Improving Decision-making in Virtual Learning Environments using

a Tracing Tutor Agent

Aluizio Haendchen Filho

1,2

, Jeferson Miguel Thalheimer

, Rudimar L. S. Dazzi

, Vinícius Santos

and Paulo Ivo Koehntopp

Laboratory of Applied Intelligence, University of the Itajaí Valley (UNIVALI), Rua Uruguay, 458, Itajaí, Brazil

Univertity Center of Brusque (UNIFEBE), Brusque, Brazil

Catarinense Association of Educational Foundations (ACAFE), Florianópolis, Brazil

pauloivo@uol.com.br

Keywords: Intelligent Tutor, Software Agents, Decision-making, Virtual Learning Environment.

Abstract: Quality of care in the Virtual Learning Environment is often compromised by large numbers of students. This

presents a difficult task for human tutors. On the other hand, Intelligent Tutoring Systems are evolving

towards a decision support system. One vision of Artificial Intelligence and education is to produce a tutor

for every student or a “community of tutors for every student”. Here we present a model of intelligent tracing

tutor agent responsible for tracking students in the virtual learning environment. We have designed the

Tracing Tutor Agent as one of the agents of a collaborative organization of intelligent tutor agents. Each agent

has his role, responsibilities and permissions. The main focus of this work is to present a model of Tracing

Tutor Agent (TTA), which is one of the organization’s agents. It has the following responsibilities: (i) monitor

students’ actions in VLE; (ii) to monitor the actions of the human tutor in the VLE; and (iii) collaborate and

interact with the other organization’s agents to supply the human tutor with information in order to improve

decision making and performance, increasing attendance and avoiding evasion.

1 INTRODUCTION

The possibility of investing in universities and private

courses have made the educational market very

attractive and profitable for national and international

groups that have been constantly involved in large,

million-dollars negotiations. With the technological

advancements and the spread of distance learning

courses, the number of students entering this

environment has significantly increased every year.

Despite all the technological advances and

concerns to ensure courses quality, Virtual Learning

Environments (VLE) continue to present significant

problems (Tomelin, 2016). The quality of care in the

VLE is often compromised by large numbers of

students. Consequently, problems including

demotivation and sentiments of isolation arise,

causing low use and high evasion rate.

According to Vicari and Giraffa (2003), with the

emergence of Artificial Intelligence (AI), the

developers began to use Computed Assisted

Instruction (CAI) techniques to make the software

more active in the process of student interaction.

From that initial experience emerged the Intelligent

Computer-assisted Instruction (ICAI) method and

following that Intelligent Tutoring Systems (ITSs).

Shifting from a problematic Expert System to an ITS

appeared to be a natural course.

In the search for automation and trying to

reproduce the teacher pedagogy, the tutors were

initially developed with the purpose of replacing the

teacher. With the evolution of virtual learning

environments and the emergence of active learning

methodologies, ITSs are evolving towards a decision

support system. One vision of Artificial Intelligence

and education is to produce a “tutor for every

student”, or a “community of tutors for every

student.” This vision includes the process of learning

in social activity, accepting multimodal input from

students (Woolf, 2010). Thus, the role of intelligent

tutors needs to be adapted to this new reality.

We have designed the Tracing Tutor Agent as a

collaborative agent of an intelligent tutor from the

organization. Each agent has his own role, responsibi-

lities and permissions. Many of the agents' activities

are executed in real time. When their lifecycle starts,

they remain in constant monitoring of the virtual

learning environment. In order to properly fulfill their

600

Filho, A., Thalheimer, J., Dazzi, R., Santos, V. and Koehntopp, P.

Improving Decision-making in Virtual Learning Environments using a Tracing Tutor Agent.

DOI: 10.5220/0007744006000607

In Proceedings of the 21st International Conference on Enterprise Information Systems (ICEIS 2019), pages 600-607

ISBN: 978-989-758-372-8

roles, the collaboration and interaction between them

are two fundamental properties.

The main focus of this work is to present a model of

Tracing Tutor Agent (TTA). It has the following respon-

sibilities: (i) monitor the actions of students in VLE; (ii)

monitor the actions of the human tutor in the VLE; and

(iii) collaborate and interact with the other organiza-

tion’s agents to supply the human tutor with information

in order to improve decision making and performance,

increasing attendance and avoiding evasion.

2 BACKGROUND

To substantiate the approach, this section presents the

following concepts: (i) intelligent tutors; (ii) data

webhouse and (iii) agent development platform.

2.1 Intelligent Tutors

Traditionally, ITSs are educational computational

systems that aim to promote immediate and

customized instruction to provide students with

individualized learning (Vos, 1995, Van Lehn, 2006,

Bernacki et al., 2014, Latham et al. Lin et al., 2014).

These instructions are performed without human

intervention, making the system simulate the

behaviour of a human tutor. In this way, tips are

introduced to help the student to correctly develop the

problem and identify errors (Wenger, 1987; Azevedo

et al., 1999; Curilem (2007), and Adams et al. (2014).

ITSs have been investigated by several authors,

including Raabe (2005), Giraffa (1999), San Pedro,

Baker and Rodrigo (2014), Rissoli, et al. (2006), and

more recently Mayer et al. (2014).

According to Rissoli, Giraffa and Martins (2006)

ITSs have the ability to learn and teach. In this way,

ITSs allow adaptation of teaching strategies to the

learning of each student's needs. The dynamic and

coherent combination of information and the content

of pedagogical aspects are related to every student.

According to Dazzi (2007), ITSs are computer

systems that tutor a student in a given domain. ITSs

models the students' understanding of a topic. As it

performs certain tasks in the system, it compares the

student's knowledge to the model they have of an

expert in that domain. If there is a difference, the

system can use its domain model to generate an

explanation to help the student to identify the mistake.

The system can also adjust student learning levels and

styles and present information, tests and more

appropriate feedback.

Ausubel (1980) proposes the Significant Learning

Theory (SLT) method, which helps to clarify the

construction of new knowledge based on the learner's

cognitive structure. In the educational context, this

theory promotes respect for the characteristics of

students. This demands more from their own education-

nal processes rather than their sociocultural particulari-

ties. It requires a personalized and interactive education,

which allows the advancement of each student consider-

ing their particular needs. This personalized education is

not feasible in traditional teaching and learning, where a

large number of students are under the pedagogical

responsibility of a single tutor (Souza et al., 2013).

Research conducted by Azevedo et al. (2016) and

Mudrick, et al. (2017) support the idea that assistance

and feedback of virtual pedagogical agents’ favour

self-regulated and complex student learning. This

depends on tracking, modelling and incentive to the

intelligent and precise feedback (apud Barcenes et al.,

2018). Other authors like (Grawemeyer et al., 2017)

developed a system to improve commitment and

learning through formative feedback based on the

emotional state of the student, called iTalk2Learn.

In this sense, we observe the great potential of the

systems that offer personalized education and

tracking of students’ actions.

2.2 Data Webhouse

According to Kimball and Merz (2000), the Web

allows the possibility of recording practically all

behavioural actions of the user in a single click. In

terms of behavioural actions, it must be understood

that one can capture not only the page accessed but

also weather and navigability information.

To make their proposal feasible a technique called

clickstream is used for the exploitation of Web access

information. The recording of all interactions made

by anyone via an application or web site, is literally

called a clickstream. Activities carried out by the user

such as click capturing, form filling, and others,

create conditions for analysis, profile identifications,

preferences and trends of each particular user.

Figure 1 shows a very simple example of a

dimensional model for DWH.

Figure 1: Simple dimensional DWH example.

Improving Decision-making in Virtual Learning Environments using a Tracing Tutor Agent

601

As for Data Warehouse, Data Webhouse (DWH)

is built with an architecture called Dimensional

Model. Dimensional modelling is a discipline that

seeks to model data for the purposes of

understandability and performance.

All dimensional models are built around the

concept of measured facts. The Facts table,

represented by the entity Clickstreams, stores users'

clicks on the VLE. The dimensions relate to the

entities that serve as perspectives of analysis in any

subject of the model. In the example, Student,

Discipline, Time and Tutor are the dimensions

connected to the fact table. Dimensions are rich in

descriptions. For example, the Student dimension

stores all the student profile data.

2.3 Agent Development Platform

The inherent difficulties of agent-based systems

development induce the search for tools in order to

minimize complexity and reduce time and cost. For

the development of the solution, we use the Midas

Platform (Haendchen Filho, Prado, and Lucena,

2007). Midas offers a flexible, extensible, adaptive,

loosely coupled, and service-oriented architecture in

order to interoperate in the Web. The platform

provides a complete runtime environment where

agents can execute concurrently within the same host.

Figure 2 shows the Midas generic architecture

detailing the Agent Container (AC). AC is a web

container where application agents and components

are hosted, providing a platform and a framework to

simplify the Multi Agent System development. The

mechanisms embedded in Midas work in two

abstraction levels: (i) the generic architectural level,

where four middleware agents are responsible for

providing infrastructure services; and (ii) in the

agents’ design level, providing abstract classes that

define hot-spots from which concrete agents and/or

components can be developed. The abstract template

already empowers the agents with communication

facilities, including event-based listeners and a

blackboard.

The reference architecture is composed of five

models: (i) Message-Oriented Model (MOM):

dedicated to the architectural issues about the

structure and transport of messages; (ii) Service-

Oriented Model (SOM): focussed in the services and

actions executed by requesters and providers; (iii)

Resource-Oriented Model (ROM): focussed in the

architectural aspects about the resources handling;

and (iv) Management Model (MGM): devoted to the

management of resources.

Blackboard is one of the most used information

exchange techniques in symbolic cognitive Multi-

Agent System. Its structure follows the basic

blackboard pattern: the knowledge sources represent

the agents, the data structure is visible to all agents,

and the controller is responsible for notifying the

agents about the changes in the environment.

Figure 2: Agent Platform (Haendchen Filho et al. 2007).

Agents are autonomous entities that have their

own thread execution and can implement adaptive or

intelligent behaviour. The abstract class provides the

interfaces and triggers the agent execution thread and

the signatures for lifecycle management. The

component is an abstract class that represents purely

reactive entities, typically used to encapsulate the

domain-specific rules of the application, including

business processes, data access objects, and legacy

application functionalities.

3 PROPOSED APPROACH

This section is intended to present the main

specifications of the system. The structure is divided

in (i) Generic Model; (ii) Learning Model; (iii) VLE

Webhouse Model; and (iv) Tracing Tutor Agent Role,

presented below.

3.1 Generic Model

The generic model is composed of an organization of

intelligent tutors, data sources, VLE and two main

actors: the human tutor and the student, as shown in

Figure 3.

Intelligent Tutor System is a software agent

organization composed of the following agents: (i)

TTA (Tracing Tutor Agent), which represents the

tracer tutor, which is the main object of this paper; (ii)

ITA (Interface Tutor Agent), which represents an

ICEIS 2019 - 21st International Conference on Enterprise Information Systems

602

interface tutor; its role is to provide information and

notification adaptive graphical interfaces in VLE; (iii)

PTA (Pedagogical Tutor Agent), that follows students’

interaction with the VLE, collect the information

required for the modelling of students’ profile used to

customise the environment assist and guide students

during the construction of their learning (Dos Santos et

al., 2002); (iv) AITA (Artificial Intelligence Tutor

Agent), which manipulates artificial intelligence

techniques like machine learning and data mining

techniques that perform predictions and prescriptions;

(v) Student representing a virtual student agent, (vi)

Tutor representing a virtual Tutor, and (vii)

Component, representing reactive objects used by

agents to obtain information stored in the database. The

databases are represented in the figure by the DWH

and Academic Information System (AIS).

Figure 3: Generic model.

The student accesses the VLE, having its clicks

captured and stored in a Data Webhouse during its

navigation. After obtaining the data and performing

the analyses, the TTA provides the information to the

human tutor. With this information, the human tutor

can make decisions and send notifications or keep in

touch with students detected on alert. Also, the TTA

may obtain permissions from the human tutor to send

notifications and help students with difficulty.

In VLE, listeners are placed in relevant spot,

waiting for the clicks to trigger the script to store the

information. Locations do not necessarily have to be

on the links since with dynamic page load, data can

be stored with simple interactions. Clickstreams,

about the click data to the user ideally, should be sent

to the webhouse data for storage.

All actions the user take can disclose knowledge

about the use of the system. Obtain such knowledge

the data webhouse is widely used to process analyses,

obtaining information from two main sources: (i)

communication protocol data, stored in the web

services logs; and (ii) behaviour seized with site

scripts after establishing a session. User’s behaviour

on pages is a critical part because it is not so simple

to change a site to capture the information.

There are two approaches to deploy scripts on site

pages: (i) placing a script within each existing clickable

element; and (ii) create a script to listen to data on all

pages and filter the relevant ones. By placing a script

on the clickable elements, it is possible to ensure that

all important information is correctly captured. The

disadvantage of this approach is to generate much work

on site maintenance. On the other hand, using a script

to listen to all page data is a simpler approach, but it

has the disadvantage of losing relevant information

about what is being clicked on.

It is important to clarify that a significant part of

student clicks, especially internal clicks in VLE, can

be collected in the log files generated by some

platforms, such as Moodle (Dougiamas, 2001) and

others. In addition to the clicks performed on the

VLE, the system may also capture clicks from other

browser open windows during the study. To enable

this, the system will prompt users for permission to

use cookies on their computer. With cookies, it is

possible to capture various preferences and interests

of users, and especially the external pages visited

during the learning process, such as collecting articles

and materials related to the topics of the classes.

3.2 Domain Learning Model

VLEs usually have a model called the domain model,

where the contents practices and learning pathways

are defined. In the domain model, the schedules are

defined to allow the contents’ presentation to the

students in the diverse pedagogical strategies of the

system. Figure 4 adapted from (Positivo, 2017)

presents a learning pathway model. The example

shows the activities of a weekly learning schedule.

Figure 4: Model of learning pathway.

In the learning process the student navigates in the

VLE looking to follow the pathways defined in the

domain model. During navigation, the student

performs several actions, such as the reading of

teaching material, participate in challenge cases,

attends video conference explanation, performs the

practical activities requested by the teacher, and

participates in virtual meetings like forums, chats or

others. One of the roles of TTA is to monitor students'

actions in the learning pathway, as will be shown in

the role model.

Improving Decision-making in Virtual Learning Environments using a Tracing Tutor Agent

603

3.3 VLE Webhouse Model

As presented in Section 2, in a Data Webhouse

(DWH), data is stored in dimensional models.

Dimensional modelling can begin by dividing it into

two parts: measurement (facts) and context

(dimensions). Measurements are usually numeric and

repeated. Facts are usually surrounded by a large

textual context (dimensions). Figure 5 shows the

DWH dimensional model.

Figure 5: VLE Webhouse Model.

The template contains a central Fact table

connected with the dimensions relevant to the context

of the VLE. The relationships among Facts table and

dimensions are made by their primary keys.

Users clickstreams and/or collected logs are

stored in the Fact Table, represented by the

click_count attribute. This attribute is an array type,

and contains the data collected in the clicks or log

files, such as the page id, the start time (second /

minute / hour / month / day / year) of the event, event

name, description of what happened in the event, the

previous source or page, the user's IP address, and so

on.

The summary descriptions of dimensions are

given below:

● Calendar date: attributes may include days of

the week, seasons, holiday, workday, tax

period, among others.

● Time_of_day: is a complement to the calendar

date, where are recorded the time slots during

the day, including hours, minutes, and time

slots like lunchtime, class time, and so forth.

● Academic date: associated with different

structures that differ on the number of modules

(semesters, four-month periods and

trimesters). These structures have a different

hierarchy, each with five levels: Year, Module

(6, 4 or 3 months), Period (classes or

examination period), Week (variable length,

defined internally by each institution) and

Day.

● Page: among the attributes of this dimension

can be cited: the page source (e.g., static,

dynamic), function (content, exercise, video,

forum) and so forth.

● Session: session is the collection of actions

taken by a visitor to a site while it navigates

through it without leaving this site.

● Causal: describes the conditions of the current

progress of the subject, such as beginning of

the subject, period of tests, etc.

● Student: information about the student profile.

● Academic records: provides information on

the student’s trajectory, like courses taken,

historical grading, and so forth.

● Discipline: the attributes can include class

hours, credits, opening and closing dates, and

so forth.

● Teacher: information about teacher's profile.

● Tutor: information about the tutor including

the degree of training, number of tutored

disciplines, etc.

● Domain_model: describes the course schedule

of the learning path in the VLE.

● Referrer: brings information about the URL

from where the user came from.

In the dimensional model, the student's historical

dimension is connected to the student dimension,

characterizing a variant of the dimensional model

named snowflakes. The main source of information

for obtaining dimension data is the institution's

Academic Information System.

3.4 Tracing Tutor Role Model

The role model has been used by many approaches

(Gonçalves, 2009, Haendchen Filho, 2017) to provide

a summary of software agents. According to the

theory, a role can be described by two basic attributes:

(i) responsibilities: they are role obligations and

indicate functionality, and (ii) permissions: they are

the rights associated with the role and indicate the

resources that the agents can use. Intuitively,

responsibilities are associated with services that an

agent must provide, while permissions are associated

with the services the agent must fulfil their

responsibilities.

TTA does not communicate directly with the

human tutor or student. Communication tasks with

human actors are carried out by ITA. TTA's main

ICEIS 2019 - 21st International Conference on Enterprise Information Systems

604

responsibilities are to collect data on learning

pathways and collaborate with other intelligent tutors

(ITA, AITA and PTA).

It provides information for ITA (detailed in the

TTA Role Model) that makes it available for VLE

managers, human tutors, and students. The data

collected in the clicks and/or log files are stored and

made available to AITA, which can use this data to

feed the machine learning and data mining

algorithms. TTA also collaborates with the PTA,

which can suggest new pathways and check which

trails the students most successfully use. As

mentioned, all information for decision making

comes to human administrators and tutors via ITA.

TTA's responsibilities are divided into four

groups: (i) student monitoring; (ii) tutor monitoring;

(iii) Provides Information for ITA and (iv) Statistical

Reports for ITA. Figure 6 shows the role model of the

agent, with its Responsibilities, Permissions and

Collaborations.

Figure 6: TTA Role Model.

The services of the Student Monitoring group

comprise tasks related to information about the

participation and interactions of the students in the

VLE, according to the domain learning model:

● Time spent by the student on VLE: identifies

students who are not entering the VLE, those

who are entering infrequently, and those

who are frequent attendees.

● Chapter download: verifies if the student

has downloaded the materials made

available for reading.

● Time spent on cases challenge: monitors the

amount of time the student spent on the

challenges assigned by the teacher.

● Student time on explanation of the scene:

monitors whether the student has attended

and participated in the video conferences.

● Student time in quiz/tests/exercises: they are

synthesized as Stop for Practice in the

learning pathway, verifies if the student is

participating in these activities and how

much time he spends in each one.

● Time spent on virtual meetings: verifies if

the student participates in the forums, chats

and online activities carried out in the VLE.

The related services in the Tutor Monitoring

group comprise activities associated with the

activities performed by the tutor in the VLE:

● Average response time for doubts: the

amount of time that the human tutor or

teacher takes on average to answer students'

questions.

● Time spent by the human tutor VLE: the

amount of time the human tutor remains in

the VLE.

● The average number of feedbacks: counts

the average number of feedbacks given to

students.

The services listed in the Provides Information for

ITA group comprise a sequence of alerts sent from

TTA to ITA:

● Students with delayed tasks: notify the

human tutor about students with overdue

assignments.

● Students with no attendance at VLE: notify

the tutor about students not entering the

VLE.

● Students with little frequency in VLE: notify

the tutor about students entering low on

VLE.

The Statistical Reports group represents reports

that can be available for ITA with information

obtained from the monitoring tasks. For each

monitoring service, reports and graphs can be

generated to facilitate visualization for analysis and

decision-making. With this information, the ITA and

the human tutor can give individual attention to the

students, taking preventive measures to avoid low

achievement or even avoidance.

Role Model Permissions, as well as its

Collaborations are placed in the right-side column of

Figure 6. As mentioned, the permissions include data

and information that the tutor agent can use to fulfil

their responsibilities. TTA uses services provided

from ITS components in order to obtain data from the

database, as well as to generate reports and organize

information.

Based on information provided by TTA on

interactions, students' course in learning pathways,

Improving Decision-making in Virtual Learning Environments using a Tracing Tutor Agent

605

and assessment, PTA can identify points of difficulty

and suggest course corrections in learning paths. This

knowledge could hardly be obtained by human tutors

in virtual environments with large numbers of

students. Moreover, intelligent tutors can be assigned

tasks to assistance students in the navigation process,

and the best course of learning paths through

recommendation systems, optimizing their

accomplishment and achievement.

Furthermore, tracking clickstreams will enable

the tutor to understand why certain groups to follow

a sequence of steps, while others follow different

ones. This will allow verification of the track defined

in the domain model. The expected quality will

hopefully permit continuous improvements in

learning pathways and GUIs.

The monitoring of the actions of the human tutor

in the VLE allows analysis and self-assessment of

their participation in the learning process.

4 DISCUSSION

The approach presented in this paper meets the

operational needs requested by the distance learning

managers of the University Center of Brusque

(UNIFEBE). Specialists in this area point out the

importance of information obtained by student

tracking. In this proposal, this information is made

available by the Tracing Tutor Agent. It provides

information for other collaborative agents in order to

improve the decision-making of the human tutors and

managers in the virtual learning environment.

As before mentioned, Intelligent Tutoring

Systems are characterized for incorporating Artificial

Intelligence techniques into their design and

development, acting as assistants in the teaching-

learning process (De Souza et al 2002). Currently,

Intelligent Agents concepts have been applied to

these systems as a way to improve them.

We chose the Multi-Agent Systems platform to

implement our proposal, considering that intelligent

software agents act dynamically and not only

reactively. They can act in a collaborative way to play

their roles more easily. In addition, intelligent agents

may learn from the knowledge engendered in the

environment, using this learning proactively for the

benefit of managers, human tutors and students.

We understand that in an Intelligent Tutors

organization each agent will have a specific role

model, acting collaboratively. For example, the

Interface Tutor Agent can use the information

provided by Tracing Tutor Agent to send notifications

and create dynamic graphical interfaces, inducing the

student to fill gaps in their learning path. PTA collect

the information required for the modelling of

students’ profile used to customize the environment

assist and guide students during the construction of

their learning. The Artificial Intelligence Tutor Agent

(AITA) can identify student clusters that have

succeeded in learning. It also provides subsidies for

the human tutor and to the managers virtual learning

environments, supporting the task of improving the

domain model. The AITA may also use machine-

learning techniques to identify potential student

evasion and prescribe actions to avoid it.

5 CONCLUSION AND FUTURE

WORKS

The main contribution of this work is to define a

model for intelligent tutors best adapted to the current

management needs of virtual learning environments.

The focus of this work was to define the role of

Tracing Tutor Agent, which is one of the

collaborative agents of the organization. Besides we

have introduced a Data Webhouse model in the

context of intelligent tutor models. We did not find

similar works in the literature, which makes the

approach innovative.

Future work will involve the implementation of

the Tracing Tutor Agent and the development of the

specifications of other intelligent tutors presented in

the model.

REFERENCES

Adams D. M., McLaren B. M., Durkin, K., Mayer, R. E.,

Rittle-Johnson, B., Isotani, S., Velsen, M. V., 2014.

Using erroneous examples to improve mathematics

learning with a web-based tutoring system Computers

in Human Behavior.

Ausubel, D. P., Novak, J. e Hanesian, H., 1980. Psicologia

educacional. Rio de Janeiro: Interamericana.

Azevedo, B. F. T., Tavares, O. L., Cury, D., 1999. Mutantis

– Um Sistema Tutor Inteligente Multi-Agente para o

Ensino-Aprendizagem de Conceitos. In: X Simpósio

Brasileiro de Informática na Educação, Curitiba. Anais

do X Simpósio Brasileiro de Informática na Educação.

Curitiba: Sociedade Brasileira de Computação.

Azevedo, R., Martin, S. A., Taub, M., Mudrick, N. V.,

Millar, G. C., & Grafsgaard, J. F., 2016. Are

Pedagogical Agents’ External Regulation Effective in

Fostering Learning with Intelligent Tutoring Systems?

In A. Micarelli, J. Stamper, & K. Panourgia (Eds.),

Intelligent Tutoring Systems: 13th International

Conference. Cham: Springer International.

ICEIS 2019 - 21st International Conference on Enterprise Information Systems

606

Barcenes, A. B., Sanz, C., Baldassari, S., Ramos, E. R.,

Mendoza, G. V., Merino, R. B., Fierro, W, 2018.

Efective computing: emotions, technologies and their

relationship with virtual education. Revista de

Investigación Talentos V.1.

Bernacki, M. L., Aleven, V., & Nokes-Malach, T. J., 2014.

Stability and change in adolescents’ task-specific

achievement goals and implications for learning

mathematics with intelligent tutors. Computers in

Human behavior.

Curilem, S. G., Barbosa, A. R., de Azevedo, F. M., 2007.

Intelligent tutoring systems: Formalization as automata

and interface design using neural networks, Computers

& Education.

Dazzi, R. L. S., 2007. Metodologia para Adaptação de

Interface e Estratégia Pedagógica em Sistemas Tutores

Inteligentes. Tese (Doutorado em Engenharia

Biomédica) – Instituto de Engenharia Elétrica,

Universidade Federal de Santa Catarina, Florianópolis.

Dos Santos, C. T., Frozza, R., Dhamer, A., Gaspary, L. P.,

2002. DÓRIS — Pedagogical Agent in Intelligent

Tutoring Systems. In: Cerri S.A., Gouardères G.,

Paraguaçu F. (eds) Intelligent Tutoring Systems. ITS

2002. Lecture Notes in Computer Science, vol 2363.

Springer, Berlin, Heidelberg.

Dougiamas, M., Taylor, P. C., 2002. Interpretive analysis

of an internet- based course constructed using a new

courseware tool called Moodle. 2002. Available:

<http://dougiamas.com/writing/herdsa2002/>

Giraffa, L. M. M., 1999. Uma arquitetura de tutor

utilizando estados mentais. UFRGS.

Gonçalves, E. J. T., 2009. Modelagem de arquiteturas

internas de agentes de software utilizando a linguagem,

MAS-ML. Dissertação (Mestrado Acadêmico em

Ciência da Computação) – Universidade Estadual do

Ceará. Centro de Ciência e Tecnologia.

Grawemeyer, B., Mavrikis, M., Holmes, W., Gutiérrez-

Santos, S., Wiedmann, M., & Rummel, N., 2017.

Affective learning: improving engagement and

enhancing learning with affect-aware feedback. User

Modeling and User-Adapted Interaction.

Haendchen Filho, A., 2017. Sistemas Multiagentes na

Internet: Uma arquitetura orientada a serviços para o

desenvolvimento de agentes inteligentes de software na

Internet. Novas Edições Acadêmicas.

Kimball R., Merz R., 2000. The Data Webhouse

Haendchen, F. A., Prado, H. A. and Lucena, C. J. P., 2007.

A WSA-based architecture for building multiagent

systems. In Proceedings of the 6th international joint

conference on Autonomous agents and multiagent

systems (AAMAS '07). ACM, New York, NY, USA.

Kimball, R., Merz, R., 2000. The Data Webhouse Toolkit:

Building the Web-Enabled Data Warehouse. John

Wiley & Sons, Inc.

Latham, A., Crockett, K., & McLean, D., 2014. An

adaptation algorithm for an intelligent natural

language tutoring system. Computers & Education.

Lin, H. C. K., Wu, C. H., Hsueh, Y. P., 2014. The influence

of using affective tutoring system in accounting

remedial instruction on learning performance and

usability. Computers in Human Behavior.

Mayer, R. E., 2014. The Cambridge Handbook of

Multimedia Learning: Second Edition. 2 ed. Cambridge

University Press.

Positivo, U., 2017. Manual do Aluno: Graduação EAD

Raabe, A. L. A., 2005. Uma proposta de arquitetura de

sistema tutor inteligente baseada na teoria das

experiências de aprendizagem mediadas. Universidade

Federal do Rio Grande do Sul, Programa de Pós-

Graduação em Informática na Educação.

Rissoli, V. R. V., Giraffa, L. M. M., Martins, J. P., 2006.

Sistema Tutor Inteligente baseado na Teoria da

Aprendizagem Significativa com acompanhamento

Fuzzy. Informática na educação: teoria & prática, v. 9,

n. 2.

San Pedro, M. O. Z., Baker, R. S. J. D., Rodrigo, M. M. T.,

2014. Carelessness and Affect in an Intelligent Tutoring

System for Mathematics. International Journal of

Artificial Intelligence in Education, v. 24, n. 2.

Souza, C. V., Shiguti, W. A., Rissoli, V. R. V., 2013.

Metodologia Ativa para Aprendizagem Significativa

com Apoio de Tecnologias Inteligentes. Conferência

Internacional sobre Informática na Educação.

Tomelin, J. F., 2016. Perfil das Instituições que ofertam

EAD no Brasil. ABED.

VanLehn, K., 2006. The Behavior of Tutoring Systems.

International Journal of Artificial Intelligence in

Education.

Vicari, R. M., Giraffa, L. M. M., 2003. Fundamentos dos

sistemas tutores inteligentes. In: BARONE, D. (org.).

Sociedades Artificiais: a nova fronteira da inteligência

das máquinas. Porto Alegre: Bookman.

Vos, H. J., 1995. Applications of Bayesian decision theory

to intelligent tutoring systems. Computers in Human

Behavior.

Wenger, E., 1987. Artificial Intelligence and Tutoring

Systems. Morgan Kaufman.

Woolf, B. P., 2010. Building intelligent interactive tutors:

Student-centered strategies for revolutionizing e-

learning. Morgan Kaufmann.

Improving Decision-making in Virtual Learning Environments using a Tracing Tutor Agent

607