.

36

Figure 3.1 – Schematic Presentation of Dissertation Research Plan

A series of qualitative interviews approved by Michigan State University Institutional Research
Board (IRB X15-377e) were conducted with a population of construction skilled trade workers
who were selected based on their availability. Anonymous information about disciplines of their
trades, and number of years they had been in a specific discipline was gathered, and the entire
population was introduced to an identical real-life scenario. That scenario was then followed up
by two (2) major questions of interests to this dissertation research in form of a non-structured,
conversation-like discussion:

Question One: How does a construction work crew act in/react to a situation where it is to make
a decision in presence of cues, and physical hints? That is, what her/his most-likely, immediate
action in such situation is.

37

Question Two: What are factors which will influence interaction among a construction work crew
members when it conducts that situation in presence of cues and physical hints?

The framework for qualitative interviews is presented in APPENDIX A. In their responses to these
open-ended inquiries, each work crew member expressed her/his opinions and shared her/his
experiences in similar situations. The contents of these responses were qualitatively analyzed and
catch-phrases were identified with regards to each of follow-up questions. In an inductive research
approach, then, these catch-phrases were clustered into mutually exclusive categories
corresponding to immediate actions (i.e., Question One) and influential factors (i.e., Question
Two). In the next step, contents of the interviews were statistically analyzed and frequencies of
repetitions were determined for each category. Finally, the most-frequently repeated categories
were identified to be critical factors and were considered in the conceptual model and –
consequently – the computer simulation.

In short, Objective One aimed to: (a) understand heuristics of a construction work crew conduct
when it encounters an ambiguous situation at the work face; and (b) identify factors influencing
these heuristics.

(3.2)

Objective Two – Research Approach

The second objective in this dissertation research, as Figure 3.1 reports, was to develop a model
which conceptually demonstrates interaction between members of a construction work crew when
a crew was to interpret cues, and physical hints at the work face. In order to deliver this objective,
first existing academic literature on Small Groups, Group Dynamics, and Group Performance
were reviewed mainly to:

38

(i)

define a group;

(ii)

identify essential characteristics of a group;

(iii)

learn about important factors which influence a group process and its outputs; and

(iv)

understand how a group (e.g., a construction work crew) functions.

A construction work crew was examined from perspective of a group, and was evaluated whether
it met a group’s criteria.

The literature was also surveyed for a reference framework that conceptually depicted a group
operate – so a general understanding of group’s process was achieved and a framework could be
developed accordingly to represent processes in a construction work crew and to conceptually
model the interactions between its members.

Second, incorporating lessons learned from analyses of qualitative interviews contents (for
Question One in Objective One) the reference framework of Small Groups proposed in academic
literature was adapted to conceptually represent how a construction work crew would conduct the
specific situation of interests in this dissertation research – i.e., interpretation of cues, and physical
hints. This model was further developed to fulfill Objective Two when integrated with factors
resulting from analyses of qualitative interviews contents for Question Two in Objective One.

(3.3)

Objective Three – Research Approach

This research dissertation deployed an Agent-based Modeling Simulation (ABMS) in order to
explore answers for the following (Question Three):

39

Question Three: What correlation exists between influential factors and construction work crew
performance in presence of cues and physical hints? Or between influential factors and
construction work crew performance when its members are to interpret cues and physical hints –
i.e., work crew’s output?

Three (3) major steps were designed and executed to achieve this objective.

Step One was to simulate actions of a construction work crew towards interpretation of cues and
physical hints on a job site.

Agent-based Model Simulation
The conceptual model developed in Objective Two was simulated in programmable environment
of NetLogo. As asserted in its User Manual, NetLogo is a “modeling environment for simulating
natural and social phenomena” based on Java virtual machine. It is a ‘simple-enough’ yet
‘advanced-enough’ tool for simulating complex systems; and can “give instructions to hundreds
or thousands of agents all operating independently. This makes it possible to explore the
connection between the micro-level behavior of individuals and the macro-level patterns that
emerge from their interaction” (Wilensky, 1999).

In order to simulate a construction operation in which a work crew exercises cue-based decisionmaking, an abstract model graphically presented in Figure 3.2 was designed:

Work Crew
In this maze, the blue circle symbolized a construction work crew of n members. This crew was to
complete an activity.

40

Construction Operation
Each activity in a construction operations is a collection of several construction tasks. The white
squares in this maze represented a set of tasks for completing the activity ‘reach the red square’ –
the activity was to reach red square through the shortest path.

Cue and Physical Hints
Black squares represented a collection of obstacles to generate cues and physical hints.
Combination of black and white squares created forks and cross-paths – which implied ‘more than
one shortest path’. The blue circle was to decide on correct shortest path.

Decision-Making
The decision making was represented by the blue circle choosing one white square over the other
one at a fork/cross-path.

41

Figure 3.2 – Graphical Presentation of Developed ABMS

Step Two was to design an experiment to execute the ABM simulation. Software SimLab (version
2.2) was employed to generate a population of input values. SimLab uses Monte-Carlo methods to
generate pseudorandom numbers based on a set of probability distributions functions. A uniform
distribution of U (0,1) was defined for each input factor not only for the matter of simplicity but
also because (specific) qualities of those factors were neither available nor within the investigation
scope.
The ‘A* Shortest Path’ algorithm was adapted to this ABMS and was employed by agents in search
of the shortest path to red square neighbor.

Step Three was to execute the ABMS for the designed experiment for a reasonable number of
cycles, and to analyze its outputs. A series of exploratory statistical analyses were conducted with

42

software R (version 3.3.3); and sensitivity of this ABMS model to its inputs values was evaluated
with software SimLab (version 2.2).

Agent-based Model Analysis
Exploratory Statistical Analysis
The statistical computations on ABMS outputs explored distributions of three (3) dependent
variables with respects to independent input factors:

(i)

Number of neighbors visited representing operation time;

(ii)

Number of neighbors re-visited representing rework and waste; and

(iii)

Number of neighbors dead-end representing need for more directives, and instructions.

Analysis Reference Value
To understand how each set of simulation outputs scattered with respect to input factors of interest,
reference thresholds were introduced for values of input factors. As illustrated in Figure 3.3, these
thresholds divided output values into four regions/quadrants.

43

Figure 3.3 – Graphical Presentation for Analysis Reference Value

The output values (e.g., number of neighbors visited) which were populated in:

(i)

Quadrant A were resulted from values greater than 0.50 for each agent’s input factors
of interest (e.g., skillfulness);

(ii)

Quadrant B were resulted from values greater than 0.50 for input factors of interest for
one agent, and values lesser than 0.50 for input factors of interest for the other agent;

(iii)

Quadrant C were resulted from values lesser than 0.50 for each agent’s input factors
of interest; and

(iv)

Quadrant D were resulted from values lesser than 0.50 for input factors of interest for
one agent, and values greater than 0.50 for input factors of interest for the other agent.
44

Then, Probability Density Functions (PDF) were constructed for the model outputs to evaluate
how (at large) the model behaved with respect to input factors values. Plotted for values in each
quadrant of input factors of interest, PDF curves compared the likelihood that a particular output
value would result from different values of that input factor. In other words, PDFs were constructed
to explore how the ABMS responded to uncertainty of input factors.

Variance-based Sensitivity Analysis (SA)
In general term, sensitivity analysis is to prorate variations in model outputs to its difference inputs.
Variance-based measure is among the known tools for sensitivity analysis in which variance of
outputs are decomposed into model components. Saltelli, et al. (2008) favored this technique for
models of “unknown linearity, monotonicity, and additivity” (p. 37); and consider this method as
model-independent, and capable to “capture the influence of the full range of variation of each
input factor” (p. 158). In addition to those, Ligmann-Zielinska and Sun (2010) argue that because
inputs in Monte-Carlo-based simulations (e.g., SimLab) are distributions of values, “it is only
logical to utilize variance as a single-value measure that reflects the uncertainty of the outputs”.

The mathematical representation of variance decomposition, adapted from Saltelli, et al. (2008),
and Ligmann-Zielinska and Sun (2010), defines a global variance V to be the summation of local
variances:
𝑉 = (∑𝑘𝑖 𝑉𝑖 ) + ( ∑𝑘𝑖<𝑗 𝑉𝑖𝑗 ) + (∑𝑘𝑖<𝑗<𝑚 𝑉𝑖𝑗𝑚 ) .............................................................. Equation 3.1

where the first term represents the independent contribution of each k input factor in variations of
outputs; the second, and the third terms express the variations of outputs due to interactions
between k input factors. Sum of these terms accounts for total contribution of k input factors in
45

variation of outputs – i.e., the left-hand side in

Equation 3.1. The sensitivity indexes are

calculated based on these variances, and as presented in the following equations (Saltelli, et al.,
2008; Ligmann-Zielinska & Sun, 2010):

First-order Sensitivity Index (Si) ................................................................................... Equation 3.2

𝑉𝑖 𝑉𝑋𝑖 [𝐸𝑋−𝑖 (𝑌|𝑋𝑖 )]
=
𝑉
𝑉 (𝑌 )
Total-effect Sensitivity Index (STi) .............................................................................. Equation 3.3

𝑉 (𝑌) − 𝑉𝑋−𝑖 [𝐸𝑋𝑖 (𝑌|𝑋−𝑖 )]
𝑉 (𝑌 )
where i (1, 2, 3, … , k) represents each input.

Software SimLab was employed to calculate these indexes; and results of were interpreted with
respect to scope of this research dissertation.

(3.4)

Summary

The research approaches discussed in this chapter illustrate how the three (3) major questions of
this investigation were to address. These approaches comprised a plan of Five (5) Phases in which
existing literatures were reviewed; real-life data was collected and analyzed; a conceptual and a
computer simulation model were developed, experiments conducted, and conclusions made.

46

(4)

Data Collection and Analysis

47

(4.1)

Construction Work Crew and Group

As stated in Section (1.4) – Goal and Objectives – this dissertation research aims to better
understand how interaction between construction work crew members may influence
interpretation of cues on a construction job. Towards this understanding, it is natural to ask what
happens in a work crew when there is a cue present, and how such dynamic influences
interpretation of that physical hint. In order to approach these inquiries, the three critical research
questions addressed in Chapter Two, were further explored.

Question One: Does a construction work crew meet the required criteria to be addressed as a
group?

Construction Work Crew
The term crew [i.e., work crew] in commonly-known English dictionaries refers to a collection of
individuals who work together. Merriam-Webster Collegiate Dictionary (2003), for example,
addresses a work crew as a number of persons who “do a specific kind of work together”; who are
involved in a “common activity”; who are associated in a “common interest”.
In Social Science literature this general and informal definition of work crew is converted to a “set
of persons within an organization” who are “highly interdependent” (McGrath, 1984, p. 7); who
perform “specialized” (Arrow, McGrath, & Berdahl, 2000) and “standardized, technical” (Forsyth,
2014, p. 405) tasks which demand “collective action” (Rousseau, Aubé, & Savoie, 2006, p. 540);
and who pursue common “goals through coordinated, interdependent interaction” (Forsyth, 2014,
p. 400).

48

In Construction Industry terminology, a work crew can refer to [almost always] a number of
individuals trained in a specific trade discipline who work together on a construction job site for a
specific period of time on a technical task that requires using equipment, tools, and/or technology
(Spatz, 2000; Forsyth, 2014). For example, at the operation level of a construction project a steelerector crew assembles a structural steel frame using a crane and wrenches; a sheet-metal crew
installs Heating, Cooling, and Ventilation (HVAC) duct-work using a drill and screws; a masonry
crew places bricks using a laser-level; a carpentry crew installs made-of-wood building materials
using a nail/stable-gun. Thus, a construction work crew refers to a few skilled trade workers of
same discipline who collectively perform a particular activity in a construction project.

Does such collective behavior classify a work crew as a group? In other words, does a construction
work crew have the required criteria and/or qualifications to be addressed a group? In order to
investigate this classification, a group will be studied in its professional/scientific terminology.
It should be noted that in academic literature a ‘group’ is sometimes defined differently from a
‘team’; that a team is a group yet a group does not always satisfy every qualification of a team
(Katzenbach & Smith, 1993; Stock, 2004; Forsyth, 2014). In this research dissertation,
nevertheless, these two terms are considered interchangeable.

Construction Work Crew as a Group
The comprehensive definition for a construction work crew which was provided in Section (4.1.1)
indeed captures at the least three (3) central features and key characteristics which are common
among different definition of a group. Specifically, those are:

(a)

size – a minimum of two (2) skilled trade workers;
49

(b)

commonality – a specific construction activity/task which has be completed; and

(c)

interaction – a continuous and collective work style to deliver a product.

Such qualification enabled the researcher to apply the abstract Input-Process-Output (I-P-O) model
by which social science addresses small group dynamics and its interactions to construction work
crew.

Question Two: How does a construction work crew function from the perspective of a group?

(4.2)

Conceptual Reference Model for Dissertation Research

As defined in Section (4.1.1), a construction work crew is a number of individuals trained in a
specific trade discipline who work together on a construction job site for a specific period of time
on a technical task that requires using equipment, tools, and/or technology.

In such a setting, individuals may be of different ages, genders, and races; each may have diverse
characteristics and personalities; and every one may have varying expertise, levels of knowledge,
sets of skills, years of experience, etc.

The group which these trained individuals of a specific discipline form in a construction project
have to follow certain norms and regulations enforced by that group, pursue directions provided
by their superiors; as members of a crew, every one may have to assume responsibilities; members
may have to work collectively and interact with other (new) members, and construct some kinds
of relationships – unless there was a past history such as friendship, work experience, etc.

The technical activity which this work crew has to perform may be of a special type, a certain level
of complexity and difficulty, and it may require specific knowledge, skills, training, etc.

50

The construction job site where the crew is expected to work may also have some characteristics.
For example, it may be located in an extreme environmental condition; it may be in a very highsecurity location, a heavy-traffic, a unique surrounding neighborhood, etc.

Question Three: What independent factors (i.e., initial states and mediators) do influence a
construction work crew performance? And which of those influencing variables are critical on a
work crew’s performance in presence of cues and physical hints?

Four types of input factors in Figure 4.1 likely trigger members of a construction work crew to
interact with one another and render a decision. Therefore, process may be considered as a vehicle
which carries initial states of a group to impact its performance outputs.

Crew

Individual

Influential
Factors

Activity/Task

Interaction

Decision

Environment

Figure 4.1 – Construction Work Crew Input-Process-Output (I-P-O) Conceptual Framework
Adapted from Source: (McGrath, 1964, p. 70; Hackman & Morris, 1975, p. 50; McGrath, 1984, p. 13)
and (Stock, 2004, p. 278; Forsyth, 2014, p. 408)

As stated previously, and also in Chapter Three, in order to identify relevant initial states (i.e.,
independent input variables) of a construction work crew I-P-O, a population of skilled trade
workers were interviewed.

51

(4.3)

Research Findings: Analysis and Discussion

Construction work crews hold major roles in construction project activities and operations, and
offer significant influence on operations performance and outputs. It is thus important that work
crew members are involved in identifying initial states which significantly influence a construction
work crew. Dai et al. (2009) claimed that construction work crews would rarely be invited to take
parts in such investigations perhaps because: (i) they would be taken away from work face and
would not be working on an activity; and/or (ii) it would be a matter of authority and would seem
to be the responsibility of a project manager.

In order to understand the heuristic of construction work crew decision-making in an ambiguous
situation, and to learn about factors which influence such heuristic in presence of cues and physical
hints, a population of hundred and seventy six (176) construction skilled trade workers were
qualitatively interviewed. Interviewees were selected based on availability, and their willingness
for participation. The characteristics of interviewee population are reported in Table 4.1.

52

Table 4.1 – Population Characteristics of Interviewee Construction Crews
Sample Size
Skilled Trade

Owner
Self-Performing

Specialty, SubContractor

Electrician

2

47

Rough Carpenter

5

33

Trim Carpenter

7

43

Tin Knocker
Population Size

An Average
of 17 Years
Experience

39
14

162

(8%)

(92%)

The interviews were conducted one-on-one at construction job sites, and each lasted for an average
of thirty minutes. Each interviewee was asked to refer to following scenario that he or she might
have encountered throughout her or his career as a construction skilled trade worker:

When you arrived on the job site, you realized that the situation did not match the
instructions (e.g., construction drawing; verbal request; work-order; etc.)
originally received; and you needed to acquire more information or informative
instruction in order to complete your activity/task. In the meantime, there was a
readily-available physical cue at the work face which might have – rightly or
wrongly – filled in for the missing pieces of information.

Two discussion topics were followed. In the first follow-up discussion each interviewee was asked
to describe her or his most-likely, immediate action in such a situation considering that he or she
would often work in a crew. Notes were taken from discussions and responses each skilled trade
worker provided.

53

From the contents analyses of interviewees responses to this follow-up question, one hundred and
eighty three (183) ‘catch-phrases’ were extracted, which were then clustered into three (3)
mutually exclusive actions categories presented in Table 4.2.

Table 4.2 – Work Crews Most-Likely Immediate Actions in an Ambiguous Situation
(from Contents Analyses of Qualitative Interviews)
Immediate Action
Autonomous Problem-Solving
Contact Customer
Contact Supervisor

Catch-phrase in Qualitative Interview
e.g., discuss with team member; review
drawing/scope of work/work order; assess the
situation; remembering similar past mistakes
e.g., communicate with customer; learn what
customer needs
e.g., ask project manager for instruction;
discuss with supervisor

Autonomous Problem-Solving
Analyses of interviews contents revealed that construction work crew members referred to
‘Autonomous Problem-Solving’ – that is, he or she would try to resolve the issue in crew and/or
between crews – one hundred and twenty seven (127) times; what made a frequency of almost
sixty nine percent (69%) of catch-phrases – as summarized in Figure 4.2.

54

Figure 4.2 – Frequency Analyses of Work Crews Immediate Actions

Skilled trade interviewees expressed that:

(i)

they had knowledge of current building codes and were aware of construction regulations
applicable to a situation;

(ii)

that often in each circumstance things could be constructed in a certain way and/or only
particular things could be constructed; and

(iii)

that within the crew they had a great number of years of experience, and a rich combination
of expertise available to them.

Therefore they were eager to evaluate the situation with respect to scope of work, to re-study and
review the construction documents, and to discuss the issue with their crew-mate(s) in order to
interpret the circumstance, and to make a sound decision.

55

Contact Supervisor
Skilled trade interviewees referred to ‘Contact Supervisor’ as their immediate response forty six
(46) times – a frequency of 25% in extracted catch-phrases. These work crew members mentioned
that they wished to respect the hierarchy of responsibilities, and to follow the lead of their superiors
– even though they themselves knew what needed to be done, these skilled trade workers preferred
to receive further instructions from their supervisors.

Contact Customer
‘Contact Customer’ was least frequent catch-phrase in the interviews contents. Skilled trade
interviewees mentioned ten (10) times – a frequency of 5% – that “the customer was to pay for the
work” and that “it was important to involve the customer in decision making”. Therefore they
contacted the customer prior to taking any action. These crew members stated they communicated
with the customers to learn about their desired work outputs, and to share with and/or teach them
about the situation.

Autonomous Problem-Solving
As shown in Figure 4.2, the majority of interviews content [for the first follow-up discussion] fitted
in the category of ‘Autonomous Problem-Solving’, and supported the Problem Statement asserted
in Section (1.3) – that is, when a construction work crew faces an ambiguous situation at their
work face, the work crew tends to rely on readily available hints in its surroundings in order to
make a decision about proceeding. On that basis, and also findings presented by Lahouti &
Abdelhamid (2012), and Lahouti (2013) a conceptual framework for such a process was developed
as illustrated in Figure 4.3.

56

Work
order

Does job-site
Match* work
order?

* e.g., Is drawing complete? Is instruction explicit? Is information perfect?

No

Interpret
missing*
information

Complete
the task

Yes

Was the
Interpretation
correct?

Additional resource is
lost to problem-solving

Yes

Complete
the task

No

Complete
the task

Undo the task
and start over

No resource
is lost

Additional resource is lost
to problem-solve &
undoing the wrong

Deliver
work ordered

Figure 4.3 – Conceptual Framework for Work Crews Autonomous Problem-Solving

The autonomous problem-solving framework of Figure 4.3 demonstrates that when a construction
work crew arrives at a job site, and finds its workface different from provided work-directives, and
that based on construction documents (e.g., shop drawings; work-orders; etc.) the work face
condition does not match with its expectation, then – according to this investigation – 69% of
actions that the work crew immediately takes on to complete its assigned work focus on finding
the missing information, interpreting present ambiguity, and decision-making by relying on
surroundings cues and hints which are readily available. When the decision leads to incorrect
and/or undesired outputs (i.e., a waste) the work crew has to undo its work, and to find additional
information from other sources, and then correctly re-complete its work. The dotted rectangle in
color ‘black’ in Figure 4.3 identifies the cue-interpretation loop.
57

Heuristics of the autonomous problem-solving were concluded from contents analyses of
qualitative interviews with construction work crews and are summarized in Figure 4.4 – where
action “Discuss with other crew member(s)” was mentioned one hundred and seven (107) times
(i.e., a frequency of 84 percent).

Figure 4.4 – Frequency Analyses of Work Crews Autonomous Problem-Solving Heuristics

This frequency is also in agreement with the Problem Statement of this research dissertation stated
in Section (1.3) that interaction between members of a construction work crew is attribute to its
decision-making. Application of Input-Process-Output reference framework of Figure 4.1 to the
autonomous problem-solving process map conceptually illustrated in Figure 4.3 may define the
heuristics of cue-interpretation within the scope of this investigation.

58

As conceptually presented in Figure 4.5, missing information about activities/tasks forces a work
crew to interpret missing information, which in turn demands interactions between crew members.
The quality of these interactions is functions of characteristics of each member – the input factor
concerned in this research dissertation. As a result of these interactions, physical hints at the work
face are interpreted and accordingly a decision will be rendered to complete the activity/task.

Interpret missing information

Missing
Information
/Cue

Individual
Characteristics

Interaction

Interpretation
of cues & hints

What-to do

Figure 4.5 – Conceptual Heuristic of Cue Interpretation in Construction Work Crews

The second follow-up discussion focused on factors which were influential on interactions
between work crew members. Each interviewee was asked to discuss the factors (e.g., individual;
environmental; etc.) which in her or his opinion were the most important in the process of making
a decision as a work crew. From the contents analyses of these unstructured, and open discussions
two thousand and seventy (2070) catch-phrases were extracted, and clustered into fourteen (14)
descriptive factors. It should be noted that these factors were described in such a fashion to
inductively represent catch-phrases, and to describe implied contents of interviews. Table 4.3
reports these influential factors alongside their repetition counts.

59

Table 4.3 – Influencing Factors on Construction Work Crews Interactions
(from Content Analysis of Qualitative Interview)
Influential Factor

Count

Catch-phrase in Qualitative Interview

Collaboration

17

e.g., willingness to work as/in a team

Common Sense

116

e.g., common sense

Communication

159

Complexity of Task

87

Confidence and Trust

298

Cost and Efficiency

181

e.g., communication; listening; sharing ideas
e.g., difficulty of execution; whether there is a need for
architectural/structural design
e.g., confidence in team member; past experience/history
with team member; trusting team member
e.g., being more productive; completing a task
cheaper/easier/faster

Customer Satisfaction

36

Feasibility of Solution

14

Knowledge and Skills

457

Morale and Personality

214

Mutual Respect

156

Open-mindedness

266

Quality of Product

39

e.g., aesthetic considerations; ease of future maintenance

Work Style

30

e.g., work style

e.g., what customer interest/need/preference/satisfaction is
e.g., code compliance; owner specifications/standards;
what best interest/option/service/value [for customer] is
e.g., team member’s ability/capability; team member’s
craftsmanship, experience, skill, and training
e.g., how team member approaches with a suggestion; team
member’s attitude; team member’s honesty
e.g., respect team member’s opinion/responsibility; respect
team relationship
e.g., accepting/considering team member’s
solution/suggestion; being open-minded; considering that
there may be a better/different way of doing things

Knowledge and Skills
As presented in the histogram of Figure 4.6, the factor Knowledge and Skills of a work crew
member was acknowledged the most frequently by interviewees. The construction work crew
members stated that degree of experience, diversity in expertise and specialties, capabilities, level
of craftsmanship, and types of trainings were the most impactful factors on their within- and

60

between- crews interactions and processes of rendering a decision on what-to do. This conclusion
from the interviews contents is in tune with Hamzeh et al., (2012) claim that knowledge, skills,
and abilities of a construction work crew member (i.e., he or she who improvises) influence the
improvisation outcome. The existing literature on Small Groups and their Dynamics also recognize
a relationship between knowledge, skills, and abilities of group members and the overall
performance of group and its outcome (Jones, 1974; Bennis & Biederman, 1997; Forsyth, 2014).

Figure 4.6 – Frequency Analyses of Influencing Factors on Work Crew Members Interactions

Confidence and Trust
Confidence and Trust was the second influencing factor to which construction work crew members
referred the most frequently. The interviewees recognized this factor from two different aspects

61

of: (a) morality – this is, whether their crew-mate could be trusted as a human being, whether he
or she was trustworthy, and (b) professionalism – that is, whether their crew-mate made a
suggestion to only complete her or his assignment; or he or she offered a well-thought, valueadding decision. The interviewees acknowledged that level of Confidence and Trust in their crewmates rooted in the past experiences – if any – they had with her or him where they witnessed her
or his approaches, decisions, and actions, whether he or she presented a bad judgment, and whether
he or she was ignorant, and/or oblivious.

Open-mindedness
The third most-frequently recognized influential factor by interviewees was Open-mindedness.
Based on their years of experiences on construction job sites, the work crew members asserted that
it was important to accept that their crew-mates had different opinions; and that he or she had a
different approach in problem-solving. They interviewees concluded that Open-mindedness of
their crew-mates and that he or she would even consider other’s suggestions/solutions were often
crucial in the crew performance.

Moral and Personality
Moral and Personality of the construction work crew members was the fourth most-frequently
mentioned factor. The interviewees stated that their crew-mates’ attitudes, how personable he or
she was, and how he or she reached out to the crew as a unit with a suggestion affected the
dynamics of interactions between them, and how the crew shaped its decisions.

62

(4.4)

Statistical Analysis

As reported in Table 4.1, a small portion of interviewee population – i.e., a sample size of fourteen
(14) which made almost 8% of entire population – was represented by Owner Self-Performing
work crews of a higher-education institution; and the rest of population was represented by
Specialty, Sub-Contractor work crews. The most prominent difference between these two workcrew types was their structures – or rather, as addressed in Section (4.2) and illustrated in Figure
4.1, characteristics of the crews. Although these characteristics were largely determined by
common, standard regulations of construction work crews, owner self-performing crews exercised
a set of uncommon norms. For example, while the specialty sub-contractors crews represented
common characteristics of typical construction work crews – as such was, hierarchy of
responsibilities and roles within a work crew (foremen, journeyman, etc.), temporary professional
partnerships of crews on construction job sites, or competition between work crews of similar
disciplines, etc., in the owner self-performing work:

(i)

members of each trade were expected to assume equal roles within their crews;

(ii)

work crews were professional partners in almost every job site;

(iii)

value was not solely in completing assignments as work crews were owner’s
representatives – hence no sense of competition; etc.

Analyses of interviews contents when controlled for influential factors revealed that trades of
diverse disciplines and crews of different structures differently weighted (i.e., mentioned with
different frequencies) the impact of each (influential) factor. Such comparison however is not to
imply that different construction skilled trades and/or work crews disregarded the influence of one
factor over another. Rather, it is to demonstrate variation for intensity of impact that different

63

skilled trades and/or work crew types believed a factor would impose on their interactions with
other crew members; that influential factors were possibility dependents of the characteristics of
work crews, their trade disciplines, and/or their activities/tasks. Statistical analyses were conducted
in order to evaluate the differences between these frequencies.

Examination for Association
Table 4.4 summarizes the frequencies of influential factors with respect to each trade’s discipline.
To determine whether differences between these frequencies suggest any association with trade’s
disciplines, Chi-Squared Test of Independence was conducted. In other words, Chi-Squared Test
was to determine whether the proportion of frequencies for factor Knowledge and Skill, for
example, was significantly different between electricians, carpenters, and tin-knockers.

64

Table 4.4 – Influencing Factors on Construction Work Crews Interactions with Respect to Trade
Electrician

Rough
Carpenter

Trim
Carpenter

Tin
Knocker

Total
(Table 4.3)

Collaboration

4

6

4

3

17

Common Sense

29

29

29

29

116

Communication

44

30

50

35

159

Complexity of Task

24

23

18

22

87

Confidence and Trust

87

61

81

69

298

Cost and Efficiency

53

36

49

43

181

Customer Satisfaction

6

8

16

6

36

Feasibility of Solution

4

2

5

3

14

Knowledge and Skills

130

94

124

109

457

Morale and Personality

62

44

58

50

214

Mutual Respect

46

33

40

37

156

Open-mindedness

72

64

72

58

266

Product Quality

8

9

13

9

39

Work Style

8

6

10

6

30

Influential Factors

The R Statistical Computing Platform returned a relatively small Chi-Squared of 18.852, and a pvalue of 0.9973 ( = 0.05). The hypothesis that influential factors were identified independent of
trade’s disciplines therefore cannot be rejected. In other words, these values suggest that frequency
of identifying a factor influential appeared to be statistically unassociated with the trade’s
discipline; and that frequencies of identifying a factor as influential were similar between different
trade’s disciplines.

Similar hypothesis that identified influential factors were also independent of crew types was
tested with respect to interviews contents summarized in Table 4.5. The R Statistical Computing
Platform returned a Chi-Squared value of 171.6 with the probability of 2.2e-16 ( = 0.05) – both

65

of which indicate a statistically significant difference in frequencies of identifying influential
factors with respect to crew types.

Table 4.5 – Influencing Factors on Construction Work Crews Interactions with Respect to Crew Type
Owner SelfPerforming

Specialty SubContractor

Total
(Table 4.3)

Collaboration

6

11

17

Common Sense

3

113

116

Communication

12

147

159

Complexity of Task

5

82

87

Confidence and Trust

7

291

298

Cost and Efficiency

5

176

181

Customer Satisfaction

13

23

36

Feasibility of Solution

6

8

14

Knowledge and Skills

16

441

457

Morale and Personality

5

209

214

Mutual Respect

5

151

156

Open-mindedness

21

245

266

Product Quality

8

31

39

Work Style

3

27

30

Influential Factors

In order to evaluate how significant these differences are, Fisher’s Exact Test for Statistical
Significance was conducted in R Statistical Computing Platform.

Analysis of Differences
In order to evaluate whether these differences in frequencies were of any significant importance,
Fisher’s Exact Test for Statistical Significance was conducted in R Statistical Computing Platform.

66

Owner Self-Performing Work Crews
While bar chart of Figure 4.6 represented each factors in relation with the entire population of
interviewees, Figure 4.7 categorized factors on which the owner self-performing work crews
emphasized with greater frequency – that is: Customer Satisfaction, Product Quality, Feasibility
of Solution, and Collaboration as those strongly influential on their interactions.

Figure 4.7 – Comparison of Influential Factors with Greater Frequencies for Owner Self-performing
Work Crew

The results of Fisher’s Exact Test for statistical significance – reported in Table 4.6 – revealed that
odds of the self-performing work crews referring to factor Feasibility of Solution was thirteen (13)
times greater than that of the specialty sub-contractor crews. Similarly, these odds for factors
Customer Satisfaction, Collaboration, and Product Quality were respectively eleven (11), ten (10),
and five (5) times greater for the owner self-performing work crews. Although factors Openmindedness, Communication, Complexity of Task, and Work Style were also acknowledged by
owner self-performing work crews more frequently, the Fisher’s Exact Test reported that their
67

differences were not statistically significant – that is, both crew types emphasized these factors as
strongly.

Table 4.6 – Fisher's Exact Statistical Significance Test Result: Influential Factors of Greater Frequencies
for Self-Performing Work Crews
Odds
Ratio

p-Value

Feasibility of Solution

13

0.00005

Statistically Significant

Customer Satisfaction

11

0.00000

Statistically Significant

Collaboration

10

0.00019

Statistically Significant

Product Quality

5

0.00103

Statistically Significant

Work Style

2

0.23060

Not Statistically Significant

Open-mindedness

2

0.08438

Not Statistically Significant

Communication

1

0.27630

Not Statistically Significant

Complexity of Task

1

0.81280

Not Statistically Significant

Influential Factor

A possible explanation for statistically-significant differences can be that these work crews did not
consider their customers the only owners, and solely the entities which pay for the work; instead
they understood that they themselves were representatives for the owner (i.e., in-house crews),
assumed owner roles in delivering the desirable end products, and paid more attention to what the
customer wished to receive. The core activities of these owner self-performing crews were focused
on alteration and improvement of facilities which aged for one year to several decades. Taking the
time to discuss things that could be done in presence of existing conditions; and to collaboratively
determine those which provided the best value(s) in the interest of the owner, while taking into
account the maintenance-ability of end products, and also the ease of future alteration and/or
improvement work promised these work crews the delivery of high-quality end products. Given
these characteristics of activities, these crews believed that feasibility of suggested solutions

68

contributed much in quality of their deliverables. The senses of ownership can also explain the
different immediate actions that these work crews emphasized in comparison with specialty subcontractor crews.

An important remark in regards with factor Complexity of Task is that work crew members
addressed this factor in two different contexts: (a) they counted complicated activities/tasks more
ambiguous and therefore influential on their interactions with crew-mates for it did invite more
discussions, and exchanges of knowledge and skills; and (b) every interviewee stated that her or
his first action when faced with an ambiguous situation was dependent on its complexity. That is,
the crew members were less likely to consult available cues, to interpret for the missing
information, and to autonomously resolve the issue the more complex an activity/a task was.

Figure 4.8 shows immediate action Contact Customer was mentioned by owner self-performing
work crews much more frequently; and Fisher's Exact Test confirmed that such a difference was
of statistical significance. The odds of these crew members consulting their customers, seeking
their opinions, and learning about their value were ten (10) times greater than that of specialty subcontractor crews.

69

Figure 4.8 – Comparison of Immediate Actions to Ambiguous Situations with Respect to Construction
Work Crews Types

In contrast, specialty sub-contractor work crews referred to immediate action Autonomous
Problem-Solving more frequently, and with no statistically-significant difference. The results of
Fisher's Exact statistical significance test are reported in Table 4.7.

Table 4.7 – Fisher's Exact Statistical Significance Test Result: Most-likely Immediate Actions
Influential Factor

Odds Ratio

p-Value

Contact Customer

10

0.00214

Statistically Significant

Autonomous Problem-Solving

2

0.08214

Not Statistically Significant

Contact Supervisor

1

1.00000

Not Statistically Significant

70

Specialty Sub-contractor Work Crews
Presented in Figure 4.9, specialty sub-contractor work crews put their emphasis on factors
Knowledge and Skills, Confidence and Trust, and Morale and Personality. And results of the
Fisher’s Exact Test revealed that specialty sub-contractor work crews had three (3) times greater
odds of referring to factor Confidence and Trust, and Morale and Personality. This odds for factor
Knowledge and Skills was two (2); and in comparison with owner self-performing crews these
differences were statistically significant.

Figure 4.9 – Comparison of Influential Factors with Greater Frequencies for Specialty Sub-contractor
Work Crews

Reported in Table 4.8, Fisher’s Exact Test for statistical significance also presented twice-as-much
odds of identifying factors Cost and Efficiency, Mutual Respect, and Common Sense as influential

71

on interaction between specialty sub-contractor work crew members did not prove differences
which were statistically significant.

Table 4.8 – Fisher's Exact Statistical Significance Test Result: Influential Factors of Greater Frequencies
for Specialty Sub-Contractor Work Crews
Influential Factor

Odds Ratio

p-Value

Morale & Personality

3

0.02676

Statistically Significant

Confidence & Trust

3

0.00610

Statistically Significant

Knowledge & Skills

2

0.02810

Statistically Significant

Cost & Efficiency

2

0.09029

Not Statistically Significant

Mutual Respect

2

0.27190

Not Statistically Significant

Common Sense

2

0.20760

Not Statistically Significant

Take-away: Responses provided by interviewees to an identical, general scenario and independent
of trade discipline and/or crew type may be statistically different when considered from alternative
points of views. Such differences are negligible because the interviews focused on an event (i.e.,
an ambiguous situation) and how construction skilled trades would act as a work crew in that
situation. These differences however would have been significant if the concentration of problem
statement was on activities/tasks a specific type of crew or a skilled trade of particular discipline
(e.g., a carpenter; an electrician) would complete.

(4.5)

Summary

In this chapter findings of investigation were presented in response to research questions: first, a
construction work crew was examined from the view-point of a group; and then group’s conceptual
Input-Process-Output (I-P-O) framework was adapted to the construction work crew. Second, data
gathered from qualitative interviews with construction work crews were analyzed revealing their
72

most-likely, and immediate action when faced with an ambiguous situation as autonomous
problem-solving. The analyses also identified three factors of knowledge and skills, confidence
and trust, and open-mindedness as the most influential on interactions between the members of
work crews while interpreting an incomplete instruction. These findings were utilized in modeling
a work crew, and in simulating interactions between its members.

73

(5)

Agent-Based Modeling Simulation (ABMS)

74

As was expressed in Section (1.3), this research addresses the of imperfect information and
incomplete instruction forcing a construction work crew during an activity execution. With
reference to such a problem, the goal of this investigation is to address an operational research
question:

Question: In a situation where a construction work crew receives an incomplete directive, how
does the interaction between crew members impact deployment and interpretation of cues and
physical hints, at the work face and its surroundings, in rendering a decision to act?

Towards approaching this goal, one of the achievable objectives in Section (1.4) proposed to
develop an Agent-based Modeling Simulation (ABMS) to explore such impacts.

(5.1)

Agent-based Model Information

The Agent-based Modeling Simulation (ABMS) simulation was constructed with two types of
essential components as summarized in Table 5.1: (i) a dynamic agent which would represent a
group of size n; and (ii) a static agent which would represent an environment of l × m neighbors
with which dynamic agents would interact.

Table 5.1 – Essential Components of Developed ABMS
Agent Type

Representation

Dynamic : User

Static : Environment

Attribute
- Cartesian coordinate
- Personal & professional
characteristics
- Cartesian coordinate
- Color

75

(5.2)

Agent-based Model Presentation

In this abstract model – as reported in Table 5.2 – the blue circle identifies agents and represent a
construction work crew of n members. This work crew was expected to reach the only red square
of agent environment to symbolize completion of a construction activity – each counted with
variable cycle. Completing an activity required conducting a collection of tasks – which were
simulated by relocating from one white square to another. White square (i.e., tasks) were replaced
with black squares to abstractly demonstrate imperfect and incomplete instructions. Black squares
as obstacles were considered to be incomplete instructions because each would only instruct agents
to NOT relocate onto them with NO further information about the ‘what-to’ do. Combinations of
black and white squares created cross-paths to symbolize presence of cues and physical hints. In
other words, a cue was represented by availability of more than one white square onto which agents
could relocate.

Table 5.2 – Presentation of Essential Components of Developed ABMS
Entity

Description

Abstraction

User A construction crew of n members

Presentation

Group of n agents

Blue circle

Von-Neumann neighbor

White square

Completion of a construction task

Destination

Red square

An incomplete instruction

Obstacle

Black square

Environment Construction activity

In short, the work crew in this agent-based model had to interact with cues and physical hints,
interpret them, to decide which activity should be conducted. The shortest path between each white
square (or rather a neighbor) and the only red square (i.e., destination) were introduced to
symbolize the most efficient set of tasks to complete a construction activity.

76

Members of this work crew, in the meantime, were in possession of three (3) behavioral factors
which were identified most influential in analyses of interview contents reported in Chapter Four.
Figure 4.6 presented these factors, in order of magnitude, to be: (i) Knowledge and Skills; (ii)
Confidence and Trust; and (iii) Open-mindedness. Aggregating these influential factors in the
ABMS framework was a challenging task.

Skillfulness
Factor Knowledge and Skills was defined as skillfulness of each crew member. Skillfulness refers
to skills that are obtained by a work crew member through education and/or experience which
enables her/him to do the right activity, right the first time. It, thus, seemed rational to present this
factor in terms of probability of correctly identifying a Von-Neumann neighbor (i.e., north; east;
south; and west) which would lead the shortest path to the destination neighbor.

Confidence and Open-mindedness
With reference to Table 4.3, factor ‘Confidence and Trust’ was presented as confidence of a work
crew member in ability, capability, and reliability of her/his crew-mate in doing right activity (the
first time). Factor open-mindedness was presented as willingness of a crew member to accept
choices of her/his crew-mate for (doing) right activity/task.

Probability of occurrence for these two factors were defined as weights to the rendered decision
based on skillfulness of a crew member. Section (5.3) further discusses definition of such weights.

77

Crew
Given that members of a work crew are usually the same while completing an activity, another
factor was introduced in addition to the three aforementioned behavioral factors in order to
represent the uniqueness of n agents in a group. In other words, factor Crew was to control
randomness in the experiments, and to ensure that operational values in each simulation cycle were
drawn from an identical population. As Table 5.3 reports, influential factors skillfulness,
confidence, and open-mindedness as well as Crew were assumed to follow uniform distributions
of probabilities.

Table 5.3 – Probability Distribution Function Introduced for Influential Factors
Influential Factor

Distribution Function

Skillfulness

Uniform  U (0,1)

Confidence

Uniform  U (0,1)

Open-mindedness

Uniform  U (0,1)

Crew

Uniform  U (0,1)

It should be noted that qualities of these input factors were not the scope of this investigation; and
neither data was collected on skillful of work crew members, their confidence/skepticism in other’s
capabilities, or how open-minded/stubborn they were; nor was such information available. For the
purpose of simplicity, then, uniform distributions were introduced to ensure equal likelihood of
occurrence for each value.

These uniformly distributed values were considered between zero and one to represent close-toreality scenarios; that is, neither of these factors was absolutely absent nor present – i.e., 0 or 1,

78

respectively. Instead, each value represented a partially-absent/present input factor. These
partialities assumed to be interactions enablers.

Agent-based Modeling Environment
As stated in Section (3.3.1), Netlogo is capable of exploring correlations between behaviors of
individual entities and patterns that emerges from interactions between those entities (Wilensky,
1999). Figure 5.1 graphically presents the aforementioned essential components of the ABMS in
NetLogo environment. Although addressed in Section (3.3.1), the use and relevance of this
depiction to the problem at hand will be explained with greater details in following sections.

Figure 5.1 – Graphical Presentation of Developed ABMS

79

(5.3)

Agent-based Model Simulation

As demonstrated by flowchart of Figure 5.2, in this simulation each of n (greater than zero) agents
would relocate from one neighbor to another in a maze in order to reach a pre-determined neighbor
destination through the shortest path:

Identify Neighbor
which is on the
Shortest Path

Visit Neighbor

Identify
Neighbors in
N, E, S, & W

Arrive at the
Destination

Figure 5.2 – Conceptual Framework for Development of an Abstract Agent-Based Model

At any given coordinate in the environment an agent would identify its available, adjacent, VonNeumann (i.e., N; E; S; and W) neighbors. Available neighbors for an agent would be those which
were neither obstacles nor previously visited by that agent. The agent would, then, move to the
neighbor which would lead the shortest path to the destination. In a situation where none of the
adjacent, Von-Neumann neighbors were available, the agent was expected to relocate into its lastvisited neighbor. A neighbor was reported dead-end in a coordinate where none of the adjacent
neighbors were available and the last-visited neighbor was already re-visited – i.e., a neighbor was
visited twice.

80

As Figure 5.3 conceptually illustrates, identifying the neighbor that leads the shortest path in any
coordinate would be dependent of agent’s skillfulness level. This assumption was made to adopt
the results of qualitative interviews – summarized in Figure 4.6 – which identified ‘Knowledge
and Skill’ as the most frequently-mentioned, influential factor.

Identify Neighbor which is on the Shortest Path :
Confidence

User i
(i = 1,2,3,…, n)

Determine a
Neighbor

Skillfulness

Decision i
(i = 1,2,3,…, n)

Open-mindedness

Figure 5.3 – Decision-Making Process for Agent i in Group of n Agents

The skillfulness level p would be drawn from a uniform probability distribution within the range
of (0 , 1) – as reported in Table 5.3. The skillfulness level p meant that agent would relocate with
probability of ‘p’ to a neighbor on the shortest path, and would contemplate to relocate with ‘1–
p’ probability to a (random) Von Neumann neighbor. This binary decision was simulated by
evaluating a random floating number ‘r’ between zero (0) and one (1) prior to a move. An agent
would relocate through the shortest path for each ‘r’ smaller than ‘p’ – that is, 0 < r ≤ p. The agent
would consider to randomly relocate to one of its Von Neumann available neighbors for any other
value of ‘r’ – i.e., p ˂ r ≤ 1. It may be noted that one of the key role for factor Crew was, for
example, to provide a seed value for floating number ‘r’ to ensure it was drawn from the same
population for each simulation cycle.

81

The greater this probability p was the more likely an agent was to correctly orient towards the
shortest path. In contrast, the lesser this probability was, the more likely the user agent was to
move into a random Von Neumann available neighbor – the latter was to simulate interpretation
of a cue and a physical hint which could result in inaccurate or rather unsuccessful identification
of the shortest path.

This output as demonstrated in Figure 5.3 would be, then, adjusted with the product (i.e.,
multiplication) of the agent’s confidence and open-mindedness:

Decision Power user agent i = (Confidence) user agent i × (Open-mindedness) user agent i ....... Equation 5.1

Multiplication of these two factors were considered to determine the weight for each rendered
decision to present a non-compensatory effect. n weighted decisions would be rendered – i.e., one
per agent i – at any given coordinate, and the decision with the least magnitude of weights would
be considered the one with which each agent would comply. The least weighted output was to
present interactions within a group and the impact of influential factor on a group output.

The least weighted output would also represent the offset that factors confidence and openmindedness impose; that is – for example – a relatively open-minded agent which was skeptical
about another agent’s decision, would have a weight that represented more of its skepticism rather
than open-mindedness; or an agent which was confident in another agent yet itself was relatively
stubborn, would have a weight that represented the level of its stubbornness more than confidence.

82

In summary, and as can be concluded from Figure 5.3, in this ABMS alternative decisions would
be generated by factor skillfulness while power of decision-making would be controlled by factors
confidence and open-mindedness.

A legitimate concern may be raised about any correlation between factor Open-minded and factor
Confidence and Trust: since a strong correlation between these two factors suggests
interchangeability of their values, the product (i.e., multiplication) of these values would do no
justice in representation of their interactions.

As summarized in Table 5.4, two hundred and ninety eight (298) catch-phrases were clustered to
describe factor Confidence and Trust – seven (7) of which were extracted from owner selfperforming crew interview contents, and two hundred and ninety one (291) from specialty subcontractor crew. Similarly, from two hundred and sixty six (266) catch-phrases describing factor
Open-mindedness, twenty one (21) were from owner self-performing crew and two hundred and
forty five (245) were from specialty sub-contractor crew.

Table 5.4 – Frequencies of Influential Factors with Respect to Crew Type
Confidence
and Trust

Openmindedness

7

21

Specialty Sub-Contractor

291

245

Total (Table 4.3)

298

266

Owner Self-Performing

Frequencies of these catch-phrases were apportioned with respect to trades within each crew type
and are reported in Table 5.5. For example, 29.21% (i.e., 85) of clustered catch-phrases for factor
Confidence and Trust were mentioned by electricians, 20.27% (i.e., 59) and 26.80% (i.e., 78) by
83

rough and trim carpenters, respectively, and 23.71% (i.e., 69) by tin-knockers in specialty subcontractor crew.

Table 5.5 – Apportioned Frequencies of Influential Factors with Respect to Trade and Crew Type
Confidence
and Trust

Openmindedness

Electrician
Owner SelfRough Carpenter
Performing
Trim Carpenter

28.57%

4.76%

28.57%

38.10%

42.86%

57.14%

Electrician

29.21%

28.98%

Specialty Sub- Rough Carpenter
Contractor
Trim Carpenter

20.27%

22.86%

26.80%

24.49%

Tin Knocker

23.71%

23.67%

To examine whether any correlation exists between factor Confidence and Trust and factor Openmindedness, the scatter plot of Figure 5.4 was constructed based on Table 5.5. The fitted linearregression line resulted in a coefficient of determination (R2) of slightly greater than 50% – which
conveniently suggests a neutral correlation between these two factors. Considering an R2 value of
greater 0.60 to be worthwhile as rule-of-thumb, R2 = 0.5105 does suggest no troublesome
correlation between these two observed, influential factors.

84

Figure 5.4 – Linear Correlation between Influential Factors Confidence & Trust and Open-mindedness

(5.4)

Agent-based Model Result

This agent-based model was simulated for a total of 199,948 cycles. Section (5.5.2) will explain
the logic for conducting this number of repetitions. One (1) cycle of simulation was completed
when the agents reached the destination. Simulation cycles were independent from one another –
i.e., each was an attempt by agents to reach the destination passing through thirty two (32) nondiagonal neighbors with NO learning from the previous experiment(s). Considering the placement
of black and white squares (i.e., layout of the maze in Figure 5.1), thirty-two represented the most
efficient (i.e., shortest) performance.

85

Each cycle was to symbolically represent that activity duration would change when construction
work crew was forced to take cues and physical hints and interpret them in order to determine
‘what-to’ do. Under the assumption of integrity, in this ABMS agents followed their moral
principles and did NOT intentionally take wrong cues nor interpret them inaccurately to benefit
from a lengthy activity.

In each simulation cycle, three (3) dependent output variables were measured:

(i)

Number of neighbors visited: number of neighbors n agents group visited to relocate
between two specific neighbors was counted to represent the operation time. With the
perspective of construction project management, this symbolizes effectiveness (i.e.,
doing right things) of resources utilization in completing an activity/a task.

(ii)

Number of neighbors re-visited: number of neighbors n agents group re-visited once –
i.e., visited twice – was collected to represent the concept of rework in construction
operations. In terms of efficiency (i.e., doing things right), it was to represent the
amount of resources utilized to undo a completed wrong activity, and to do the correct
activity.

(iii)

Number of neighbors dead-end: number of neighbors dead-end that the n agents group
visited was tallied. Relocating into a dead-ended neighbor was to simulate that every
feasible option was exhausted. With respect to construction project management/
operation, this output variable would represent demands for more instruction/
information – which would often dictate inputs from superintendent, project engineer,
etc.
86

The descriptive measures of these outputs summarized in Table 5.6 reveal two (2) major points:
first – population size of agents contributed in variations of each outputs category. The maximum
number of neighbors visited and dead-end were the least (i.e., 2,123 and 69, respectively) for the
group of 5 agents, while those for a group of two agents were the most (i.e., 3,886 and 195,
respectively). In a different fashion, groups of two and three agents re-visited respectively the least

(i.e., 159) and the most (i.e., 176) number of neighbors. These measures may imply that a group
of agents – in comparison to one – influence the outputs of this model.

Table 5.6 – Statistical Descriptive Measures for ABMS Outputs Variables
Neighbors

Visited

Re-visited

Dead-end

Minimum

Maximum

Mean

1 agent

32

2,711

71.74

Standard
Deviation
111.05

2 agents

32

3,886

64.93

82.47

3 agents

32

2,353

60.74

63.85

5 agents

32

2,123

56.82

48.20

1 agent

0

167

6.61

11.97

2 agents

0

159

6.62

11.83

3 agents

0

176

6.56

11.65

5 agents

0

168

6.40

11.35

1 agent

0

116

0.70

3.71

2 agents

0

195

0.32

1.96

3 agents

0

91

0.15

0.98

5 agents

0

69

0.04

0.46

Population

Second – values in each outputs category were more clustered around smaller values as population
of agents grew larger. For instance, mean number of neighbors visited decreased by (roughly) 20
percent from ≈ 72 to ≈ 57 for one and five agents respectively and the numbers of neighbors 5

87

agents visited were 57 percent more clustered around mean than those of 1 agent (i.e., standard
deviations of ≈ 111 and ≈ 49 for one and 5 agents, respectively).

(5.5)

Agent-based Model Analysis

In context of this research dissertation topic, the aforementioned dependent output variables were
analyzed to determine: (a) whether influential factors on interaction among work crew members
would cause variation in operation time and efficiency of resources utilization; would result in
rework and generate waste; and (b) whether they would increase the likelihood of demand for
assistance – e.g., supervisor instruction/involvement.

Exploratory Analysis
In order to learn about distributions of outputs, and to evaluate changes of outputs with respect to
input factors a series of exploratory statistical analyses were conducted.

Construction Work Crew Size
The first step in process of analyses was set to understand whether changes in outputs would be
observed at large with respect to a group size greater that one agent. Therefore, the ABM was
simulated with population of one, two, three, and five agents, for 199,948 cycles each. Then, to
make interpretation straightforward, probability density functions were constructed for outputs of
each population size. The PDF plot enables an observer to compare magnitudes of probability ––
i.e., the area under each distribution curve corresponding to a range of data.

In Statistics terms, a Probability Density Function (PDF) estimates the probability that a random
variable belongs within a defined range of values. In other words, it defines the likelihood of

88

occurrence for a specific event. Figure 5.5, the probability density of number of neighbors visited
for different group sizes, demonstrates a pattern in which the likelihood that an operation time
would fall within a particular range varies as group sizes change. This change may be interpreted
in favor of the optimal number of neighbors visited that there is a greater likelihood that this
number is optimal in a group of fewer agents. It can also be interpreted in favor of group size –
that non-optimal number of neighbors visited would be less likely to occur in groups of more
agents in comparison with those of fewer – in case of this investigation from five to one; although
the variations of outputs were not as significant in magnitudes between groups of two and five
agents as was between those of one and five (or even one and two).

The vertical dotted line in Figure 5.5 shows the value of optimum number of neighbors visited
established at the thirty two (32) steps in the maze (i.e., Figure 5.1).

89

Figure 5.5 – Variation in Number of Neighbors Visited with Respect to Group Size

Although work crew size by itself was not the focal point in this research dissertation as one of the
sources for variation of outputs, it was discussed as the facilitator of interaction between members
of a crew (James, 1951; O'Dell, 1968). With reference to Figure 5.3 it could be claimed that in a
case of only one agent, neither factor confidence nor open-mindedness affected outputs as there
existed no interactions; and that skillfulness was the solely influential one. In contrast, all three
factors influenced outputs in a case of two or more agents. The next step was set to evaluate the
impact(s) of these three factors on ABMS outputs.
A population of size two was assumed to represent the concept of ‘fair share’ in participations;
and to reasonably eliminate the ‘majority-rule’ effect.
90

Take-away: A group contributes to change of ABMS outputs.

Throughout analyses, as Section (3.3.2) discussed in details, a set of reference values were
introduced not only to make the process simple and straightforward, but also more meaningful and
realistic.

Analysis Reference Value
Table 5.3 reported that influential factors were assumed to carry probability distributions between
zero and one – i.e., U (0,1). Arithmetic mean value (i.e., 0.50) of these independent input factors
was chosen as the dividing point in evaluating their influence on dependent output variables. As
demonstrated in Figure 5.6, four regions/quadrants were constructed each of which assign an agent
with a value of greater or lesser than 0.50.

91

Figure 5.6 – Graphical Presentation for Analysis Reference Value

Construction Work Crew Skillfulness
In order to explore whether a correlation existed between variation in outputs and skillfulness level
of each agent, frequency analysis for number of neighbors visited was conducted. Reported in
Table 5.7, (roughly) similar portions of 199,948 simulation cycles resulted in optimal and least
optimal number of neighbors visited – i.e., respectively, 17.52% cycles visited 32 neighbors and
18.30% visited more than 72. Levels of skillfulness for agents were plotted in terms of its
corresponding value being either ends of this range.

92

Frequency

Percent
of Total

≤ 32

35,022

17.52%

> 32 and ≤ 40

70,062

35.04%

> 40 and ≤ 48

27,229

13.62%

> 48 and ≤ 56

15,052

7.53%

> 56 and ≤ 64

9,599

4.80%

> 64 and ≤ 72

6,390

3.20%

36,594

18.30%

> 72

81.70%

Number of
Neighbors Visited

29.15%

Table 5.7 – Frequency Analysis of ABMS Output: Number of Neighbors Visited by Two Agents

In Figure 5.7, population in color ‘blue’ represents simulation cycles in which 32 neighbors were
visited; and population of those in which more-than 72 neighbors were visited is in color ‘red’.
The scattered population was then divided into four quadrants to match the ‘analysis reference
value’. It can be observed that:

(v)

Quadrant A accommodated majority of simulation cycles in which 32 neighbors were
visited;

(vi)

Quadrant C clustered with a great portion simulation cycles in which more-72
neighbors were visited; and

(vii)

Quadrants B & D were populated by a combinations of simulation cycles – those with
optimal number of neighbors visited, and those with non-optimal.

93

Figure 5.7 – Skillfulness Level for Two Agents with Respect to Simulation Cycles Outputs

In order to better understand the influence of different levels of skillfulness on likelihood that
number of neighbors visited would deviate from its optimum, PDF curves were constructed.
Illustrated in Figure 5.8, numbers of neighbors visited were presented in terms of independent
input factor skillfulness. In compliance with analysis reference value, factor skillfulness was
divided into four levels – quadrants A, B, C, and D. Each curve represents a quadrant; and the area
under each curve defines the likelihood that a range of particular values would populate that
quadrant.

The range [32 , 40] was the case in more than fifty-two percent of simulation cycles. According to
Table 5.7, the likelihood that number of neighbors visited would fall within limits of this range

94

was the greatest for skillfulness values of quadrant A, the quadrants B and D. This likelihood was
the least for skillfulness values of quadrant C. In contrast, for the range (40 , 72] which according
to Table 5.7 was the number of neighbors visited also in (roughly) one-third of simulation cycles,
skillfulness values of quadrant C provided the greatest likelihood in comparison with those of
quadrants A (i.e., the least), B, and D.

Dotted vertical lines in color ‘black’ define the boundaries of these ranges. It should be noted since
density distributions for dependent outputs of quadrants B, and D are similar, only population of
outputs correspond with quadrants B was considered for further analyses.

Figure 5.8 – Number of Neighbors Visited with Respect to Skillfulness Level for Two Agents

95

With reference to Section (5.3), a rational explanation for clusters of values in quadrants A and C
of Figure 5.7 and also their greater likelihoods in Figure 5.8 would be the likelihood that n
decisions – two, in this case – at a given location were similar. Moreover, at any location the
likelihood of dissimilar decisions was greater for skillfulness values in quadrants B and D. The
latter explanation is based on the fact that those values were drawn from different sections of
skillfulness spectrum – value of skillfulness for one agent was greater than 0.50 while that of the
other agent was lesser than 0.50. This unlikelihood for similarity of decisions, unarguably
translates to greater likelihood of interaction between agents in process of determining the interim
destination.

It can be argued, in other words, that greater volumes of interactions between agents took place in
simulation cycles populated the quadrants B and D; and consequently a trigger for greater
involvement of factors confidence, and open-mindedness. Thus, further analyses focused on those
populations (i.e., curves in color ‘light green’ and ‘dark green’).

Take-away: Although layout of agent environment could play roles in similarity and dissimilarity
of decisions since neighbors’ availability was a requisite, nevertheless the more similar were levels
of skillfulness for each agent, the greater was likelihood that both agents would choose the same
neighbor.

Construction Work Crew Confidence and Open-mindedness
In Figure 5.9 probability density functions are presented with respect to different levels of
confidence and open-mindedness – given a particular range of skillfulness: the continuous curve
in color ‘black’ represents probability density for number of neighbors visited in simulation cycles

96

that skillfulness values were drawn from quadrant D. This PDF curve was then adjusted for
confidence and open-mindedness values of quadrants A, B, and C. It can be observed that
likelihood for number of neighbors visited to fall in range [32 , 40] was the least in magnitude
when confidence and open-minded values were withdrawn from quadrant C. This likelihood rose
for those values of confidence and open-mindedness withdrawn from quadrant B.

Figure 5.9 – Number of Neighbors Visited for Skillfulness Level ‘D’ with Respect to Confidence, &
Open-mindedness

One explanation for such behavior would be the intensity of interaction between agents. With
reference to Figure 5.3, different quadrants of values for skillfulness level were likely to result in
dissimilar decisions; and that in turn prompted interactions. Confidence and open-mindedness
97

levels from different quadrants intensified such interactions to a greater degree, which
consequently influenced outputs of the process. In contrast, the behavior of PDFs in
Figure 5.10 can be explained by lesser likelihood of rendering dissimilar decisions with
skillfulness values of quadrant C – which lessened the intensity of interactions between agents;
hence the smaller area under adjusted PDF curve for quadrant B values of confidence and openmindedness.

Figure 5.10 – Number of Neighbors Visited for Skillfulness Level ‘C’ with Respect to Confidence, &
Open-mindedness

Take-away: Dissimilarities of decisions agents rendered triggered interactions between input
factors; and these interactions in turn contributed to variations of outputs.
98

This section, so far, demonstrated that interactions between n agents – two in this case – would be
enabled by different levels of their skillfulness; that such interactions would be intensified based
on agents’ levels of confidence and open-mindedness; and that intensity of these interactions in
turn would influence output variables of this ABMS. In order to explore how influential each of
these three factors were – or rather to investigate how sensitive the outputs of ABMS were to these
three factor – or even how much each of these three factors contributed in variation of ABMS
outputs, Sensitivity Analysis was conducted.

Sensitivity Analysis (SA)
The main objective of analyses presented in this section was to explore sensitivity patterns induced
by input factors in ABMS outputs. SimLab software, as mentioned in Section (3.3.2), was
employed, and variations in outputs of 199,948 ABMS cycles were decomposed with association
to their corresponding input factors. Table 5.8 summarizes results of analyses in which values of
columns ‘First-Order Index’ and ‘Total-Order Index’ respectively represent the first and last
segments of the variance-based sensitivity analysis equation:

𝑉 = (∑𝑘𝑖 𝑉𝑖 ) + ( ∑𝑘𝑖<𝑗 𝑉𝑖𝑗 ) + (∑𝑘𝑖<𝑗<𝑚 𝑉𝑖𝑗𝑚 ) ................................................. Equation 5.2

99

Table 5.8 – Sensitivity Indexes for Variation of Output: Number of Neighbors Visited
Influential
Factor

First-Order
Index

Total-Order
Index

Difference
of Indexes

Skillfulness

28.68%

91.76%

63.08%

Confidence

8.50%

66.32%

57.82%

Open-Mindedness

8.77%

67.82%

59.05%

0.0035%

55.90%

55.89%

Independent

Interaction

45.95%

54.05%

Crew

Apportioned
Contributions

The first-order indexes values define the contributions of each input factors, when independently
fixed, in variation of outputs. The summation of these values explain how much of outputs
variations was caused by input factors, independently. The values of total-order indexes determine
overall contributions of each input factors in variations of outputs – that is, the contribution of each
input factors when independently fixed plus interactions of such factor with other ones. Table 5.9
discusses the importance of these values in terms of orders of their magnitudes.

100

Table 5.9 – Guideline for Interpretation of SimLab Sensitivity Analysis Output
Adopted from Source: (Ligmann-Zielinska & Sun, 2010)
SimLab Analysis Result
Interpretation
First-Order Index  A relatively great value for this measure represents a
corresponding significantly-influential input factor. Sum
of these indexes determines the portion of outputs
variations caused by input factors independently (i.e., no
interaction with other input factors)
Total-Order Index  A relatively small value for this measure represents a
corresponding non-influential/insignificantly influential
input factor.
Difference of a First- and  A relatively great value of difference represents a
a Total-Order Index
corresponding input factor with a significant intensity of
interaction with other input factors. Such input factor
influences variation of outputs through interactions with
other input factors, regardless of its significant/
insignificant independent influence.

Based on sensitivity indexes summarized in Table 5.8, it can be concluded that input factors
skillfulness, confidence, open-mindedness, and crew independently caused 45.95 percent of
outputs variations in this AMBS – with the most significant influence of 28.68 percent by factor
skillfulness; and that interactions between all four input factors, as presented in Figure 5.11,
contributed to more than a half of variations in the outputs – i.e., 54.05%. The fact that values for
first-order indexes summed less than 100 percent demonstrate the complexity of this model and
an additive correlation between impacts of these input factors on variations of model outputs.

101

Figure 5.11 – Independent Contribution of Input Factors to Outputs Variations: Number of Neighbors
Visited

Figure 5.12 – Contribution of Input Factors Interactions to Outputs Variations: Number of Neighbors
Visited

102

Moreover, considering the total-order index values it can be understood that although factors
confidence and open-mindedness did not independently influence model outputs by much, both
interacted with other factors and contributed to outputs variations. Similarly, factor crew-type
which represented ‘randomness’ in ABMS independently impacted outputs by close to zero
percent, and yet interacted with other factors to contribute in outputs variations. Intensities of such
interactions were revealed by the differences of sensitivity-index values: factor skillfulness
brought about the largest portion of interactions; then factors open-mindedness and confidence;
and factor crew-type interacted with other factors the least.

Comparable to those for output Number of Neighbors Visited, Table 5.10 summarizes sensitivity
indexes for output Rework. Based on values of first-order and total-order indexes it can be
concluded that the majority of variations in output – i.e., 54.63 percent – was caused by interactions
between input factors in which factor skillfulness led the intensity followed by factors openmindedness, confidence, and crew-type.

Furthermore, it can be understood that only 45.37 percent of variations in outputs might be
explained by independent contributions of each input factor – with the greatest contribution of
28.64 percent by factor skillfulness. The only notable difference between sensitivity indexes of
these two output variables was for input factor crew-type: although negligible in comparison to
others, its independent contributions to variations of outputs was increased by 60 percent from
0.0035 to 0.0056.

103

Table 5.10 – Sensitivity Indexes for Variation of Output: Rework
Influential
Factor

First-Order
Index

Total-Order
Index

Difference
of Indexes

Skillfulness

28.64%

92.09%

63.45%

Confidence

8.34%

65.90%

57.56%

Open-Mindedness

8.38%

66.20%

57.82%

0.0056%

52.05%

52.04%

Independent

Interaction

45.37%

54.63%

Crew

Apportioned
Contributions

As a pattern in this ABMS, interactions between input factors apportioned for greater contributions
than their independent influences in outputs variations. As Table 5.11 reports, this contribution of
57.65 percent was an increase of (about) 7 percent from those of outputs Number of Neighbors
Visited, and Rework. Intensity of interactions between input factors increased in case of this output
with a notable spike for crew-type – i.e., from 55.89 (for output number of neighbors visited) and
52.05 (for output rework) to 71.92 percent.

Independent influences of input factors confidence, and open-mindedness in variations of outputs
also experienced an increase of (roughly) 25 percent in comparison with those of other outputs.

104

Table 5.11 – Sensitivity Indexes for Variation of Output: Call for Assistance
Influential
Factor

First-Order
Index

Total-Order
Index

Difference
of Indexes

Skillfulness

21.18%

89.58%

68.40%

Confidence

10.53%

76.60%

66.07%

Open-Mindedness

10.34%

77.04%

66.70%

Crew

0.0042%

71.92%

71.92%

Independent

Interaction

42.05%

57.95%

Apportioned
Contributions

Figure 5.13 graphically compares the intensities of interactions between input factors for each
output variable.

Figure 5.13 – Comparison of Interactions Intensity of Input Factors with Respect to Outputs Variables

105

Number of Simulation Cycles
Figure 5.14 demonstrates the altering behaviors of first-order sensitivity indexes values for
numbers of simulation cycles between 500 and 300,000. The dotted, vertical line in color ‘black’
marks 110,000 simulation cycles; and those horizontal represent the range of 3.50 percent
difference in sensitivity indexes. As can be observed, values of each index relatively stabilized for
a number of 110,000 and greater simulation cycles.
The number of 199,948 repetitions – displayed by a solid, vertical line in color ‘dark green’ – was
selected since it roughly averaged the range of stabilized simulation cycles. Note that the number
of 199,948 receptions was in accord with the sample of values for input factors generated in
SimLab. As introduced in Section (3.3.1), software SimLab uses Monte Carlo random sampling
technique to generate pseudorandom numbers in some (user defined) iterative fashion. This sample
is then supplied as input factors for Uncertainty Analysis (UA) and Sensitivity Analysis (SA) of
models.

106

Figure 5.14 – Variations in Sensitivity Indexes with Respect to Number of Simulation Cycles

Take-away: Input factors contributed to variations in outputs not only independently, but also
through interactions with one another; and in case of this ABMS, the latter apportioned greater
contributions than former. Therefore, even though an input factor insignificantly influenced the
variations of outputs, its interactions with others were found significant and impactful in variations
of outputs.

107

(5.6)

Summary

In Chapter Five an agent-based model was developed based on conceptual frameworks of a
construction work crew’s Input-Process-Output (I-P-O) and cue-interpretation introduced in
Chapter Four. This ABM was translated into NetLogo programmable modeling environment, and
then simulated for 199,948 cycles with a numerical experiment designed in software SimLab.
Number of neighbors visited (i.e., operation time), re-visited (i.e., rework), and dead-end (i.e., callfor-assistance) were collected as outputs of simulation cycles. Exploratory statistical and
sensitivity analyses were conducted revealing that population size for agents, levels of skillfulness,
confidence, and open-mindedness contributed to variations of ABMS outputs; that greater
skillfulness, confidence, and open-mindedness increased the likelihood of visiting lesser number
of neighbors; and that this ABMS was the most sensitive to level of skillfulness, and then levels
of confidence and open-mindedness.

108

(6)

Conclusion

109

The problem addressed in this research dissertation was the interpretation of cues and physical
hints by a construction work crew in an ambiguous situation and where explicit directives were
absent. As explained in Chapter One, this investigation further posited that interactions between
members of a construction work crew would impact interpretation of cues and would thus impose
influences on work crew performances. In order to explore this conjecture the following objectives
were achieved:

(i)

Factors which would influence interactions between members of a construction work
crew were identified;

(ii)

(Reference) frameworks to conceptually address construction work crew interactions,
and cue-interpretation were developed; and

(iii)

An abstract agent-based model based on the conceptual reference frameworks was
developed and exploited for deriving insights.

To satisfy these aforementioned objectives, five (5) phases were undertaken: in Phase One
literature relevant to Small Group research was surveyed. As reported in Chapter Two, it was
understood that a group was defined to be:

(a)

a collection of two or more individuals;

(b)

who are functionally related;

(c)

who continuously engage in a direct (e.g., face-to-face) interaction(s); and

(d)

who purposively perform toward a common objective(s).

Focusing on the dynamics of a small group – and NOT its groupness – it was also learned that a
reference framework Input-Process-Output (I-P-O) had been used to conceptually present group
110

function; and that because previously-identified factors influencing a group process were almost
always case-specific, it was important to identify factors relevant to construction industry and
decision-making in presence of cues and physical hints.

Review of literature pertinent to construction work crew decision making identified the gap in
addressing improvisation versus cue interpretation – that is, in absence of explicit directives, and
when instructions were incomplete and information was imperfect, the work crews were known to
improvise; and this research dissertation argued that work crews also take cues, respond to physical
hints, and accordingly take actions.

In Phase Two qualitative interviews with a population of construction skilled trade workers
revealed that majority of work crews were likely to autonomously solve a problem caused by
ambiguity in workface – instead of contacting their supervisors, or consulting with their customers.
Also, as Chapter Four summarized, interviewees identified fourteen factors influential to their
decision-making processes in presence of surroundings cues and physical hints. The factors
Knowledge and Skill, Confidence and Trust, and Open-mindedness were the most-frequently
mentioned ones.

In Phase Three, a references frameworks were developed to conceptually present: (i) phenomenon
of cue interpretation in construction workface; and (ii) a work crew operation in process of
interpreting a cue and physical hint.
Phase Four focused on translating these frameworks into an agent-based modeling simulation –
ABMS.

111

In Phase Five, ABMS experiments were conducted for a total of 199,948 cycles in which a series
of three outputs were determined: (i) number of neighbors visited; (ii) number of neighbors revisited; and number of dead-end neighbors reached. Results of exploratory statistical analyses and
sensitivity analyses (SA) were reported in Chapter Five and are discussed in the next Section (6.1).

(6.1)

Discussion

The analyses presented in Chapter Five were conducted to explore interactions between work crew
members and to better understand influences such interactions impose on interpretations of cues
and physical hints. In order to discuss these findings with respect to this research dissertation the
following research question is considered:

Question: How can these analytical results be interpreted with respect to interpretation of cues in
a construction work crew?

It was explored whether being a group contributed to variations of ABMS outputs. Results of
analyses demonstrated that levels of uncertainties in model outputs would vary for work crews of
different size (i.e., number of members). This variation is an implication of some processes or
rather interactions between members of a crew, and it does not imply that a construction work
crew of one member may be extra efficient, more productive, and/or likely successful than that of
two or more members in accurately interpreting the “what-to do’s” based on surroundings cues
and physical hints. In other words, (activity) performance would be negatively impacted by
presence interpretation of cues and physical hints, even though it appears to improve with
increasing crew size.

112

As Figure 6.1 suggests, a crew of larger size would experience a better performance in variation
of operation time due to greater volume of interactions between crew members in decision-making
and cue-interpretation. However, it does not necessarily improve the performance mean.

Figure 6.1 – Average Operation Time and Variation Performance with Respect to Crew Size

Crew size facilitates the intensity of interactions between crew members and, consequently,
contributes to predictability and reliability of work flow. Nevertheless, interaction is not enough
to eliminate variations in performance and/or work flow. In the meantime, it is important to note
that increasing size of a crew may not be advised in solitary and with no considerations of technical
necessities, and adverse impacts of crowding phenomenon.

The analyses also showed that, with respect to level of skillfulness, a homogeneous group rendered
(more) similar decisions; and that heterogeneity of skillfulness values propagated greater
113

uncertainties in distributions of outputs. Skillfulness in this investigation symbolized the
knowledge of how-to technically (i.e., from a technical stand point) accomplish a task. The
knowledge about what-to do considered in terms of directives (e.g., pieces of information and
instructions) – the absence of which this research argues is the trigger for cue-interpretation. Those
in considerations, the analyses can imply that a construction work crew of comparable (i.e.,
homogeneous) skillfulness levels may complete identical tasks more efficiently in comparison
with a heterogeneously-skilled crew. Homogeneity of crew members in skillfulness is argued to
result in similar decisions which in turn decreases not only the intensity of interactions but also
their impacts. That is, crew members take similar cues and/or interpret them similarly in order to
determine the “what-to do’s”; which eliminate great intensities of interactions to introduce the
final decision.

On the other hand, it may be inferred from these analytical results that heterogeneity of skillfulness
enables and encourages work crew members to interact for rendering the final decision as each
member takes a different cue and/or render a different interpretation of that – which in turn
becomes more influential. It can be further inferred that the quality of interactions output depends
on confidence and open-mindedness levels of each construction work crew member: the more the
work crew members are heterogeneous in confidence and/or open-mindedness, the greater
influences their interactions impose.

Sensitivity analyses of ABMS outputs, as compared in Figure 6.2, revealed that skillfulness of
crew members may be introduced as a factor which will need more attention since it strongly
impacts amount of rework and waste, operation time, and efficiency in resources utilization, and
also frequency of call-for-assistance, respectively. As previously mentioned, one side of

114

knowledge and skills is about directives (information and instruction), lack of which invites cues
and surroundings physical hints in determination of what-to do. This is, thus, aligned with the
problem stated in Section (1.3) since the more complete and more explicit is a directive, the more
efficiently it is executed; the less amount of waste is generated; and the fewer occasions when
supervisory-assistance is required. It is of value to note that how-to do represents the technical
aspect of knowledge and skillfulness – which greatly influences amount of rework and waste, and
frequency of call-for-assistance. These in turn will impact operation time due to lack of knowledge
on how to complete a task.

115

Figure 6.2 – Comparison of Independent Contributions of Input Factors with Respect to Outputs
Variables

Open-mindedness, that is willingness of a work crew member to consider others’ opinions, can be
identified, based on the sensitivity indexes, as the next most-important factor in frequencies of
call-for-assistance. It is of lesser impact on operation time, and is the least influential on causing
rework and waste. A stubborn crew member may be seen, to insist, for example on her/his own
interpretation of a cue which, depending on her/his skillfulness, can deviate from accuracy. That
may cause rework, and lead to longer operation time/inefficient utilization of resources.

The confidence in ability, and reliability of a crew member seems to impose the least impact among
the three factors for amount of rework and waste as the contributions of sensitivity indexes in
Figure 6.2 reports. Confidence can be introduced as the next most-influential factor in frequency
of call-for-assistance, and operation time/efficiency in resources utilization. It can be inferred that
lesser confidence that a crew-mate will evaluate the situation appropriately and, when needed, will
116

interpret a cue accurately increases the likelihood of call to the supervisor for assistance to seek
(more) explicit directives. Greater level of confidence in crew-mate, on the other hand, contributes
to a longer and less efficient operation time/resource utilization, and may also lead to some amount
of rework depending on how skillful that crew-mate is.

These effects will be more meaningful and sensible when considered in combination with one
another.

Take-away: Interactions between members of a homogeneous, in comparison with a
heterogeneous, construction work crew results in outputs which may deviate less from expected
ones.

The reality, however, is that a work crew is what it is! That is, with the fair assumption that
members of a construction work crew are trained in a specific discipline, and are knowledgeable
about how-to do their assigned activities, it is arguable that traits they bring to the crew will not
change nor can it be altered. And as mentioned previously, exploratory and sensitivity analyses
concluded that heterogeneity of such traits drives their interactions. Thus, it may be argued that
those interactions may be directed towards positive influences. Lean Construction principles
embodied in tools such as The Last PlannerÂŽ System (LPSÂŽ) of Production Planning and Control
may be named as appropriate means. As was referred to in Chapter Two Section (2.4), the LPSÂŽ
mandates that an activity to be definable, sound, in sequence, and appropriately sizeable. LPSÂŽ
declare soundness as a state in which every requirement an activity needs are available.
Soundness criteria can be particularly emphasized as important for two reasons: first – it ensures
that directives an activity needs are present, explicit, and complete. Therefore, what-to do is clearly

117

known and there will be minimal likelihood – if any – of cue interpretations. Second – it channels,
directs, and focuses the interactions between crew members towards an advantageous output – i.e.,
making an activity ready for execution
Either it is in the make-ready stage or towards the weekly-work planning, two phases of LPSÂŽ, the
soundness criteria unarguably minimizes ambiguity, and serves as a channel to effectively employ
the inevitable interactions between crew members so implicitness is uncovered. In other words, it
minimizes – if not eliminate – the iterative nature of cue interpretation illustrated in Figure 6.3,
generation of waste, and inefficiency of operation.

No

Interpret
missing*
information

Complete
the task

Was the
Interpretation
correct?

Yes

No

Undo the task
and start over

Additional resource is lost
to problem-solve &
undoing the wrong

Figure 6.3 – Schematic Presentation of Iterative Process of Cue Interpretation
(i.e., adopted from Figure 4.2)

118

(6.2)

Contribution

Illustrated in Chapter Three Figure 3.1, contributions of this research dissertation to the body of
construction-industry knowledge can be summarized as follows:

Influential Factors on Cue Interpretation
Instrumenting a framework for qualitative interviews as Chapter Four discussed in details, fourteen
implied factors were identified by construction skilled trade workers. These factors, summarized
in Table 6.1, signified by one hundred and seventy six (176) construction work crew members as
influential in ambiguous situations and where decisions are to be made in presence of cues and
physical hints.

Table 6.1 – Frequency Analyses of Influencing Factors on Work Crew Members Interactions

Knowledge and Skills

Frequency of
Implication
22.1%

Confidence and Trust

14.4%

Open-mindedness

12.9%

Morale and Personality

10.3%

Cost and Efficiency

8.7%

Communication

7.7%

Mutual Respect

7.5%

Common Sense

5.6%

Complexity of Task

4.2%

Product Quality

1.9%

Customer Satisfaction

1.7%

Work Style

1.4%

Collaboration

0.8%

Feasibility of Solution

0.7%

Influential Factor

119

Contents of qualitative interviews also implied that, as reported in Figure 6.4, crews were more
eager to autonomously resolve such ambiguity and most-likely by discussing the issue with their
crew-mates instead of contacting their supervisors, or customers.

Figure 6.4 – Overall-Frequency Analyses of Work Crews Immediate Actions in Ambiguous Situations

Conceptual Reference Framework of Cue Interpretation
Presented in Chapter Four, a reference framework was developed to conceptually describe the
phenomenon of cue-interpretation in a construction work crew. This conceptual framework was
developed based on contents of qualitative interviews with 176 construction skilled trades workers.
Concept of cue interpretation in construction job site and in absence of complete instruction and
perfect information was introduced by Lahouti and Abdelhamid (2012).

120

Abstract Agent-based Model
This reference framework was then employed to develop an agent-based model in Chapter Five.
The ABM was translated in programmable modeling environment of NetLogo (Wilensky, 1999);
and was further developed to simulate dynamics (e.g., interaction) of a construction work crew to
explore how interactions impact cue-interpretation. Sensitivity analyses of simulations outputs
identified most-influential factors on interactions of crew members in case of cue-interpretation.

Overall Contribution
This investigation as a whole offers a computational validation of the quality-tasks concept, and
constraint screening utilized in the Last PlannerÂŽ System. The LPSÂŽ ability to prevent planning
failures is undisputed given its mounting empirical evidence of the past quarter-century. The
conclusion is that ‘it works.’ Why does it work has still not been adequately answered. This work
has demonstrated a principled explanation that leads to a stronger understanding for why the LPSÂŽ
works in maintaining reliable and predictable workflow on construction sites. The core of LPSÂŽ
summarizes that:

(a)

execution, and planning should not be separated; and

(b)

projects are not scripted performance – rather, they are ongoing conversations.

In the absence of the routines and practices of the LPSÂŽ, the crew is not shielded from the damaging
effects, of having to interpret cues and physical feature of workplaces as proven in this research.
The cost is manifested in the delays caused by waiting for directive clarifications, and possibility
of rework.
121

Moreover, this research dissertation provides a sound basis for crew member recruitment, and crew
make-up. Skill is an intuitive requirement. However, it is worthwhile to interview and screen for
the additional traits of confidence, and open-mindedness.

Take-away: Interaction between members of work crew is important in activity performance. Or
in “layman’s terms”, it is crucial for work crew to discuss and finalize its plan for action before
start the task.

(6.3)

Limitation

With reference to Section (4.2) Figure 4.1 and from four types of inputs factors, which have been
introduced in literature as influential on outputs of a work crew, only factors of type “individual”
were modeled in the ABMS of this investigation. Such consequently assumes no influence or
rather fixed contribution from other types of inputs factors – which indeed is a limitation of this
investigation.

The literature relevant to small group research (to the extent surveyed in process of this
investigation) did not define a unique mathematical representation for correlation between
behavioral characteristics of group members. Therefore, a best-fit correlation was defined to
describe interactions between two factors – Confidence and Open-mindedness. As Chapter Five
Section (5.3) discussed the product (multiplication) was presented as an attempt of approximation
in accordance with the literature best-fit approach.

Also discussed in Chapters Three and Five, real-life data was not available for factors introduced
to ABMS; and therefore the simulation was executed with limited experiments designed based on

122

probabilistically distributed values. Indeed future development of this research would benefit from
distributions of collected data.
The potential artifact imposed by the specific layout of (developed) ABMS – graphically presented
in Figure 3.2 and Figure 5.1 – can also be nominated as a limitation in this research. That is,
although sensitivities of simulation outputs would be the same in quality to the input factors, their
orders of magnitudes could be influenced by a specific layout.

(6.4)

Future Research

It was previously concluded that the attention of this investigation was directed towards
interactions between members of a construction work crew as representation of a small group; and
accordingly the Input-Process-Output (I-P-O) reference framework of Figure 4.1 was adapted from
small group research literature to illustrate such interactions. It is believed by the researcher of this
dissertation topic that Social Network Analysis (SNA) process of investigating social structures
(e.g., a work crew) will enhance this work as SNA has the capabilities to shed lights on impacts of
inputs factors on one another – one instance for the case of this research dissertation will be the
impacts of interactions between input factors “Confidence and Trust” and “Open-mindedness” on
variations in outputs.

As discussed in Chapters Three and Five, ABMS cycles did not represent the concept of past
experience. In other words, agents did not have a learning attribute. z simulation cycles presented
z groups of two agents each completed the same activity. Considering that an agent’s confidence
in another agent can be influenced by past experience – as is likely in a real-life situation – it will
be a major development to correlate level of confidence with respect to decisions made by the

123

corresponding agent. Enabling an agent to learn, and to carry-over its experience from one cycle
to another was not within in the scope of this investigation; and is definitely a potential for future
development.

Also mentioned in Chapter Five that visiting thirty two (32) neighbors to reach the destination
(neighbor) exemplified the most efficient performance. Such conclusion was based on the
assumption that visiting each neighbor (i.e., completing each tasks) would demand the same
amount of time – in other words, different tasks were of equal durations. Relaxing such assumption
will be a development for the future of this research in which different tasks will have completion
time apportioned to their complexity.

124

APPENDICES

125

APPENDIX A: Framework for Qualitative Interview

Decision-Making Process Heuristic among Construction Work Crew
INTERVIEW GENERAL INFORMATION
Number: _ _ _ _ _ _

Date: _ _ _ _ _ _

Venue: _ _ _ _ _ _

WORK CREW (ANONYMOUS) INFORMATION
Skilled Trade Discipline:
__________

Current Title / Rank:
__________

(Total) Number of Years of Experience:
__________

Primary Category of Expertise:
__________

Typical Structure of Work Crew:

_ _ _ _ _ _ members _ _ _ _ _ _ + _ _ _ _ _ _

Situation:
When you arrived on the job site, you realized that the situation did not match the instructions
(e.g., construction drawing; verbal request; work-order; etc.) originally received; and you needed
to acquire more information or informative instruction in order to complete your activity/task. In
the meantime, there was a readily-available physical cue at the work face which might have –
rightly or wrongly – filled in for the missing pieces of information:

DECISION-MAKING AT PRESENCE OF PHYSICAL CUE(S)
Since you often work as a crew, how will you most likely respond / react to such a situation?
__________
Will you please briefly elaborate?
__________
Is there often an alternative task for you if your planned task cannot be completed?
______

126

EXAMPLE OF A DECISION-MAKING
Activity Information: _ _ _ _ _ _ _ _ _ _
Description of Situation:
__________
‘WHAT’ did you do: _ _ _ _ _ _ _ _ _ _
‘HOW’ did you decide on the ‘WHAT’ to do?
__________
‘WHY’ did you decide on that “WHAT’? _ _ _ _ _ _ _ _ _ _
Was that approach the best possible one: _ _ _ _ _ _ _ _ _ _
How did the actual outcome of decision compare with the anticipated / expected result?
__________

MORAL OF THE STORY
In your opinion, did ‘HOW’ [i.e., the process] you made a decision impose an impact on the
task performance?
__________
What are the factors that have influenced [in descending order] ‘HOW’ you made decisions in
similar situations [on operation level] in years of your experience on various projects?
______
______
______
______
______
______

CRITICAL LESSEN(S) LEARNED / KEY TAKEAWAY(S)
__________

127

APPENDIX B: Agent-based Model

NetLogo Programmable Environment Code

To Setup
No-Display
Clear-All
Clear-Patches
Ask Patch 17 19 [Set pColor Red]
Create-Turtles 1 [
No-Display
Set Call 0
Set Color Blue
Set Counter 0
Set per-Level 0
Set ReVisit [ ]
Set ReWork [ ]
Set Supervisor [ ]
Set Step 0
Set Visit [ ]
Move-To Departure]
Display
Reset-Ticks
End
To Find-Shortest-Path-to-Destination
Set Path Find-A-Path Patch-Here Patch X Y
Set Optimal-Path Path
Set Current-Path Path
End
To-Report Find-a-Path
[Short-Departure Short-Destination]
Let Search-Done? False
Let Search-Path [ ]
Let Current-Patch 0
Set Open [ ]
Set Closed [ ]
Set Open lPut Short-Departure Open
While [Search-Done? != True ]
[IfElse Length Open != 0
[Set Open Sort-By [[F] of ?1 < [ F ] of ?2 ] Open
128

Set Current-Patch Item 0 Open
Set Open Remove-Item 0 Open
Set Closed lPut Current-Patch Closed
Ask Current-Patch
[IfElse Any? Neighbors4 with [(pXCor = [pXCor] of Short-Destination) And (pYCor =
[pYCor] of Short-Destination)]
[Set Search-Done? True]
[Ask Neighbors4 with [pColor != Black And (Not Member? Self Closed) And (Self != ParentPatch)]
[If Not Member? Self Open And Self != Short-Departure And Self != Short-Destination
[Set Open lPut Self Open
Set Parent-Patch Current-Patch
Set G [G ] of Parent-Patch + 1
Set H Distance Short-Destination
Set F (G + H)]]]]]
[User-Message ( "No Patch Exists !" )
Report [ ]]]
Set Search-Path lPut Current-Patch Search-Path
Let Temp First Search-Path
While [Temp != Short-Departure]
[Ask Temp
[Set pColor 2]
Set Search-Path lPut [Parent-Patch] of Temp Search-Path
Set Temp [Parent-Patch] of Temp]
Set Search-Path fPut Short-Destination Search-Path
Set Search-Path Reverse Search-Path
Report Search-Path
End
To Re-Run
Set Color Blue
Set Heading One-of [0 90 180 270]
Set ReVisit [ ]
Set Step 0
Set Visit [ ]
Move-To Departure
End
To Factor-Confidence-&-Trust
Set Confidence-Trust [ ]
ForEach (n-Values Crew-Size [? + Crew-Size])
[Set Confidence-Trust lPut Confidence-Trust Confidence-Trust]
End
To Factor-Knowledge-&-Skill
Set Current-MoveTo-Patch [ ]
129

Set MoveTo-Patch-Crew [ ]
Set Crew ( n-Values Crew-Size [1] )
Set Current-Knowledge-Skills [ ]
Set Knowledge-Skill [ ]
ForEach (n-Values Crew-Size [?])
[Set Knowledge-Skill lPut Knowledge-Skill Knowledge-Skill]
Repeat Crew-Size [
Let Temporary-Knowledge-Skills (Random-Float 1)
Set Current-Knowledge-Skills Sentence Current-Knowledge-Skills Temporary-KnowledgeSkills]
ForEach Current-Knowledge-Skills
[IfElse ? > (Item (Position ? Current-Knowledge-Skills) Knowledge-Skill)
[Let Index (Position ? Current-Knowledge-Skills)
Find-Available-Patch
Find-HeadingTo-Patch
If Empty? Current-MoveTo-Patch
[Set Current-MoveTo-Patch "N/A"
Set Crew Replace-Item Index Crew (10 ^ 3)]
Set MoveTo-Patch-Crew Sentence MoveTo-Patch-Crew Current-MoveTo-Patch]
[Set MoveTo-Patch-Crew Sentence MoveTo-Patch-Crew (Item 1 Optimal-Path)]]
End
To Factor-Open-Mindedness
Set Open-Mindedness [ ]
ForEach (n-Values Crew-Size [? + (Crew-Size * 2)])
[Set Open-Mindedness lPut (Item ? Factor-Data) Open-Mindedness]
End
To Find-Available-Patch
Set Available [ ]
Set Current-MoveTo-Patch [ ]
Set Neighbor [0 90 180 270]
Set Sign [ ]
Set Target [ ]
ForEach Neighbor
[IfElse NoBody = Patch-At-Heading-And-Distance ? 1
[Set Neighbor Remove ? Neighbor]
[If (Black != [pColor] of Patch-At-Heading-And-Distance ? 1)
[Set Available Sentence Available ?]]]
ForEach Available
[IfElse Member? (List [pXCor] of Patch-At-Heading-And-Distance ? 1 [pYCor] of Patch-AtHeading-And-Distance ? 1) Visit
[Set Available Remove ? Available]
[If Red = [pColor] of Patch-At-Heading-And-Distance ? 1
[Set Target Sentence Target Patch-At-Heading-And-Distance ? 1
Set Current-MoveTo-Patch Target]
130

If (Green = [pColor] of Patch-At-Heading-And-Distance ? 1 And Empty? Target)
[Set Sign Sentence Sign Patch-At-Heading-And-Distance ? 1
Set Current-MoveTo-Patch Sign]]]
ForEach Available
[Set Available Replace-Item (Position ? Available) Available Patch-At-Heading-And-Distance ?
1]
End
To Find-HeadingTo-Patch
If (Empty? Current-MoveTo-Patch And Not Empty? Available)
[Set Current-MoveTo-Patch (List One-of Available)]
If (Empty? Current-MoveTo-Patch And Empty? Available)
[Escape-Dead-End]
End
To Log-Visit-Patch
Let Location (List ([pXCor] of Patch-Here) ([pYCor] of Patch-Here))
If (Member? Location Visit And Not Member? Location ReVisit)
[Set ReVisit Sentence ReVisit (List Location)]
Set Visit Sentence Visit (List Location)
Set Visit Remove-Duplicates Visit
End
To Escape-Dead-End
Set Available [0 90 180 270]
ForEach Available
[IfElse (NoBody = Patch-At-Heading-And-Distance ? 1
Or Black = [pColor] of Patch-At-Heading-And-Distance ? 1
Or Member? (List [pXCor] of Patch-At-Heading-And-Distance ? 1 [pYCor] of Patch-AtHeading-And-Distance ? 1) ReVisit)
[Set Available Remove ? Available]
[Set Current-MoveTo-Patch Sentence Current-MoveTo-Patch Patch-At-Heading-And-Distance
? 1]]
If (Length Current-MoveTo-Patch > 1)
[Set Current-MoveTo-Patch (List One-of Current-MoveTo-Patch)]
End
To Contact-Supervisor
Set Call Call + (1)
Log-Visit-Patch
Set Departure Patch 3 1
Set Visit [ ]
Move-To Departure
End
To Apply-Factors
131

Factor-Knowledge-&-Skill
Factor-Confidence-&-Trust
Factor-Open-Mindedness
Aggregate-Factors
End
To Aggregate-Factors
Set Performance-MoveTo-Patch-Crew [ ]
Set Performance-MoveTo-Patch-Crew Remove-Duplicates MoveTo-Patch-Crew
Let Weight (Map [?1 * ?2 * ?3 ] Crew Confidence-Trust Open-Mindedness )
Set Move-Factors [ ]
Set MoveTo-Patch-Crew (List (Item (Position (Max Weight) Weight) MoveTo-Patch-Crew))
Set MoveTo-Patch-Crew (List (Item (Position (Min Weight) Weight) MoveTo-Patch-Crew))
Set MoveTo-Patch-Crew Remove "N/A" MoveTo-Patch-Crew
ForEach (List Position ( Min Weight ) Weight )
[Set Move-Factors lPut Item ? Knowledge-Skill Move-Factors
Set Move-Factors lPut Item ? Confidence-Trust Move-Factors
Set Move-Factors lPut Item ? Open-Mindedness Move-Factors]
End
To Move
IfElse Empty? MoveTo-Patch-Crew
[Contact-Supervisor]
[Measure-Performance
Move-To One-of MoveTo-Patch-Crew
Set Step Step + (1)]
ForEach Optimal-Path
[If (Position ? Optimal-Path) > 0 And (Position ? Optimal-Path) < (Length Optimal-Path - 1)
[Ask ?
[Set pColor White]]]
End
To Count-A-Move
Set Collect-Step Sentence Collect-Step Step
Set per-Level per-Level + 1
IfElse (per-Level = Iteration)
[Let M-Move Max Collect-Step
Let N-Move Round Mean Collect-Step
Let R-Move Length ReVisit
Set Average Sentence Average N-Move
Set Supervisor Sentence Supervisor Call
Set Maximum Sentence Maximum M-Move
Set ReWork Sentence ReWork R-Move
Set Collect-Step [ ]
Set Counter (Counter + 1)
Set Call 0
132

Set per-Level 0
Set Performance [0 0 0 0]
Re-Run]
[Re-Run]
End
To Make-Decision
Ask Turtles
[IfElse Patch-here = Destination
[Count-A-Move]
[Log-Visit-Patch
Find-Shortest-Path-to-Destination
Apply-Factors
Move]]
Tick
End

133

BIBLIOGRAPHY

134

BIBLIOGRAPHY
Abdelhamid, T. S. (2011). Variation in Production : AGC's Lean Construction Education
Program - Unit 1. Arlington, VA: The Assciated General Contractors of America.
Ahn, S. (2014). Construction Workers' Absence Behavior Under Social Influence. Doctoral
Disseration - University of Michigan. Ann Arbor, MI, U.S.A.
Arashpour, M., Wakefield, R., Blismas, N., & Lee, E. W. (2013). Analysis of Disruptions Caused
by Construction Field Rework on Productivity in Residential Projects. Journal of
Construction Engineering and Management, 140(2).
Arrow, H., McGrath, J. E., & Berdahl, J. L. (2000). Small Groups as Complex Systems: Formation,
Coordination, Development, and Adaptation. Thousand Oaks, CA: Sage Publications, Inc.
Babbie, E. (2010). The Practice of Social Research - Twelfth Edition. Belmont, CA: Wadsworth
Cenage Learning.
Bales, R. F. (1950). A Set of Categories for the Analysis of Small Group Interaction. American
Sociological Review, 15(2), 257-263.
Bales, R. F. (1950). Interaction Process Analysis. Cambridged, MA: Addison-Wesley Press, Inc.
Ballard, H. G. (2000). The Last Planner System of Production Control. University of Birmingham,
United Kingdom.
Ballard, H. G., & Howell, G. A. (1997). Implementing Lean Construction: Stabilizing Work Flow.
In L. Alarcon (Ed.), Lean Construction (pp. 101-110). Rotterdam, The Netherlands: A. A.
Balkema Publishers.
Ballard, H. G., & Howell, G. A. (1998). Shielding Production: An Essential Step in Production
Control. Journal of Construction Engineering and Management, 11-17.
Bankes, S., Lempert, R., & Popper, S. (2002). Making Computational Social Science Effective:
Epistemology, Methodology, and Technology. Social Science Computer Review, 377-388.
Bennis, W. G., & Biederman, P. W. (1997). Organizing Genius: The Secret of Creative
Collaboration. Cambridge, MA: Perseus Books.
Bijlsma, K., & Koopman, P. (2003). Introduction: Trust within Organizations. Personnel Review,
32(5), 543-555.
Bonabeau, E. (2002). Agent-Based Modeling: Methods and Techniques for Simulation Human
Systems. 99(3), 7280-7287.
Brodbeck, M. (1958). Methodological Individualisims: Definition and Reduction. Philosophy of
Science, 25(1), 1-22.
135

Burke, P. J. (2006). Interaction in Small Groups. In J. D. Delamater, Handbook of Psychology (pp.
363-387). Boston, MA: Springer.
Cartwright, D., & Zander, A. (1968). Groups and Group Membership : Introduction. In D.
Cartwright, & A. Zander (Eds.), Group Dynamics: Research and Theory (pp. 45-62). New
York, NY: Harper & Row, Publishers.
Castiglione, F. (2006). Agent-Based Modeling, 9. Retrieved June 29, 2012, from Scholarpedia:
http://www.scholarpedia.org/article/Agent_based_modeling
Chiocchio, F., Forgues, D., Paradis, D., & Iordanova, I. (2011). Teamwork in Integrated Design
Projects: Understanding the Effects of Trust, Conflict, and Collaboration on Performance.
Project Management Journal, 42(6), 78-91.
Chua, D. K., Kog, K. H., & Loh, P. K. (1999). Critical Success Factors for Different Project
Objectives. Journal of Construction Engineering and Management, 125(3), 142-150.
Ciborra, C. U. (1999). Notes on Improvisation and Time in Organizations. Accounting,
Management and Information Technologies, 9(2), 77-94.
Crosbie, P. V. (1975). Interaction in Small Groups. New York, NY: MacMillanPublishing
Company, Inc.
Cunha, M. P. (2005). Bricolage in Organizations. Lisboa, Portugal: Universidade Nova de Lisboa.
Cunha, M. P., Cunha, J. V., & Kamoche, K. (1999). Organizational Improvisation: What, When,
How, and Why. International Journal of Management Reviews, 1(3), 299-341.
Dai, J., Goodrum, P. M., & Maloney, W. F. (2009). Construction Craft Workers' Perceptions of
the Factors Affecting Their Productivity. Journal of Construction Engineering and
Management, 138(3), 217-226.
Davis, J. H. (1969). Group Performance. Reading, MA: Addison-Wesley Publishing Company.
Desai, A. P., & Abdelhamid, T. S. (2012). Exploring Crew Behavior During Uncertain Jobsite
Conditions. Proceedins for 20th Annual Conference of the International Group for Lean
Construction. San Diego, CA.
Dozzi, S. P., & Abourizk, S. M. (1993). Productivity in Construction. Institute for Research in
Construction. Ottawa: Natinal Research Council Canada.
Du, J. (2012). Investigation of Interpersonal Cooperation in Construction Project Teams: An
Agent-Based Modeling Appraoch. Doctoral Disseration - School of Planning, Design, &
Construction; Michigan State Univesity. East Lansing, MI, U.S.A.
Eden, C., Williams, T., Ackermann, F., & Howick, S. (2000). The Role of Feedback Dynamocs in
Disruption and Delay on the Nature pf Disruption and Delay (D&D) in Major Projects.
Journal of the Operational Research Society, 51(3), 291-300.
136

Foley, J., & MacMillan, S. (2005). Patterns of Interaction in Construction Team. CoDesign, 1(1),
19-37. doi:10.1080/15710880412331289926
Formoso, C. T., Sommer, L., Koskela, L., & Isatto, E. L. (2011). An Exploratory Study on the
Measurement and Analysis of Making-Do in Construction Sites. Proceedins for 19th
Annual Conference of the International Group for Lean Construction, (pp. 13-15). Lima,
Peru.
Forsyth, D. R. (2014). Group Dynamics (Sixth ed.). Belmont, CA: Wadsworth Publishing Cengage Learning.
Fortino, G., & North, M. J. (2013). Simulation-based Development and Validation of Multi-agent
Systems: AOSE and ABMS Approaches. Journal of Simulation, 7, 137-143.
GalĂĄn, J. M., Izquierdo, L. R., Izquierdo, S. S., Santos, J. I., Olmo, R., LĂłpez-Paredes, A., &
Edmonds, B. (2009). Errors and Artefacts in Agent-Based Modeling. Journal of Artificial
Societies
and
Social
Simulation,
12(11).
Retrieved
from
http://jasss.soc.surrey.ac.uk/12/1/1.html
Goldberg, L. R. (1990). An Alternative Description of Personality: The Big-Five Factor Structure.
Journal of Personality and Social Psychology, 59(6), 1216-1229.
Gomes, D., Tzortzopoulos, P., & Kagioglou, M. (2014). Collaboration Through Shared
Understanding in Early Design Stage. Proceedins for 24th Annual Conference of the
International Group for Lean Construction, (pp. 63-72). Boston, MA.
Gürcan, Ö., Dikenelli, O., & Bernon, C. (2013). A Generic Testing Framework for Agent-based
Simulation Models. Journal of Simulation, 7, 183-201.
Hackman, J. R. (1968). Effects of Task Characteristics on Group Products. Journal of
Experimental Psychology, 4, 162-187.
Hackman, J. R., & Katz, N. (2010). Group Behavior and Performance. In S. T. Fiske, D. T. Gilbert,
& G. Lindzey, Handbook of Social Psychology (Fifth ed., Vol. 2, pp. 1208-1251).
Hoboken, NJ: John Wiley & Sons, Inc.
Hackman, J. R., & Morris, C. G. (1975). Group Tasks, Group Interaction Process, and Group
Performance Effectiveness: Review and Proposed Integration. Advances in Experimental
Psychology, 8, 45-99.
Hackman, J. R., Brousseau, K. R., & Weiss, J. A. (1976). The Interaction of Task Design and
Group Performance Strategies in Determining Group Effectiveness. Organizational
Behavior and Human Performance, 16, 350-265.
Halligan, D. W., Demsetz, L. A., & Brown, J. D. (1994). Action-Response Model and Loss of
Productivity in Construction. Journal of Construction Engineering and Management,
120(1), 47-64.

137

Hamzeh, F. R., Morshed, F. A., Jalwan, H., & Saab, I. (2012). Is Improvisation Compatible with
Look-Ahead Planning? An Exploratory Study. In I. D. Tommelein, & C. Pasquire (Ed.),
Proceedins for 20th Annual Conference of the International Group for Lean Construction.
San Diego, CA.
Handa, V. K., & Rivers, D. (1983). Downgrading Construction Incidents. Journal of Construction
Engineering and Management, 109(2), 190-205.
Homma, T., & Saltelli, A. (1996). Importance Measures in Global Sensitivity Analysis of NonLinear Models. Reliability Engineering and System Safety, 52, 1-17.
Hopp, W. J., & Spearman, M. L. (2008). Factory Phsyics. Long Grove, IL: Waveland Press, Inc.
Howell, G. A., & Ballard, H. G. (1997). Implementing Lean Construction: Reducing Inflow
Variation. In L. Alarcon (Ed.), Lean Construction (pp. 93-100). Rotterdam, The
Netherlands: A. A. Balkema Publishers.
Howell, G. A., Ballard, H. G., Tommelein, I. D., & Koskela, L. (2004). Discussion of ‘‘Reducing
Variability to Improve Performance as a Lean Construction Principle". Journal of
Construction Engineering and Management, 130(2), 299-300.
Hsu, W., & Liu, B. (2000). Conceptual Design: Issues and Challenges. Computer-Aided Design,
32, 849-850.
Ilgen, D. R., Hollenback, J. R., Johnson, M., & Jundt, D. (2005). Teams in Organizations: From
Input-Process-Output Models to IMOI Models. Annual Review of Psychology, 56, 517543.
Jackson, R. L. (2010). The Encyclopedia of Indentity. Thousand Oaks, CA: SAGE Publications.
James, J. (1951). A Preliminary Study of the Size Determinant in Small Group Interaction.
American Sociological Review, 16(4), 474-477.
Johnson, D. W., & Johnson, R. T. (1998). Cooperative Learning and Social Interdependence
Theory. In R. S. Tindale, L. Heath, J. Edwards, E. J. Posavac, F. B. Bryant, Y. SuarezBalcazar, . . . J. Myers, Theory and Research on Small Groups (pp. 9-35). New York, NY:
Plenum Press.
Jones, M. B. (1974). Regressing Group on Individual Effectiveness. Organizational Behavior and
Human Performance, 11(3), 426-451.
Kaming, P. F., Olomolaiye, P. O., Holt, G. D., & Harris, F. C. (1997). Factors Influencing
Craftmen's Productivity in Indonesia. International Journal of Project Management, 15(1),
21-30.
Katzell, R. A., Miller, C. E., Rotter, N. G., & Venet, T. G. (1970). Effects of Leadership and Other
Inputs on Group Processes and Output. The Journal of Social Psychology, 80, 157-169.

138

Katzenbach, J. R., & Smith, D. K. (1993). The Wisdom of Teams: Creating High-Performance
Organizations. Boston, MA: McKinsey & Company, Inc.
Kent, R. N., & McGrath, J. E. (1969). Task and Group Characteristics as Factors Influencing
Group Performance. Journal of Experimental Social Psychology, 5, 429-440.
Kerr, N. L., & Tindale, R. S. (2004). Group Performance and Decision Making. The Annual Review
of Psychology, 55, 623-655.
Koskela, L. (1992). Application of The New Production Philosophy to Construction. Center for
Integrated Facility Engineering. Stanford University.
Koskela, L. (1999). Management of Production in Construction a Theoretical View. Proceedins
for 7th Annual Conference of the International Group for Lean Construction, (pp. 241252). Berkeley, CA.
Koskela, L. (2000). An Exploration Towards a Production Theory and Its Application to
Construction. VTT Technical Research Centre for Finland.
Koskela, L. (2004). Making-Do - The Eight Category of Waste. Proceedins for 12th Annual
Conference of the International Group for Lean Construction. Helsingor, Denmark.
Kozlowski, S. J., & Bell, B. S. (2012). Work Groups and Teams in Organizations. In I. B. Weiner,
W. C. Borman, D. R. Ilgen, R. J. Klimoski, S. Highhouse, & N. W. Schmitt, Handbook of
Psychology: Industrial and Organizational Psychology (Vol. 12, pp. 412-469). Hoboken,
NJ: Wiley.
Lahouti, A. (2013). Cue-Based Decision-Making in Construction Job Site: An Agent-Based
Modeling Approach. Master Thesis - School of Planning, Design, & Construction;
Michigan State Univesity. East Lansing, MI, U. S. A.
Lahouti, A., & Abdelhamid, T. S. (2012). Cue-Based Decision-Making in Construction: An AgentBased Modeling Approach. In I. D. Tommelein, & C. Pasquire (Ed.), Proceedins for 20th
Annual Conference of the International Group for Lean Construction. San Diego, U.S.A.
Retrieved from www.iglc.net
Lewin, K. (1948). Resolving Social Conflicts : Selected Papers on Group Dynamics. NY: Harper
& Brothers Publishers.
Liberda, M., Ruwanpura, J., & Jergeas, G. (2003). Construction Productivity Improvement: A
Study of Human, Management, and External Issues. Winds of Change Intgration and
Innovation in Construction; Proceedings of the Construction Research Congress (pp. 18). Honolulu: American Society of Civil Engineers.
Ligmann-Zielinska, A. (2010). Agent-Based Models. In B. Warf, Encyclopedia of Geography (pp.
28-31).
Thousand
Oaks:
SAGE
Publications,
Inc;.
DOI:
http://dx.doi.org/10.4135/9781412939591.n14.

139

Ligmann-Zielinska, A., & Sun, L. (2010). Applying Time-Dependent Variance-Based Global
Sensitivity Analysis to Represent the Dynamics of An Agent-based Model of Land Use
Change. International Journal of Geographical Information Science, 24(12), 1829-1850.
Mabry, E. A., & Barnes, R. E. (1980). The Dynamics of Small Group Communication. Englewood
Cliffs, NJ: Prentice-Hall, Inc.
Macht, G. A., & Nembhard, D. A. (2015). Measures and Models of Personality and Their Effects
on Communication and Team Performance. International Journal of Industrial
Ergonomics, 49, 78-89.
Marks, M. A., Mathieu, J. E., & Zaccaro, S. J. (2001). A Temporary Based Framework and
Taxonomy of Team Processes. Academy of Management Review, 26(3), 356-376.
McDavis, J. W., & Harari, H. (1968). Social Psychology: Individuals, Groups, Societies. New
York, NY: Harper & Row Publishers, Inc.
McGrath, J. E. (1964). Social Psychology: A Brief Introduction. Holt, Rinehart, and Winston, Inc.
McGrath, J. E. (1984). Groups: Interaction and Performance. Englewood Cliffs, NJ: PrenticeHall, Inc.
McGrath, J. E., Arrow, H., & Berdahl, J. L. (2000). The Study of Groups: Past, Present, and Future.
Personality and Social Psychology Review, 4(1), 95-105.
Memon, A. H., Rahman, I. A., & Abdul Azis, A. A. (2012). Time and Cost Performance in
Construction Projects in Southern and Central Regions of Penisular Malaysia.
International Journal of Advances in Applied Sciences, 1(1), 45-52.
Menches, C. L., & Chen, J. (2013). Using Ecological Momntary Assessment to Understand a
Construction Worker's Daily Disruptions and Decisions. Construction Management and
Economics, 31(2), 180-194.
Menches, C. L., & Chen, J. (2014). A Diary Study of Distruption Experiences of Crew Members
on a Jobsite. Jounal of Management in Engineering, 30(1), 60-68.
Merriam-Webster Collegiate Dictionary. (2003). Crew. In Merriam-Webster Collegiate
Dictionary (Eleventh ed., p. 295). Soringfield, MA: Merriam-Webster, Incorporated.
Merton, R. K. (1968). Social Theory and Social Structure. New York, NY: The Free Press.
Mills, T. M. (1967). The Sociology of Small Groups. Englewood Cliffs, NJ: Prentice-Hall, Inc.
Mohsini, R. A., & Davidson, C. H. (1992). Determinants of Performance in the Traditional
Building Process. Construction Management and Economics, 10, 343-359.
Moore, H. (2013). Exploring Information Generation and Propagation from the Point of
Installation on Construction Jobsites - An SNA-ABM Hybrid Approach. Doctoral
140

Disseration - School of Planning, Design, & Construction; Michigan State Univesity. East
Lansing, MI, U. S. A.
Morris, C. G. (1966). Task Effects on Group Interaction. Journal of Personality and Social
Psychology, 4(5), 545-554.
Murphy, J. T. (1980). Getting the Facts: A Fieldwork Guide for Evaluators and Policy Analysts.
Santa Monica, CA: Goodyear Publishing Company, Inc.
Newcomb, T. M. (1951). Social Psychology Theory. In J. H. Roher, & M. Sherif (Eds.), Social
Psychology at the Crossroads (pp. 31-49). New York, NY: Harper.
O'Dell, J. W. (1968). Group Size and Emotional Interaction. Journal of Personality and Social
Psychology, 8(1 Pt. 1), 75-78.
Oreskes, N., Shrader-Frechette, K., & Belitz, K. (1994). Verification, Validation, and
Confirmation of Numerical Models in the Earth Sciences. Science, 263(5147), 641-646.
Petty, R. E., Cacioppo, J. T., & Harkins, S. G. (1983). Group Size Effects on Congnitive Effort
and Attitude Change. In H. H. Blumberg, A. P. Hare, V. Kent, & M. F. Davies, Small
Groups and Social Interaction - Volume 1 (pp. 165-181). New York, NY: John Wiley &
Sons.
Pikas, E., Sacks, R., & Priven, V. (2012). Go Or No-Go Decisions At The Construction Workface:
Uncertainty, Perceptions of Readiness, Making Ready and Making-Do. In I. D.
Tommelein, & C. Pasquire (Ed.), Proceedins for 20th Annual Conference of the
International Group for Lean Construction. San Diego, CA.
Radosavljević, M., & Horner, M. W. (2002). The Evidence of Complex Variability in Construction
Labor Productivity. Construction Management and Economics, 20, 3-12.
Railsback, S. F., & Grimm, V. (2012). Agent-Based and Individual-Based Modeling : A Practical
Introduction. Princeton, New Jersey: Princeton University Press.
Rilett, L. R. (1998). Identifying Component Variability of End Product Specification Tests.
Journal of Constrution Engineering and Management, 124(2), 133-138.
Ronen, B. (1992). The Complete Kit Concept. The International Journal of Production Reserch,
30(10), 2457-2466.
Rousseau, V., AubĂŠ, C., & Savoie, A. (2006). Teamwork Behaviors - A Review and an Integration
of Frameworks. Small Group Research, 37(5), 540-570.
Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., . . . Tarantola, S.
(2008). Global Sensitivity Analysis; The Primer. England, U.K.: John Wiley & Sons, Ltd.
Sanders, S. R., & Thomas, H. R. (1991). Factors Affecting Masonry-Labor Productivity. Journal
of Construction Engineering and Management, 117(4), 626-644.
141

Sawhney, A., Walsh, K., & Mulky, A. R. (2003). Agent-Based Modeling and Simulation in
Construction. Proceedings of the 2003 Winter Simulation Conference, (pp. 1541-1547).
New Orleans.
Schatzman, L., & Strauss, A. L. (1973). Field Research: Strategies for a Natural Sociology.
Englewood Cliffs, NJ: Prentice-Hall, Inc.
Seidman, I. (1998). Interviewing as Qualitative Research - Second Edition. New York, NY:
Teachers College Press.
Shaw, M. E. (1981). Group Dynamics: The Psychology of Small Group Behavior (Third ed.). U.
S. A.: McGraw-Hill, Inc.
Shohet, I. M., & Frydman, S. (2003). Communication Patterns in Construction at Construction
Manager Level. Journal of Construction Engineering and Management, 129(5), 570-577.
SimLab. (2.2). Retrieved 2017, from https://ec.europa.eu/jrc/en/samo/simlab
Simon, H. A. (1957). Models of Man: Social and Rational; Mathematical Essays on Rational
Human Behavior in Society Setting. New York: Wiley.
Smith, M. (1945). Social Situation, Social Behavior, and Social Group. Psychological Review,
224-229.
Sniezek, J. A., & Henry, R. A. (1989). Accuracy and Confidence in Group Judgment.
Organizational Behavior and Human Decision Processes, 43, 1-28.
Spatz, D. M. (2000). Team-Building in Construction. Practice Periodical on Structural Design
and Construction, 5, 93-105.
Srbljinovic, A., & Skunca, O. (2003). An Introduction to Agent-Based Modeling and Simulation
of Social Processes. Interdisciplinary Description of Complex Systems, 1-8.
Steiner, I. D. (1972). Group Process and Productivity. New York, NY: Academic Press, Inc.
Sterman, J. D. (1992). System Dynamics Modeling for Project Management. Cambridge, MA:
System Dynamics Group.
Stock, R. (2004). Drivers to Team Performance: What Do We Know and What Have Still To
Learn. Schmalenbach Business Review, 56, 274-306.
Syer, J., & Connolly, C. (1996). How Teamwork Works. London, United Kingdom: McGraw Hill.
Thomas, H. R., & Sakarcan, A. S. (1994). Forecasting Labor Productivity Using Factor Model.
Journal of Construction Engineering and Management, 120(1), 228-239.
Thomas, H. R., Horman, M. J., de Souza, U. L., & Zavrski, I. (2002). Reducing Variability to
Improve Performance as a Lean Construction Principle. Journal of Construction
Engineering and Management, 128(2), 144-154.
142

Tommelein, I. D., Riley, D. R., & Howell, G. A. (1999). Parade Game: Impact of Work Flow
Variability on Trade Performance. Journal of Construction Engineering and Management,
304-310.
VandenBos, G. R. (Ed.). (2015). APA Dictionary of Psychology (Second ed.). Washinton D. C.:
American Psychological Association.
Wambeke, B. W., Hsiang, S. M., & Liu, M. (2011). Causes of Variation in Construction Project
Task Starting Times and Duration. Journal of Construction Engineering and Management,
663-677.
Watkins, M., Mukherjee, A., & Onder, N. (2008). Using Situational Simulations to Collect
Analyze Dynamic Construction Management Decision-Making Data. Proceeding of the
2008 Winter Simulation Conference, (pp. 2377-2386). Austin.
Wilensky, U. (1999). NetLogo. http://ccl.northwestern.edu/netlogo/. Center for Connected
Learning and Computer-Based Modeling. Northwestern University. Evanston, IL.
Retrieved February 29, 2016
Womack, J. P., & Jones, D. T. (2003). Lean Thinking. New York, NY: Free Press.
Yin, R. K. (1994). Case Study Research: Design and Methods - Second Edition. Thousands Oaks,
CA: Sage Publications Inc.
Yin, R. K. (2011). Qualitative Research from Start to Finish. New York, NY: The Guilford Press.

143