THE COLLABORATIVE PROCESS OF DECISION-MAKING IN
THE COOLED LEARNING ENVIRONMENT
Patricia Cristiane de Souza, Eunice P. dos Santos Nunes and Guilherme M. Armigliatto
Institute of Computing, Federal University of Mato Grosso - UFMT, Cuiabá, Mato Grosso, Brazil
Keywords: Collaborative Virtual Environment, Decision-making, Intelligent Agent.
Abstract: This paper presents decision-making processes and collaboration and coordination mechanisms with the use
of an intelligent agent in the CoolED Environment. The CoolED is a collaborative learning environment
developed by the LAVI research group at the Federal University of Mato Grosso (UFMT). This system aims
to help computer science undergraduate students master the content of Nonlinear Lists, which is a subject of
the Data Structuring class, emphasizing insertion, removal and search operations. The group decision-
making is made by a leader chosen by the agent and it follows some evaluation criteria stipulated in a
leadership model developed on this work. This model allows any user a chance to lead the group at least one
work section and respects individual interests on the process of becoming a leader.
1 INTRODUCTION
This paper presents decision-making processes and
collaboration and coordination mechanisms with the
use of an intelligent agent in the CoolED –
Collaborative Learning Environment for the Study
of Nonlinear Lists. This environment is one of the
results of researches in collaborative environments
of the Research Group Laboratory of Interactive
Virtual Environments (Laboratório de Ambientes
Virtuais Interativos - LAVI) of the Computer
Science Institute at the Federal University of Mato
Grosso (UFMT).
Decision-making is a very important factor in a
collaborative environment, but working
collaboratively demands additional effort for
coordinating its members. Without coordination,
many communication efforts are not put to good use
during collaboration. That is to say, in order for the
group to satisfactorily act together and make correct
and coherent decisions, compromises must be made
during conversations among participants. In this
context, the agent plays the role of a mediator and
contributes to the decision-making process,
improving communication among participants and
helping develop collaborative learning.
This paper is divided into six sections. Section 2
describes how the decision-making process operates.
Section 3 presents a few studies related to the topic.
Section 4 presents the CoolED system and shows
how it works, describing the decision-making
process and the role of the agent in mediating this
process. Section 5 presents the conclusions of this
study and the next steps for continuing the research.
Lastly, Section 6 presents bibliographic references.
2 DECISION-MAKING PROCESS
The decision-making process characterizes a
situation that displays a certain degree of uncertainty
and a certain level of complexity. When occurring in
a group, this process also considers individual
knowledge based on each person’s values, creeds
and experiences. The process is initially based on
information exchange between individuals within
the group by the resources offered by the
environment. From this interaction and group
reasoning, a so-called collective knowledge is
created, which determines the decision through a
sequence of actions that will be executed. At this
moment, the decision-making process may be
considered pre-finalized because, when discerning
the result, a new discussion is generally initiated,
creating a certain unbalance within the group and
often producing changes in the collective knowledge
about a given situation. This experience, in turn,
leads the group to acquire new and richer
227
Cristiane de Souza P., P. dos Santos Nunes E. and M. Armigliatto G. (2010).
THE COLLABORATIVE PROCESS OF DECISION-MAKING IN THE COOLED LEARNING ENVIRONMENT.
In Proceedings of the 2nd International Conference on Computer Supported Education, pages 227-232
DOI: 10.5220/0002780202270232
Copyright
c
SciTePress
experiences, due to the debate generated by the
encountered problem.
Figure 1 illustrates a schema of the decision-
making process in a group, also called collective
decision-making process based on the work authored
by Rodriguez et al. (2007), which presents a social
software system for the collective decision-making
process called Smartocracy.
Figure 1: Collective decision-making stages.
According to Moron (1998), the decision-making
process is characterized by phases that ensues the
so-called decision in an election, for instance, where
people generally research, discuss and exchange
information. Decision-making can be considered an
attempt to resolve a certain problem.
Making a decision means choosing among many
well-considered alternatives. Thus a decision must
be made as a response to a problem that has many
possible solutions and each solution has positive and
negative sides (Gomes et al., 2006).
Also, when certain decision involves a high
degree of complexity, its resolution simultaneously
influences many different objectives. A decision is
made considering quantitative and qualitative
parameters and their impacts are not easily
predictable; they can result on the possibility or
necessity of other decision-making processes.
It is also worth noting that not all individuals
who participate in a decision-making process
necessarily have the power to decide. However,
these participants can contend with deciders in order
to influence them.
Therefore, the decision theory is neither a
descriptive nor explanatory theory, since it does not
attempt to describe or explain how and/or why
people act a certain way or make certain decisions
(Maciel, 2008). Conversely, Gomes, Gomes and
Almeida (2006) believe the theory to be at times
prescriptive and at other times normative, in the
sense that it intends to help people make better
decisions according to their basic preferences. The
decision theory thus assumes that individuals are
capable of expressing their basic preferences and are
rational when they face simple decision-making
situations.
There are different methods that aid decision-
making and the most commonly consensual is voting
and leader-based decision-making. The latter is the
method used in the CoolED system. This choice is
based on an experience that uses a voting tool
described in Souza’s work (2003) and also on the
fact that the proposed environment works with small
groups and with non complex problems, demanding
some discussion without, however, much time for
choosing, for instance, a consensual method. In
addition, this choice is also based on the fact that
leadership method allows verifying some behavior
aspects on students such as diligence and
responsibility, among others.
Leader-based decision-making relies on the
choice of one member to lead the group and this
person will be invested with the power to make a
certain decision, though this leader’s choice will be
influenced by other members. This method requires
the establishment of criteria for choosing the leader
who will not necessarily remain the whole time with
this function. The length a leadership will last
depends on the methodology employed – it may
have a fixed leader during the whole collaborative
process or other members of the group may also play
this role. A model of leadership was elaborated and
it is described on section 4.3.
After the decision is made, the group must
organize itself to assess the efficiency of the adopted
solution. This evaluation will serve as a parameter
for next decisions. Therefore, we can consider a
decision positive when it is democratically
accomplished and generates harmony, satisfaction
and motivation among participants.
3 RELATED STUDIES
The use of intelligent agents has been widely
explored by researchers of this field in order to
monitor learner behavior and to facilitate the
teaching-learning process in collaborative learning
environments, interchanging collaboration and
interaction (Wang, 2002).
Souza (2003) developed a collaborative learning
environment for problem solving in the field of civil
engineering. The presented prototype uses an agent
that simulates the behavior of a social constructivist
INPUT
pro
blem
individual
solution
ideas
generate
potential
solutions
collective
solution
rankin
g
final
solution
selection
OUTPUT
solution
CSEDU 2010 - 2nd International Conference on Computer Supported Education
228
professor aiming to mediate student interaction. This
agent possesses a plan-driven and information-based
behavior regarding learners that is dynamically built
during the interaction course. The role of the
mediator agent is to stimulate and verify the
existence of a learning situation, thus it monitors
individual and collective participation of the group
within the environment and acts according to
defined mediation strategies. The agent also assists
group members during the decision-making process
through voting tool. Despite the existence of this
tool, the agent uses its communication strategy to
stimulate discussion among members about the
decision being made before using the tool.
Goodman et al. (2005) presents the use of an
intelligent agent to recognize the dialog between
participants and interact with them in a collaborative
learning environment. The agent participates in the
discussion and interacts with participants when
detecting learning problems, for instance, when a
participant dominates the debate but does not
interact with other participants or when participants
do not understand the problem at hand. The agent is
also concerned with group interaction during
collaborative learning and encourages participation,
asking the group questions and stimulating members
to think. When a problem is detected, the agent posts
a question or comment in order to facilitate learning
or improving interaction, serving as assistance or a
clue to guide the group in solving the problem.
Therefore, the investigation’s key concern was
determining when and how the intelligent agent
should intervene in learning.
After the simulations, improvements were
identified in the collaboration process (hence
learning process) and showed that the role of the
instructor, represented by the agent, should not be
lost when the student enrolls in distance learning.
Therefore, studies by Goodman et al. (2005) show
that the dialogue between participants (students and
instructor) provides strong evidence to check
whether or not group dynamics in the teaching-
learning process is efficient.
Accordingly, Aguilar, Antonio & Imbert (2006)
propose an intelligent collaborative virtual
environment (ICVE) that incorporates a Pedagogical
Virtual Agent (PVA) to assist one or more team
members during task execution in situations where
the real learning environment is impossible or
undesirable due to costs or danger, while acting as
an advisor and coordinator in virtual group
meetings.
The Pedagogical Virtual Agent (PVA)
incorporated into ICVE will take as its bases for the
tutorial process, the knowledge of the task, as well
as a mechanism for modeling the group based on the
interaction process. It may communicate with the
apprentices or make them suggestions during the
execution of their activities, if necessary. The PVA
will offer its help to the team, giving preference to
activities that are critical for the task success. The
incorporation of a PVA as a team member (leader)
with capacity to carry out behaviors similar to those
of a human tutor will require integrating with the
IVET a cognitive architecture for the virtual agent.
The aim of the study was to develop a strategy
for the assisted formulation of small groups in
ICVEs called Team Training Strategy (TTS), in
order to assist the formulation process. The proposed
strategy consists four interdependent phases in
which the trained team follows an iterative self-
evaluation process for executing a certain task.
4 THE COOLED
The CoolED system is a collaborative learning
environment developed by the LAVI research group
at the Federal University of Mato Grosso (UFMT).
This system aims to help computer science
undergraduate students master the content of
Nonlinear Lists, which is a subject of the Data
Structuring class, emphasizing insertion, removal
and search operations.
4.1 Functions
The CoolED system includes two types of users –
student and professor – who are identified by login.
The professor’s first task is to make a pre-
registration of students containing only essential
data. At this point the professor can either create
work groups with two or three students or let the
system randomly perform this operation. The
professor’s main activities in CoolED are: including
new exercises in the environment, since CoolED
already has some available exercises; inserting
didactic texts to download; recording appointments
in a planner; creating topics for discussion in the
forum; and visualizing and/or printing reports of
student activity in the environment.
The agent that welcomes the student receives this
student’s first access in CoolED and provides a brief
description of the operation and aims of the
environment. The agent then asks the student to
finalize his/her registration through the option
profile, which is made available in the environment.
THE COLLABORATIVE PROCESS OF DECISION-MAKING IN THE COOLED LEARNING ENVIRONMENT
229
The main menu shows the options concerning the
domain (binary trees, AVL trees, and B trees),
didactic material for download, and an option to
activate the Help system and the following
submenus include concepts, tree operation
simulations and proposed exercises about insertion,
removal and search operations.
CoolED also presents chats to allow student
group discussion; a forum, which acts as a repository
of topics of interest for discussion, where students
and professors may participate; an agenda or
planner, where the professor posts deadlines for
activities and meetings; and an animated agent that
acts as a mediator in the collaborative process,
stimulating participation of all group members,
observing the development of activities, and
assisting the decision-making process as well.
4.2 The Agent
The agent is responsible for the student-environment
communication interface. It acts as mediator,
intervening when the student, for instance,
experiences difficulties in using a CoolED tool or in
conducting exercises and also when choosing the
group leader. The term mediator was constructed
from the concept of mediated relation described by
Vygotsky and is based on the behavioral analysis of
a teacher in a classroom who follows a social
constructivist approach.
In the developed prototype the agent’s figure is
represented by Peedy, an animated agent developed
by Microsoft that is available free of charge. Peedy
possesses a few movement functions, appearance
and text balloons. The remaining functionalities are
being developed by LAVI.
The following functions pertain to the agent:
Warning – shown to the students every time
new information is included by the professor in
the Planner, when material is available for
download, and when interest topics are inserted
in the Forum;
Assistance – provided whenever the agent
notices that a student faces difficulties in solving
the exercise, when there is lack of
communication between group members or when
communication is unilateral;
Correction – at the end of an exercise, the agent
compares the group answer to the exercise
solution;
Decision-making – for a solution of each
exercise, the agent uses the leadership model for
determining which student will lead the group.
4.3 CoolED’s Decision-making Process
In solving exercises collaboratively, group decision-
making is conducted by a leader, who is selected by
the agent, considering evaluation criteria described
in the leadership model. The student chosen as a
leader is the only one who conducts insertion,
removal and search operations in trees, and the role
of the remaining students is to participate in the
solution of the exercise, expressing their opinions
through the chat room.
In the model of leadership each student has a
leadership reputation grade that corresponds to the
probability of being chosen to be leader and a
variable that indicates how many times the student
has already been a leader. The system initially
bestows grade 5 to each student, hence all students
are equally likely to become leaders. As students
lead the exercises, their grades vary from 0 to 10,
where 0 means least likely to be chosen and 10
means most likely to be chosen. Grades start at 5
because, after each leadership, a student can have his
or her grade increased or decreased depending on his
or her performance. Whenever there is someone in
the group who has not yet been leader, this person
will always be chosen. The number of times a
student has been a leader will also be taken into
consideration, and priority is given to those who
were leaders the least often. In case of a tie between
the grade and number of times two people were
leaders, both will have the same chances of being
chosen, and the agent will randomly select one of
them.
At the end of each collaborative student-led
exercise, the group will answer an evaluation survey
that aims to analyze leadership performance, and
thus serve as a positive or negative weight for the
leadership grade. CoolED possesses two types of
surveys, one for group members lacking leadership
functions and another for leader self-evaluation. The
self-evaluation questions induce leaders to reflect on
their performance in this role and their interest in
repeating the experience. The other survey makes
the remaining participants of this work session to
analyze the leader’s role and investigate whether
they wish to work with this leader again. The result
after compilation is presented by the agent to the
leader.
4.4 Implementation
The CooIED system is in a final stage of
development taking as a basis the Document of
Requisites Specification, the Entity-Relation
CSEDU 2010 - 2nd International Conference on Computer Supported Education
230
Figure 2: Interface of the instructor’s module.
Diagram, and the Interface Project. The CoolED
interface project was elaborated according to
methodology proposed by Pressman (2006) to create
interface projects for web. Different criteria of
usability were adopted during the whole project and
a session of inspection was implemented using
heuristic evaluation.
After defining MySQL as management language,
the data basis was created, as well its tables and its
relations. In addition, some tests were taken in order
to verify their integrity. For its implementation Java
was the language selected and this choice was made
for its particularities, for non additional costs for the
project, and for been known for the development
team. After studies, the following frameworks were
defined: JSF (with facelets, richfaces, and Ajax4Java
– A4J), Hibernate JPA, and JavaMail. Proper
implementation started with the process of building
the classes which are related to the data basis entities
and its maps using Hibemate JPA specifications.
After that, insertions of random data were made on
the data basis and a few tests were taken as well.
In accordance to the CoolED Specification
Document, the implementation of the system
followed MVC architecture and the layer model was
subdivided on business and persistence. On the
persistence layer several interfaces (Java) were
defined to be implemented. A number of tests were
taken in order to verify the mapping made by classes
and the XLM file, the Hibermate framework
behavior inside the application, and also the Java
code necessary for access operations on data basis.
The layer of vision is on a process of
development based on images available for the
interface project team, where mapped actions and
attributes have been used on the pages. The
Controller has been developed as an intermediate
stratus among layers of vision, persistence and
business. It is responsible for providing required
information and registering it, applying some
validation rules.
After dividing responsibilities between creations
of data basis, its application and the classes that
support that structure, the construction of the pages
of web application was started attending to the
student´s and the instructor´s modules. The
construction of the pages and the application of the
tests have been finalized and soon the integration
between web application and the agent will be done,
concluding the construction of the first version of
the CoolED System.
To illustrate how the CoolED works, figure 2
presents an interface for the instructor’s module of
insertion of the binary tree activity. On the screen
the instructor determines what type of activity:
insertion, removing or search. In this example, the
teacher has chosen a sort of tree (null or not null),
and typed some keys and the complete statement of
the activity as well. After finishing (save key) the
activity it is storage on the CoolED system data
basis and ready for use at the student module.
THE COLLABORATIVE PROCESS OF DECISION-MAKING IN THE COOLED LEARNING ENVIRONMENT
231
5 CONCLUSIONS AND FUTURES
STUDIES
Investigation of related studies still leads us to
believe in the need for an environment that considers
collaborative learning, where the professor can be
present during the whole work session represented
by the mediator agent and where the professor’s role
is not limited to analyzing the end result of
collaborative activity. Every issue involving
communication, activity coordination, and above all
decision-making processes is related to the
opportunities that the environment bestows, favoring
constant dialogue between group members and
especially interest in teamwork. Leadership is a
crucial issue in this work since it is not fixed and
continuous in all work sessions and is thus able to
permeate all group members. Besides, there is an
attempt to respect individual interest in leading the
group by means of an evaluation conducted
according to previously established criteria in the
leadership model.
As regards the investigative aspects of this study, the
next step is to conduct an empirical study with a data
structuring class of students enrolled in the
undergraduate department of Computer Science and
Information Systems at UFMT, thus analyzing the
agent’s performance as a mediator of student
interaction and especially in the decision-making
context.
In this experiment the researchers intent to observe
the CoolED support during the learning process of
non linear lists (trees), the decision making process
behavior of the students, and also the efficiency of
the developed leadership model. The study is not
restricted to the environment, as we also intend to
use the interviewing method with users in order to
conduct a more accurate analysis.
In this final stage of the development of the system,
the methodology for this empirical study and
interview is on the process of elaboration. It is
expected that in middle of March those activities
will be done and thus the experiment will be
concluded in April of 2010.
It is worth noting that one of the members of the
research group LAVI is a professor who teaches data
structuring classes. She is the responsible for
elaborating the whole contend related to the domain
of the system: texts, examples, simulations, and tests
on trees.
ACKNOWLEDGEMENTS
CNPq and Propeq-UFMT.
REFERENCES
Aguilar, R. A; de Antonio, A.; Imbert, R. An Intelligent
Collaborative Virtual Environment for Team Training
– A Preliminary Report. (November 21-24, 2006). In
Proceedings of the 15th International Conference on
Computing CIC.IEE Computer Society, Washington,
DC, 236-239. doi:10.1109/CIC.2006.23
Gomes, L. F. A. M.; Gomes, C. F. S.; Almeida, A. T. de.
(2006). Tomada de Decisão Gerencial: enfoque
multicritério. (2nd ed) São Paulo: Atlas.
Goodman, B. A.; Linton, F. N. Gaimari, R. D.; Hitzeman,
J. M.; Ross, H, J.; Zarrella, G. (Mar, 2005). Using
Dialogue Features to Predict Trouble During
Collaborative Learning. User Modeling and User-
Adapted Interaction 15, 1-2, 85-134. doi:
10.1007/s11257-004-5269-x.
Maciel, C. Um Método para Mensurar o Grau de
Maturidade na Tomada de Decisão E-Democrática.
(2008). Doctoral Thesis in Computer Science.
Universidade Federal Fluminense. Niterói (RJ), Brazil.
Moron, M. A. M. Concepção, Desenvolvimento e
Validação de Instrumentos de Coleta de Dados para
Estudar a Percepção do Processo Decisório e as
Diferenças Culturais. (1998). Master Thesis in
Administration. Universidade Federal de Rio Grande
do Sul. Porto Alegre (RS), Brazil.
Pressman, R. S. (2006). Software Engineering: a software
practitioner´s approach. (6nd ed). Mcgraw-Hill.
Rodriguez, M. A.; Steinbock, D. J.; Watkins, J. H.;
Gershenson, C.; Bollen, J.; Grey, V.; de Graf, B.
Smartocracy: Social Networks for Collective Decision
Making. (2007). Proceedings of the 40th Hawaii
International Conference on System Sciences.
Souza, P. C. Diretrizes para a Construção de Mediadores
Sócio-Construtivistas em Sistemas de Aprendizagem
Colaborativa Apoiada por Computador. (2003).
Doctoral Thesis in Engineering. Universidade Federal
de Santa Catarina. Florianópolis (SC), Brazil.
Wang, F. Self-organising communities formed by middle
agents. (2002). In Proceedings of the first international
conference on autonomous agents and multi-agent
systems, Bologna,Italy.
CSEDU 2010 - 2nd International Conference on Computer Supported Education
232