The Application of Learning Theories into Abdullah
An Intelligent Arabic Conversational Agent Tutor
Omar G. Alobaidi, Keeley A. Crockett Smieee, Jim D. O'Shea Mieee and Tariq M. Jarad
School of Computing, Mathematics, and Digital Technology, Manchester Metropolitan University,
Manchester, M1 5GD, U.K.
Keywords: Conversational Agents, Intelligent Tutoring System, Knowledge Base, Learning Theory, Pattern Matching.
Abstract: This paper outlines the research and development of a Conversational Intelligent tutoring System (CITS)
named Abdullah focusing on the novel application of learning theories. Abdullah CITS is a software
program intended to converse with students aged 10 to 12 years old about the essential topics in Islam in
natural language. The CITS aims to mimic human Arabic tutor by engaging the students in dialogue using
Modern Arabic language (MAL), and classical Arabic language (CAL), utilizing supportive evidence from
the Quran and Hadith. Abdullah CITS is able to capture the user’s level of knowledge and adapt the tutoring
session and tutoring style to suit that particular learner’s level of knowledge. This is achieved through the
inclusion of several learning theories implemented in Abdullah’s architecture, which are applied to make the
tutoring suited to an individual learner. There are no known specific learning theories for CITS therefore the
novelty of the approach is in the combination of well-known learning theories typically employed in a
classroom environment. The system was evaluated through end user testing with the target age group in
schools in Jordan and the UK. The initial evaluation has produced some positive results, indicating that
Abdullah is gauging the individual learner’s knowledge level and adapting the tutoring session to ensure
learning gain is achieved.
1 INTRODUCTION
Arabic language is the tool which carried the Arabic
culture since the old period until the approach of
Islam, when Arabic became the most important
language in the Islamic world. Arabic language is an
official language of more than twenty countries, and a
major spoken language by over 300 million people
worldwide (Habash, 2012). There are two forms of
Arabic language, which are the modern standard
Arabic (MSA), and the classical Arabic language
(CAL). MSA is used in everyday language, in the
media, education, and literature (Ryding, 2005). MSA
is mainly derived from CAL, which is the standard
form of the language used in the holy Quran.
Intelligent tutoring systems (ITS) are computer based
learning systems, which can adapt to learners’ current
knowledge and skills, provide the necessary feedback
when mistakes are made and provide consistent
tutoring any time, 24 hours a day. ITS are adaptive
educational systems that employ intelligent
technologies to provide individualized instruction, by
adapting to learners’ skill level closely to
individualized lesson provided by the system
(Ghadirli and Rastgarpour, 2013). The main goal of
an ITS, is deliver knowledge by mimicking a human
tutor through a computer-based system. Developing a
CITS for the Arabic language faces many challenges
due to complexity of the morphological system, non-
standardization of the written text, ambiguity, and
lack of resources. However the main challenge for the
developed Arabic CITS is how the user utterances are
recognized and responded to by the Conversational
Agent (CA), as well as how the domain is scripted
and maintained (Alobaidi et al., 2013). This paper
focuses on a novel methodology with regards to
implementing learning theories within Arabic CITS to
adapt the tutoring session to suit the individual
learner’s level of knowledge related to the tutoring
subject. This makes the tutoring session less rigid,
more adaptive, and capable of delivering personalized
learning to suit individual users’ knowledge levels.
This ensures that a more holistic approach to the
tutoring session is implemented making the Arabic
CITS more adaptive and engaging throughout the
tutoring session based on the learners interaction.
Abdullah the Tutor is a web based CITS with a CA
interface which leads the tutoring session, asking
361
Alobaidi O., Crockett K., O’Shea J. and Jarad T..
The Application of Learning Theories into Abdullah: An Intelligent Arabic Conversational Agent Tutor.
DOI: 10.5220/0005197003610369
In Proceedings of the International Conference on Agents and Artificial Intelligence (ICAART-2015), pages 361-369
ISBN: 978-989-758-074-1
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
structured questions, moving from topic to topic in a
session and offering intelligent feedback to students.
This paper is organized as follow: Section 2 will
describe the Arabic CA, Section 3 describes
intelligent tutoring system, Section 4 describes the
learning theories and methods of learning, Section 5
introduces the Abdullah CITS, Section 6 describes the
experimental methodologies, Section 7 will discuss
the results, and section 8 will conclude and highlight
areas for further research and development.
2 ARABIC CONVERSATIONAL
AGENT (ARABIC CA)
A Conversational Agent (CA) is a software program
intended to converse with a human in natural
language (Crockett et al., 2011). CAs have been used
effectively in many applications, such as database
interfaces (Owda et al., 2011), and student’s debt
management guidance (O'Shea et al., 2010). Two
forms of Conversational Agents have been developed,
that being, ‘Embodied CAs’ and ‘Linguistic CAs’.
Linguistic CA’s handle conversation in written or
spoken forms (Yin et al., 2010). Existing CA’s can be
categorized according to the development process
into three main approaches: These approaches are
Natural Language Processing (NLP) (McNamara et
al., 2013); short text semantic sentence similarity
measures (STSM) and Pattern Matching (PM)
(O’Shea et al., 2010). NLP based CAs focus on
translating user utterances and then determine the best
actions to respond to user. Arabic language
theoretically has a number of limitations, it consider a
language of complexity and ambiguity (i.e. Arabic
word might have more than one meaning or the
sentence might have more than one structure).
Consequently, for the above mentioned reasons, the
NLP approach is not suitable to build a CA based on
the Arabic language (Monem et al., 2008). The
second approach is STSM measures. STSS can be
used to measure the semantic similarity between short
texts of sentence length (10 -25 words long) (O’Shea,
2012). In order to build a CA based on STSS a
number of resources are required, such as an
appropriate Arabic Wordnet (AWN) (Boudabous et
al., 2013). AWN is only available for modern Arabic
language. However the lack of resources as well as
the ambiguity of the Arabic language (such as
Morphological and Syntactic ambiguity), led the
researchers to adopt the traditional approach for
building a CA using pattern matching techniques.
Pattern matching is considered as being a good
solution for text-based CA’s as they do not require
grammatically correct or complete input (Hijjawi,
2011). Text-based CA’s use a form of pattern
matching in order to organize their scripts into
contexts consisting of a number of rules which
themselves consist of a number of patterns and a
stimulus response pairs in the CA’s knowledge base.
A rule is the subtopic that belongs to a context that a
user utterance may be matched with a given rule in a
given topic related to the context of the discussion. A
rule can have a number of different patterns that
might be matched with a user’s utterance. Patterns
consist of a collection of words and a wildcards,
which are used to match a portion of the user’s
utterance (Alobaidi et al., 2013).
3 INTELLIGENT TUTORING
SYSTEM (ITS)
Intelligent Tutoring System (ITS) are computer based
learning systems, which assists learners in their
learning process. The main goal of the ITS is to
provide the benefits of one to one instruction
automatically and cost effectively (Sottilare and
Proctor, 2012). ITS typically have four main
components, which are the domain model, learner
model, tutor model, and interface model. The domain
model contains all the elements required to represent
the knowledge to the learners, such as the strategies
or theories, and identify errors (Huertas and Juárez-
Ramírez, 2013). The learner model can track the
learners understanding for the learner, it can make
the right decisions to adapt the tutoring session
(Abdelsalam, 2014). The tutor model is the model
that is concerned about the instructional methods,
such as choosing an appropriate teaching methods
that suit each individual learner (Sani and Aris,
2014). In an intelligent graphical user interface is
responsible for communication with learner and the
CITS (Ghadirli and Rastgarpour, 2013). CA
interfaces to ITS can add more naturalization to the
tutoring, allowing students to experience cooperative
problem solving similar to with a human tutors.
Using a CA interface to an ITS has shown some
success in learning, for example:
• AutoTutor is a CITS that assists the student in
actively constructing knowledge, about computer
literacy through discussion (Cheng et al., 2013).
The main goal of the AutoTutor CITS is
encourage students to show lengthier answers to
questions that exhibit deep reasoning such as
(answers to why, an how questions), while
ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence
362
directing the students towards constructing a
solution (Graesser and McNamara, 2010).
However, the Arabic conversational interface to ITS
is a new area of research. To our knowledge, no
academic research exists on the conversational
intelligent system based on the Arabic language.
4 LEARNING THEORIES AND
LEARNING METHODS
Learning theories focus primary on how the
information is achieved, organized, and recalled
(Groff, 2013). According to the cognitive
information process theory there are a number of
techniques been used to simplify the learning such
as, Gagne’s learning theory (Gagné and Gagné,
1985), and Piaget’s learning theory. This section will
describe each of these learning theories, and method
in more detail.
4.1 Gagne Learning Theory
Gagne’s theory focuses on intentional or purposeful
learning, which is the type of learning that occurs in
school (Gagne et al., 2005). This type of learning
follows a sequence of steps starting from gaining the
attention of the learner to recall of prior learning, to
connecting to previous knowledge, and finally to
transfer of knowledge to the long-term memory
(Gagne et al., 2005). According to Gagne’s theory
there are nine instructional events, which should be
fulfilled to provide the necessary conditions for
learning (Kruse, 2010). These events are (Gain
attention, identify objective, recall prior learning,
present stimulus, guide learning, elicit performance,
provide feedback, assess performance, and enhance
retention).
4.2 Piaget’s Learning Theory
Piaget’s theory of learning is considered one of the
most accurate theories which defines a child
cognitive state (Piaget and Mussen, 1970). Piaget’s
theory is considered more important when teaching
the younger age groups (2-11 years old), as it helps
to determine how much and in what way the learner
will understand the topic being taught (Kim et al.,
2014). Piaget’s theory is used in ITS as a support tool
on many domains (Stipek, 2013). It has been used as
a helping guide for the learner by giving the learner
the information they request, based on their
knowledge in the taught domain (Carmona and
Bueno, 2007).
4.3 Storytelling Learning Method
Storytelling is one of the most powerful and simplest
methods for learning. The use of stories in education
has been found to be most useful in language
learning, such as religious subjects (van Gils, 2005).
Interactive digital storytelling through multimedia is
a valid educational tool to teach literacy and narrative
skills and has been shown to excite people about
learning (Yang and Wu, 2012). Stories must also be
learner designed, in that they need to be tailored for
the specific audience they are delivered to (Mokhtar
et al., 2011). In most religious texts, such as the
Quran, storytelling is the natural way in which
information about fundamental beliefs is taught
(Moll, 2010).
5 ABDULLAH CITS
This section provides a brief overview of the
Abdullah CITS (Alobaidi et al., 2013). Abdullah is a
novel conversational intelligent tutoring system,
which can ask questions and offer problem-solving
support rather than simply presenting the answers.
Abdullah was designed to model a human tutor by
directing a tutoring conversation.
5.1 Abdullah CITS Architecture
The proposed framework for Abdullah CITS consists
of three main components as shown in Figure 1.
These are: the ITS (to personalize teaching according
to individual learner’s characteristics such as the
knowledge of the subject, and the behavior), the
knowledge base (to provide the sources/material of
the learning topics), and the CA (to lead the tutorial
through natural language dialog).
Figure 1: Abdullah CITS Architecture.
TheApplicationofLearningTheoriesintoAbdullah:AnIntelligentArabicConversationalAgentTutor
363
5.2 Implementing Learning Theories to
Abdullah CITS
The Abdullah CITS incorporates a number of
learning theories and methods such as, Gagne theory
of learning, Piaget theory of learning and storytelling
learning method to deliver the tutoring session. The
main aim behind the implementation of these
learning theories is to make Abdullah a more like a
human tutor. The theories allow Abdullah to adapt
and adjust the tutoring session based on the learner’s
interaction with the system. Abdullah is able to
capture several variables to gauge the user/learners
level of understanding and perception in relation to
the tutoring subject. The variables are used to adapt
the session to most suit the learner and apply the
different learning theories, and ensure some level of
learning gain. The next section will outline the
implementation of these theories into Abdullah CITS
and how it utilizes them throughout the tutoring
session.
5.2.1 Application of Gagnes Theory into
Abdullah CIS
Gagne’s theories outline a number of instructional
events, which are briefly described along with a
description of how they are applied in Abdullah
CITS.
Gaining Learners Attention (reception)
Capturing learning attention is considered the first
and the most important process for learning. Two
techniques have been used to deal with this event in
the design of Abdullah CITS:
• A graphical user interface (GUI) that begins
with an animated title screen accompanied by
sound effects, to increase children's visual
orientation (Marco et al., 2009).
• Each lesson will start with a thought-
provoking question or interesting fact about
the selected topic to be taught, curiosity
motivates students to learn (Li, 2013).
Informing Learners of the Objective
To help the learners to complete the lesson and to
achieve the goal of the presented topic, the learning
objective must be listed early in each tutoring
session. In Abdullah, CITS an initial image is
displayed at the beginning of the tutorial describing
that on completion of the lesson, the leaner will have:
• A brief understanding of the selected topic.
• A link to all the supportive evidence for the
topic (Quran and Hadith).
Stimulating Recall of Prior Learning (retrieval)
Associating new information with prior knowledge
can facilitate the learning process (Gagne et al.,
2005). A simple way to stimulate recall is to ask
questions about how well the learners understand
previous concepts or the body of contents in general.
However, all the tutorial questions are organized in
the Tutorial Knowledge Base as questions with
answers in a default style (A normal basic question
which designed to suit different level of learner’s
knowledge), or as questions with answers in basic
detailed style (A type of questions for the learner’s
with low level of knowledge). During the tutoring
session, Abdullah CITS will measure the
understanding of the tutoring topic by the learner by
counting the number of correct default and detailed
answers. The learner’s knowledge will be measured
during the tutorial by a variables, an example of such
variables are:
• The percentage of the correct answers.
• Whether or not the learners ask, a question
related to the main topic.
• Have the learners provided any information
using CAL (i.e. Quran or Hadith). This would
indicate a high level of understanding.
Presenting the Tutorial Content
The tutorial content is designed to include all the
necessary information the learner requires in order to
achieve the learning outcome. The learning contents
of Abdullah CITS were organized based on the book
of monotheism, which is used in primary school
education for learners in years 3, 4 and 5. This book
has been printed and organized by the Ministry of
Education in Saudi Arabia (Al-Sadan, 2000). The
topics were then structured using knowledge
engineering, and involving a real expert teacher.
Figure 2: Abdullah CITS.
ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence
364
Subsequently Gagne learning theories were applied
to give the tutoring a better structure to ensure
effective learning was applied through the Abdullah
CITS. In accordance with Gagne’s and Piaget’s
learning theories, the system utilizes a variety of
media to appeal to different learning styles; media
such as text, graphics and audio as show in
Figure 2.
Providing Feedback (reinforcement)
As learners practice new topics it is very important to
provide specific and immediate feedback on the
learner’s performance. The Abdullah CITS will
display an appropriate encouragement after each
correct answer, partially correct and low near miss
answers. As an example, a match of 80% between
utterance and pattern script is classified as a correct
answer and Abdullah CITS will respond with a
message like “excellent God bless you” (Alobaidi et
al., 2013). It also can provide an appropriate response
in case of learner’s bad behavior, or wrong attitude
about the contents of the tutoring lesson.
5.2.2 Implementation of Piaget’s Theory in
Abdullah CITS
Piaget diagnoses the cognitive processes of the
learners through a number of highly interactive tasks
aimed at learners aged 8-12 years old. Piaget theory
is implemented in Abdullah, through the
determination of the learner’s level of perception and
understanding related to the domain. Abdullah CITS
implements some interactive tasks such as learner’s
promotion, confusion detection, and hint selection
(Anglo and Rodrigo, 2010). Each of these tasks will
now be defined along with an explanation of how
they are applied in Abdullah CITS.
Learner’s Promotion
Learners with a high level of cognitive development
require fewer problems to solve than a learners with
low level of cognitive development (Roll et al.,
2011). For that purpose, Abdullah CITS is designed
with a number of questions allocated for each sub
topics covered during the tutorial. The learners will
only be allowed to move from one sub topic to
another when most questions related to sub topic
been answered correctly (more than 80%).
Confusion Detection
Learners with a low level of understanding require
more time to solve problems, than learners with high
level understanding (Felder and Brent, 2005).
Thus, during the tutorial if the learners are identified
to have a low level of understanding if they take a
long time to answer a question. In this scenario,
Abdullah CITS assumes that the learner is struggling
with the tutoring content or the learners have not
understood the question. Therefore, Abdullah CITS
will either rephrase the questions or present the
question with an illustrated media like (i.e. picture, or
sound) to help the learner.
Hint Selection
Learners with a low level understanding require more
concrete visual hints, while the learners with high-
level understanding need more abstract hints (He et
al., 2009). As with the confusion detection, Abdullah
CITS produces hints in the form of pictures and
sounds to help the learner answer the question.
5.3 Storytelling in Abdullah CITS
Abdullah CITS implements a story telling based
learning strategy allocated to support the adoption of
knowledge to the learner. Abdullah is able to generate
multimedia presentations to tell the stories that are
related to each topic in the tutoring session using a
mixture of natural language, pictures and sounds
(Rahimtoroghi et al., 2013). Furthermore, the tutoring
content is structured and presented in way that groups
the entire learning context into related sub topics.
This ensure that each tutoring session has related
content which promotes recall and transfer of
knowledge into long term memory (Banaszewski,
2005).
6 EXPERIMENTAL
METHODOLOGY
This section will describe the experimental
methodology to test the ability of Abdullah CITS to
provide an effective tutoring session. The tutorial that
was given by Abdullah CITS was based on the
Islamic education modules, to teach the three
branches of Islam for the selected age group (10-12
years old). However, the Abdullah CITS tutorial
model is suitable for any students that are fluent with
the Arabic language, and have little previous
experience with the fundamental principles of Islam.
The sample size consisted of 58 participants in total
(38 from UK and 20 from Jordan). The sample
included both genders, and the participant ages were
ranged between year 5 and 6. The participants were
categorized into a number of groups:
TheApplicationofLearningTheoriesintoAbdullah:AnIntelligentArabicConversationalAgentTutor
365
• Group 1: 10 participants year 5 (Jordan)
• Group 2 : 10 participants year 6 (Jordan)
• Group 3: 22 participants year 5 (UK)
• Group 3: 16 participants year 6 (UK)
During the tutoring /interaction, students were
randomly presented with one of the three branches of
Islamic religion (to know you God, to know your
prophet, and to know your religion of Islam) as the
tutoring topic. During the tutoring session several
variables were captured (e.g. students questions,
answers, and behavior) and recorded, for further
analysis to predict the success of the tutoring session.
At the end of the tutoring session, students were
asked to complete a usability questionnaire. The data
gathered from the experiments was analyzed to
determine how well Abdullah CITS helped to
improve the tutoring in the taught subject.
Hypothesis 1: The success of students in a particular
tutoring method is indicative of participant’s
knowledge improvement in the taught subject.
Hypothesis 2: It is possible to adapt to the student’s
knowledge level from the tutoring discourse with an
intelligent tutoring conversational agent.
7 RESULTS AND DISCUSSION
The data gathered from the experiments was
analyzed to determine how well Abdullah CITS
helped to improve the tutoring in the taught subject.
There were two experiments designed to test the
hypotheses in the previous section. The results and
analysis of the experiments designed to answer the
hypotheses are outlined in the following sections.
7.1 Experiment 1: Tutoring Success
This experiment tests the hypothesis H1, and is
conducted to test the tutoring success of the Abdullah
CITS. This experiment is based on the log file that
records the dialogue between the user and the
system. A number of objective and subjective
metrics (illustrated in
Table 1), were used to verify if
Abdullah CITS led to satisfactory learning results).
Table 1: Experiment 1 Metrics.
Metric to be
Evaluated
Mode of
Evaluation
Subjective /
Objective
Pre and Post Test Log file Objective Metric
Answers Log file Objective Metric
Completion Time Log file Objective Metric
Quality of tutorial Questionnaire Subjective Metric
Tutoring content Questionnaire Subjective Metric
The generalized linear model (GLM) has been
employed in this experiment to analyze the score of
pre-test and post-test between different factors (i.e.
year group, location, and gender). The results suggest
a strong statistically significant relation, (p value less
than 0.001) between the students score before the
tutoring (pre-test) and after the tutoring (post-test)
scores. There is a significant difference (p value
equal to 0.001), between the students in year 5 and
the students in year 6, independent of the students
location (UK or Jordan). There is also a significant
difference (p value of 0.023) between the students in
the UK, and the students in Jordan. During the
tutoring Abdullah CITS recorded a value between, 1
to 3 for each response that is answered correct, and
then calculate the accumulative average for each sub
topic until the end of the tutoring session. The log
file record the score value (Highly corrected,
partially corrected, and near miss answers).The
accumulative average for each branch will compare
against the best accumulative average. As an
example the first branch cover 15 subtopics and
assuming that student got the highly corrected
answer for each subtopic, that will gave a value of 45
as the best accumulative average for first branch at
the end of tutorial. Comparing the best accumulative
average for each branch against the observed values
gave percentages of (67.78%, 72.35%, and 58.26%
for the first second and third braches respectively).
Completion time is an important metric for most
dialogue systems, and can be measured in terms of
how much time, a given task takes to complete
(Forbes-Riley and Litman, 2011). The log file
captured the completion time of each tutoring branch
delivered to the user. This time was used as a metric
in order to gauge the student level of knowledge and
understanding and used in future tutoring branches to
adapt the learning style in order to increase tutoring
success. The quality of tutoring was examined after
the Abdullah CITS tutoring session has ended by
giving the students a questionnaire which aims to
find out whether they are learning from the tutoring
session with Abdullah CITS or not. Tutoring quality
was examined after the students are rating the
questioner, such as the question (Do you agree that
there are too much to learn in one tutorial?). The
questionnaire results show that the students are quite
happy about the information content that was given
by Abdullah with a percentage of 44.8% stated they
have a neutral feeling about the learning content in
the tutoring session. 13.8% of students were not
happy with amount of information in the tutoring
session. Tutoring content was examined by asking
the students (Does Abdullah the tutor overload you
ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence
366
with information?). Students in general found
Abdullah CITS not overloads them with information,
the majority of students from the whole sample have
a neutral feeling (58.6%), a quarter of the students
(25.9%) are happy, and (15.5%) of the students felt
not happy when rating this question.
7.2 Experiment 2: Adaptability to
Student Knowledge Level
This experiment tests the hypothesis H2 (It is
possible to adapt to the student’s knowledge level
from a tutoring discourse with an intelligent tutoring
conversational agent). Hypothesis H2 was tested
based on a number of metrics recorded from the log
file as well as user questioner, as shown in Table 2
:
Experiment 2 Metrics.
.
Table 2: Experiment 2 Metrics.
Metric to be
Evaluated
Mode of
Evaluation
Subjective /
Objective
Answers
classification
Log file Objective Metric
Questions
classification
Log file Objective Metric
Quality of tutorial,
and teaching
Questionnaire Subjective
Metric
Results obtained from the log file, have been used for
measuring and evaluating two metrics, which are
(answers classification, and questions classification).
Answers classification metrics show that the
Jordanian year 6 students demonstrate more
understanding and comprehension to the tutoring
topics during the tutoring session than year 5 UK
students, with a statistically significant relationship
between the two groups (p value = 0.0001). In
addition, year 6 Jordanian students are more likely to
recognize the questions been asked in the tutoring
session in comparison with year 5. The results
obtained show a statistically significant difference
between the two year groups (p value = 0.00005).
Two questions related to H2, were asked to the
students subsequent to their interaction with
Abdullah CITS, to measure their feeling to support
the hypothesis. The first question (Does Abdullah the
Tutor provide you with information that you
understand?), and the second question (Is it right, that
Abdullah the tutor does not provide too much
information to remember?). The first question is
related to the tutoring content, the results show that
the majority of UK students have a neutral feeling
when rating this question (48.6%), while there are
only a small portion of Jordanian students have the
same feeling (12.01%). The second question is linked
to the quality of teaching, the results reveal that the
Jordanian year 6 students are happier with the
teaching from Abdullah (79.4%), compared to the
year 5 UK students (20.6%). It can be concluded that
Arab students who are being taught Arabic and the
Islamic education in their curriculum, and the effect
of the Arabic Islamic environment in the Arabic
country such as Jordan enjoyed their interaction with
Abdullah more than the UK Students who are less
exposed to this curriculum.
8 CONCLUSIONS
This paper has presented a novel methodology for
implementing learning theories within Abdullah
CITS. The aim of Abdullah CITS is to teach the
students between the ages of 10-12 years old the
fundamentals of Islam, using both modern and
classical Arabic language. Gagne, and Piaget
learning theory, and storytelling learning method are
implemented in Abdullah CITS to teach new
knowledge to students. The results highlighted that
the adoption of several key learning theories has
made the Abdullah CITS a more intelligent and
realistic tutor. The results demonstrate Abdullah is
able to adapt and adjust the level of the tutoring
session in order to keep the student engaged, through
adjusting the questions (based on the students
understanding), adjusting the material (sounds,
picture etc.), and providing feedback (encourage-
ment, and hints). The learning theories implemented
in this paper illustrated the benefits of incorporating
learning theories to develop an ITS system to make
the ITS more effective as a tutor. Through implemen-
tation of the well-established learning theories that
are used throughout modern education systems into a
CITS the student learning experience and knowledge
gain has been enhanced. The comparison between the
UK and Jordanian students shows Abdullah’s ability
to adapt to the different users abilities, the UK
students study Arabic part time, whereas the
Jordanian students study Arabic full time. The
results demonstrate that Abdullah CITS was able to
adapt the tutoring to suit both levels.
REFERENCES
Abdelsalam, U. M. A Proposal Model Of Developing
Intelligent Tutoring Systems Based On Mastery
TheApplicationofLearningTheoriesintoAbdullah:AnIntelligentArabicConversationalAgentTutor
367
Learning. The Third International Conference On E-
Learning In Education, 2014. 106-118.
Ahmad, I. 2011. Religion And Labor: Perspective In Islam.
14, 589-620.
Al-Sadan, I. 2000. Educational Assessment In Saudi
Arabian Schools. Assessment In Education: Principles,
Policy & Practice, 7, 143-155.
Alobaidi, O. G., Crockett, K. A., O'shea, J. D. & Jarad, T.
M. Abdullah: An Intelligent Arabic Conversational
Tutoring System For Modern Islamic Education.
Proceedings Of The World Congress On Engineering,
2013.
André, E. & Pelachaud, C. 2010. Interacting With
Embodied Conversational Agents. Speech Technology.
Springer.
Anglo, E. A. & Rodrigo, M. M. T. Can Affect Be Detected
From Intelligent Tutoring System Interaction Data.
Intelligent Tutoring Systems, 2010. Springer, 260-262.
Banaszewski, T. M. 2005. Digital Storytelling: Supporting
Digital Literacy In Grades 4–12. Georgia Institute Of
Technology.
Boudabous, M. M., Chaaben Kammoun, N., Khedher, N.,
Belguith, L. H. & Sadat, F. Arabic Wordnet Semantic
Relations Enrichment Through Morpho-Lexical
Patterns. Communications, Signal Processing, 2013 1st
International Conference On, 2013. Ieee, 1-6.
Carmona, C. & Bueno, D. 2007. Evolution Of An
Educational Game For Spanish Orthography. Journal
Of Computers, 2, 9-16.
Cheng, Q., Cheng, K., Li, H., Cai, Z., Hu, X. & Graesser,
A. Autotutor 2013: Conversation-Based Online Its
With Rich Media. Artificial Intelligence In Education,
2013. Springer, 930-931.
Crockett, K., James, O. S. & Bandar, Z. 2011. Goal
Orientated Conversational Agents: Applications To
Benefit Society. Technologies And Applications.
Springer.
D’mello, S., Lehman, B., Sullins, J., Daigle, R., Combs, R.,
Vogt, K., Perkins, L. & Graesser, A. A Time For
Emoting: When Affect-Sensitivity Is And Isn’t
Effective. Its, 2010. Springer, 245-254.
Felder, R. M. & Brent, R. 2005. Understanding Student
Differences. Journal Of Engineering Education, 94,
57-72.
Ferguson, K., Arroyo, I., Mahadevan, S., Woolf, B. &
Barto, A. Improving Intelligent Tutoring Systems:
Using Expectation Maximization To Learn Student
Skill Levels. Its, 2006. Springer, 453-462.
Forbes-Riley, K. & Litman, D. 2011. Designing And
Evaluating A Wizarded Uncertainty-Adaptive Spoken
Dialogue Tutoring System. Computer Speech &
Language, 25, 105-126.
Gagné, R. M. & Gagné, R. M. 1985. Conditions Of
Learning And Theory Of Instruction.
Gagne, R. M., Wager, W. W., Golas, K. C., Keller, J. M. &
Russell, J. D. 2005. Principles Of Instructional
Design.Online Library.
Ghadirli, H. M. & Rastgarpour, M. 2013. A Web-Based
Adaptive And Intelligent Tutor By Expert Systems.
Advances In Computing And Information Technology.
Springer.
Graesser, A. & Mcnamara, D. 2010. Self-Regulated
Learning In Learning Environments With Pedagogical
Agents Interact With Natural Language.Educational
Psychologist, 45, 234-244.
Groff, J. S. 2013. Expanding Our “Frames” Of Mind For
Education And The Arts. Harvard Educational Review,
83, 15-39.
Habash, N. 2012. Mt And Arabic Language Issues.
He, Y., Hui, S. C. & Quan, T. T. 2009. Automatic
Summary Assessment For Intelligent Tutoring
Systems. Computers & Education, 53, 890-899.
Hijjawi, M. D. 2011. Arabchat: An Arabic Conversational
Agent. Huertas, C. & Juárez-Ramírez, R. 2013.
Developing An Its For Vehicle Dynamics. Procedia-
Social And Behavioral Sciences, 106, 838-847.
Kim, H.-H., Taele, P., Valentine, S., Liew, J. & Hammond,
T. 2014. Developing Intelligent Sketch-Based
Applications To Support Children’s Self-Regulation
And School Readiness.
Kruse, K. 2010. Gagne's Nine Events Of Instruction: An
Introduction. Beginner Basics.
Li, N. 2013. Integrating Representation Learning And Skill
Learning In A Human-Like Intelligent Agent. Marco, J.,
Cerezo, E., Baldassarri, S., Mazzone, E. & Read, J. C.
Bringing Tabletop Technologies To Kindergarten
Children. Proceedings Of The 23rd British Hci Group
Annual Conference On People And Computers:
Celebrating People And Technology, 2009. British
Computer Society, 103-111.
Mcnamara, D. S., Crossley, S. A. & Roscoe, R. 2013.
Natural Language Processing In An Intelligent Writing
Strategy Tutoring System. Behavior Research
Methods, 45, 499-515.
Mokhtar, N. H., Kamarulzaman, M. F. A. H. & Syed, S. Z.
2011. The Effectiveness Of Storytelling In Enhancing
Communicative Skills. Procedia-Social , 18, 163-169.
Moll, Y. 2010. Islamic Televangelism: Religion, Media
And Visuality In Contemporary Egypt. Arab Media &
Society, 10, 1-27.
Monem, A. A., Shaalan, K., Rafea, A. & Baraka, H. 2008.
Generating Arabic Text In Multilingual Speech-To-
Speech Machine Translation Framework. 22, 205-258.
O'shea, K., Crockett, K. & Bandar, Z. Application Of A
Semantic-Based Conversational Agent To Student
Debt Management. Fuzzy Systems, 2010 Ieee
International Conference On, 2010. Ieee, 1-7.
O’shea, K. 2012. An Approach To Conversational Agent
Design Using Semantic Sentence Similarity. Applied
Intelligence,
37, 558-568.
O’shea, K., Bandar, Z. & Crockett, K. 2010. A
Conversational Agent Framework Using Semantic
Analysis. International Journal Of Intelligent
Computing Research (Ijicr), 1.
Owda, M., Bandar, Z. & Crockett, K. 2011. Information
Extraction For Sql Query Generation In The
Conversation-Based Interfaces To Relational
Databases. Agent And Multi-Agent Systems:
Technologies And Applications. Springer.
ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence
368
Piaget, J. & Mussen, P. 1970. Carmichael’s Manual Of
Child Psychology. Vol. New York, 703-730.
Rahimtoroghi, E., Swanson, R., Walker, M. A. &
Corcoran, T. Evaluation, Orientation, And Action In
Interactive Storytelling. Ninth Artificial Intelligence
And Interactive Digital Entertainment Conference,
2013.
Roll, I., Aleven, V., Mclaren, B. M. & Koedinger, K. R.
2011. Improving Students’ Help-Seeking Skills Using
Metacognitive Feedback In An Intelligent Tutoring
System. Learning And Instruction, 21, 267-280.
Ryding, K. C. 2005. A Reference Grammar Of Modern
Standard Arabic, Cambridge University Press.
Sani, S. & Aris, T. N. 2014. Computational Intelligence
Approaches For Student/Tutor Modelling. Sottilare, R.
A. & Proctor, M. 2012. Passively Classifying Student
Mood And Performance Within Intelligent Tutors.
Journal Of Educational Technology & Society, 15.
Stipek, D. 2013. Mathematics In Early Childhood
Education: Revolution Or Evolution? Early Education
& Development, 24, 431-435.
Van Gils, F. Potential Applications Of Digital Storytelling
In Education. 3rd Student Conference On It, 2005.
Yang, Y.-T. C. & Wu, W.-C. I. 2012. Digital Storytelling
For Enhancing Student Academic Achievement,
Critical Thinking, And Learning Motivation.
Computers & Education, 59, 339-352.
Yin, L., Bickmore, T. & Cortés, D. E. The Impact Of
Linguistic And Cultural Congruity On Persuasion By
Conversational Agents. Intelligent Virtual Agents,
2010. Springer, 343-349.
TheApplicationofLearningTheoriesintoAbdullah:AnIntelligentArabicConversationalAgentTutor
369
