Aljameel, Sumayh, O’Shea, James, Crockett, KA ORCID logoORCID:
https://orcid.org/0000-0003-1941-6201 and Latham, Annabel ORCID lo-
goORCID: https://orcid.org/0000-0002-8410-7950 (2018) Can LANA CITS
Support Learning in Autistic Children? A case study evaluation. In: Intelli-
gent Systems and Applications Proceedings of the 2018 Intelligent Systems
Conference (IntelliSys) Volume 2, 06 September 2018 - 07 September 2018,
London, UK.
Downloaded from:
https://e-space.mmu.ac.uk/620505/
Version: Accepted Version
Publisher: Springer
DOI: https://doi.org/10.1007/978-3-030-01057-7
Please cite the published version
https://e-space.mmu.ac.uk
Intelligent Systems Conference 2018
6-7 September 2018 | London, UK
1 | P a g e
Can LANA CITS Support Learning in Autistic
Children? A case study evaluation
Sumayh S. Aljameel, James D. O’Shea, Keeley A. Crockett, Annabel Latham
Department of Computing, Math and Digital Technology
Manchester Metropolitan University
sumayh.s.aljameel@stu.mmu.ac.uk, {j.d.oshea, k.crockett, a.latham}@mmu.ac.uk
Abstract—LANA CITS is a Conversational Intelligent
Tutoring System that uses the Visual, Auditory, and Kinaesthetic
learning style (VAK). It supports learning in autistic pupils, who
are studying in mainstream primary schools. Facilitating the
learning of these pupils using traditional teaching within
mainstream schools is complex and poorly understood. This paper
present investigation into how LANA CITS using VAK learning
style model can adapt to autistic pupils learning style and improve
their learning in mainstream schools. This paper provides a case
study evaluation of three children with high-functioning autism
examining the effectiveness of learning with LANA CITS. The case
study took place in primary school in Saudi Arabia. The results
were positive with the students engaged in the tutorial and the
teacher noticed some improvement over classroom activities. This
results support for the continuing development, evaluation, and
use of CITS for pupils with autism in mainstream schools.
Keywords— Intelligent Tutoring System; Autism Spectrum
Disorder; Conversational Agent; VAK learning style.
I. INTRODUCTION
Pupils with Autism Spectrum Disorder (ASD) are affected in
different degrees in terms with their level of intellectual ability.
Some people with Asperger syndrome or high functioning
autism are very intelligent academically but they still have
difficulties in social and communication skills [1]. In recent
years, many of these pupils are taught within mainstream
schools. However, the process of facilitating their learning and
participation remains a complex and poorly understood area of
education [2, 3]. Many teachers in mainstream schools
encourage the principles of inclusive education, but they do not
feel that they have the enough training to teach autistic pupils
[4, 5]. Many pupils with ASD tend to be affected differently as
a result of the same symptoms, so it is important to know how
each child is impacted on in terms of their cognitive and
interaction abilities. Hence, the specific requirements of each
child, which is known as learning style, should be identified [6].
Pupils with ASD are more likely to rely on one style of learning,
which could be determined by observing the person [7]. VAK
theory is considered to be one of the classical learning theories
in the educational field [8]. It classifies learners by sensory
preferences: Visually (V), Auditory (A), and Kinaesthetic
(K)[9]. Many researches [10-12] found that many people with
autistic spectrum are visual learners as they can learn well
through watching videos, pictures and movies. Another study
[13] found that Autistic people have learning preferences of
kinaesthetic, visual, then auditory. Roberts [14] found that
people with autism who are kinaesthetic learners, learn better
when include “hands-on” activities with the teacher taught
material to achieve learning. Nevertheless, research that has
been conducted with autistic children regarding the impact of
learning style is limited and need further research.
Technologies have attracted increasing attention in the autism
community for their educational [15-17]. Several studies [18,
19] reported that the majority of people with autism feel
positive towards technology and computer-based training. This
is attributable to the fact that software programs offer a
structured and predictable environment, which accommodates
the autistic children preferences for routine and repetitive
behaviours [20]. Truong [21] reviewed 51 studies of the recent
applications which examined the integration of learning styles
theories in adaptive learning system. In this review, it reported
a positive impact of adaptive systems using learning styles on
students satisfaction. No known intelligent tutoring system
using the VAK learning style model has been specific
developed for autistic children.
This paper is a part of our investigation within LANA CITS
project [22, 23]. This project examines the research and
development of a novel Arabic conversational intelligent
tutoring system (CITS) for children with autism spectrum
disorder (ASD) using the Visual, Auditory and Kinaesthetic
learning styles model (VAK). The project aims to design and
implement an Arabic CITS for children with autism aged 10-16
using the (VAK) model. The CITS architecture used the pattern
matching (PM) and the Arabic short text similarity (STS) to
extract the responses from resources in particular domain
(Science subjects). This paper presents a case study evaluation
of LANA CITS used by three pupils with high-functioning
autism or Asperger’s, who are studying in mainstream schools
in Saudi Arabia. The case study approach was chosen for this
study because characteristics of pupils with ASD differ widely
from each other. In addition, large groups of pupils with ASD
who have similar characteristics are difficult to recruit. The case
study examines the effectiveness of using LANA CITS and
whether the autistic pupils satisfied with the usability of LANA
CITS. The main contribution of this paper is that it is the first
case study evaluation of CITS using VAK model for pupils with
high functioning autism in mainstream schools. This paper is
Intelligent Systems Conference 2018
6-7 September 2018 | London, UK
2 | P a g e
organised as follows: Section II provides a brief overview of
LANA CITS. Section III describes the case study methodology.
Section IV presents the results and discussion and Section V the
conclusion.
II. LANA CONVERSATIONAL INTELLIGENT TUTORING
SYSTEM
A. Background
Conversational Intelligent Tutoring Systems (CITS) provide a
natural language interface to an online tutoring system to allow
users to learn topics through discussion as they would in the
classroom. Many researchers have developed CITS for
different domains, such as AutoTutor [24] which is a computer
tutor that is designed to assist students within university to learn
an introductory computer literacy course. However, AutoTutor
does not consider the student learning styles during a tutoring
session. Oscar [25] is another example of CITS, which
overcomes AutoTutors limitation by predicting and adapting
dynamically to a student’s learning style. Oscar is developed
for students in University’s level and teaches SQL course. In
the Arabic language, CITS is considered as a new area of
research. Abdullah CITS [26] is a recent of Arabic CITS for
modern Islamic education. Abdullah CITS is an online system
that aims to teach children aged 10 to 12 years old essential
topics in Islam. Abdullah CITS does not consider the student
learning styles. To our knowledge, no academic research exists
on the Arabic conversational intelligent tutoring system for
Autistic children.
B. LANA CITS
LANA CITS [22] [23] is a novel Arabic CITS, which delivers
topics related to the science subject by engaging with the user
in Arabic language. LANA uses Visual, Auditory, and
Kinaesthetic (VAK) learning style model to personalise the
lesson according to the user learning style preference. The main
feature of LANA CITS are:
Ability to control the conversation through context.
The domain used to script the CA was from the
Science curriculum for the age (10-12 years old). The
material was provided by the Ministry of Education in
Saudi Arabia [27]. Scripts consisting of a context
structured according to the topics in the science book
such as the Earth, Moon, Solar System, eclipse, etc.).
Ability to personalise the lesson with the user’s
learning style (VAK) and provide suitable material to
the user according to the user’s learning style (images
and videos for Visual learner, Sound for Auditory
learner, physical models for Kinaesthetic learner).
A conversation agent scripting language, which
provides Arabic dialogue for LANA CITS allowing
capturing variables from the user utterance.
A novel conversation agent engine using the pattern
matching (PM) technique was first used to find a
suitable response to the user’s utterance. If it is not
found, the engine uses the Short Text Similarity (STS)
technique to find a response. Combining the
techniques is beneficial to reduce the number of
scripted patterns.
Figure 1 shows a screenshot of the conversation interface of
LANA CITS, where the system converses with the student by
asking him/her a question about a specific science topic and
then providing the suitable response based on the student’s
utterance.
Fig. 1. LANA CITS graphics user interfaces.
III. CASE STUDY
This section presents the case study, which took place in a
primary school in Saudi Arabia. Ethical informed consent was
obtained from all participants, parents and the School. The
study aims to test hypothesis H1 (LANA CITS can support
learning in autistic children).
A. Participants
In this study, three participants who met the following criteria
were interviewed and observed:
Participant with age group (10-12) years old.
He/she had previously received a community clinical
diagnosis of high functioning autism spectrum
disorder or Asperger’s, indicating that they have no
problem with language and intellectual impairments.
His/her first language was Arabic.
Studying in mainstream primary schools in Saudi
Arabia.
No previous experience using LANA CITS.
To ensure data privacy, the names of participants have been
changed in this paper. In this study, three autistic students are
involved and all of them are female (Mariam, Fatimah, and
Dana). Two of them (Mariam and Fatimah) are in (Grade 5),
and their age is 11 years old, the third students (Dana) is in
(Grade 4) and her age is 10 years old. The three participants
are studying in mainstream primary school, which provided
educational provision for children with ASD in Saudi Arabia.
All three participants have Arabic as their first language and
none of them had a previous experience using LANA CITS.
Each child had previously received a community clinical
diagnosis of Asperger’s, who they have no problem with
language and intellectual impairments.
Intelligent Systems Conference 2018
6-7 September 2018 | London, UK
3 | P a g e
B. Ethical Considerations
Because the participants were children, an informed consent
form and information sheet were sent to the participant’s
parents and the School. The information sheet outlined the
details and purpose of the study, and the benefits for
participating in the study. They also informed that the
information gathered from the participant is stored securely and
none of children’s real names will be recorded.
C. Data Collection Methods
There are several data collections methods in qualitative
research. Marshall and Rossman [28] identified three primary
methods that qualitative researchers use for gathering
information. Primary methods involve gathering information
from the source directly. These three methods are: 1)
participating in the setting, 2) observing directly, and 3)
interviewing in depth. In this study, two primary methods are
used: Observation and Interviewing. “Observation involves the
systematic noting and recording of events, behaviors, and
artifacts in the social setting,” [28]. Kvale and Brinkman [29]
defined the interview as a conversation with structure and a
purpose. There are secondary methods that may also be used to
gather information. This involves gathering information from
someone or something else outside of the original source by
analysing documents such as: videos, photography, interaction
analysis, and questionnaires [28].
1. Interviewing
Interviewing is a universal methods in qualitative research [30,
31]. Interviews are typically divided into three categories:
structured, unstructured, and semi-structured [32]. In structured
interviews, all participants will have the same questions using
the same wording and in the same order [33]. However, while
a structured interview ensures that comparable responses are
gathered from each participant, it leaves no room for
elaboration [34]. The second category is an unstructured
interview, which contains open questions and subsequent
questions depending on the participant’s responses[35]. The
drawbacks of this type of interview are the difficulty of
processing data and the effort required. It is difficult to find
similar statements from different participants and hence the
links are often difficult to make. The most common type of
interviews used in qualitative research are the third category,
which is semi-structured interviews. It involves structured
questions and the researcher is free vary the order and wording
of the questions [32], depending on the direction of the
interview, and to ask additional questions. According to Berg
[34], the researcher is able to adjust the wording of the
questions, and add or delete probes, as necessary. With this
structure, the interviewer guided the interview, to obtain a good
description from the participant’s perspective. Moreover, this
structure allowed the researcher the flexibility to dive deeper
when appropriate. Creswell [36] asserted that the semi-
structured process lets the researcher to explore and ask follow-
up questions if necessary to clarify the particular subject. This
was an extremely beneficial process in the field of qualitative
interviewing.
The Semi-structured interviewing was used in this study
because the participants are children and the level of
comprehension may vary. The interview was interactive and the
participants were able to ask questions. In addition, notes were
taken during the interview to capture the core of the
participant’s comments and to note any non-verbal behaviour
and facial expressions. In this study, two interviews were
conducted for each student, the first interview was with the
autistic student and the second interview was with the teacher.
The first interview, which was with the autistic student,
contains of nine questions and the interviewer was free to ask
follow-up questions. Table 1 shows the nine questions that are
used in the interview. The questions were asked in Arabic.
TABLE I. THE INTERVIEW SHEET WITH THE AUTISTIC STUDENT.
1. Do you think LANA Tutor is enjoyable to use?
Which part do you like?
Why do you like this part?
2. Do you think LANA had a right amount of information in one
Tutorial?
3. Do you remember what LANA has taught you in the tutorial?
Tell me in one or two things you learned from
LANA
4. Did you like using your learning style (Visual – Auditory -
Kinaesthetic) in LANA?
Would you like to use your learning style with all
subjects and tutorials?
5. Do you like the way LANA looked?
What do you think of the color of LANA?
Any suggestion to add in the design or the
character.
6. Do you think you can learn from LANA better than your
teacher?
7. Would you like to use LANA again with other lessons?
8. Is talking with LANA like talking with your friend?
9. Would you like to use LANA in the classroom?
The second interview was conducted with the teacher. The aim
of this interview is to measure the student’s interaction and
behaviour from the teacher’s point of view. The interview
contains five questions as shown in Table 2.
TABLE II. THE INTERVIEW SHEET WITH THE TEACHER.
1. Is there any behavioural issue with the student?
2. What is the student’s education level in the school?
3. How was the student’s interaction during the tutorial?
4. Did the student interact with the LANA’s tutorial in the same
way as the class tutorial?
5. Do you have any suggestions to add to LANA?
2. Participant Observation
Intelligent Systems Conference 2018
6-7 September 2018 | London, UK
4 | P a g e
In addition to semi-structured interviews, the participant has
been observed during the interview by Sumayh, the lead
researcher of this study. Observations allow the interviewer to
observe the context in which the participants interact [36].
Additionally, observations allow the researcher to notice the
participant’s behaviors, facial expressions, and the interactions.
D. Data Analysis
Following the interviews the data was analysed using the
following methodology [36].
1. Data management: This involves organizing and
gathering all the data. In this study, the interview was
transferred into a written format.
2. Reading and noting: This involves reading through all the
transcriptions and making notes in the margins regarding
the interviewee’s comments.
3. Describing: This involves describing the participant’s
experience with the study.
4. Classifying: The researcher then makes a list of significant
statements, quotes that emerged from the data and then
grouped together.
5. Interpreting: the researcher develops textural and
structural descriptions of the data. Textural descriptions
are the written descriptions of what the participants
experienced. After the textural descriptions are written,
the researcher outlined the contextual influence, and this
is the structural description.
6. Representing and visualizing: After both the textual and
structural descriptions are written, the researcher writes
the findings of the study. In addition, tables will be created
to help illustrate the data.
E. Participant’s interaction
The study started by detecting the participant’s learning style
using a child friendly-customised VAK learning style
preferences questionnaire, which was answered by the student
with the help of the teacher. The LANA CITS registration
interface was then shown to the participant and the lead
researcher registered the participant with her learning style into
the system in order to adapt the tutorial according to her
learning style. Then, a pre-test interface started (see Figure 2)
and the participant was asked to complete the pre-test by
choosing the right answer using the touch screen. When the pre-
test was completed, the tutorial started and the participant
interacted with the tutorial and answered the questions related
to the science subject. During the tutorial, materials were used
according to the learning style (Visual: video and images,
Auditory: sounds, Kinaesthetic: physical models). After the
participant completed the tutorial, a post-test interface started
and the participant answered the questions in the post-test.
When the tutorial was completed a verbal interview (see Table
1) was conducted in order to measure the participant
satisfaction and their opinion about using LANA CITS. During
the tutorial and the interview, the student was observed by the
lead researcher and the teacher to notice the participant’s
behaviors, facial expressions, and the interactions. Support was
provided to the participant if asked. When the student finished
and left the room, the teacher was interviewed by the lead
researcher (see Table 2).
Fig. 2. Pre-test Questions interface.
IV. RESULTS AND DISCUSSION
In this section, the data that was gathered from the interviews
(Table 1, Table 2) and the observation were described and
classified in order to test Hypothesis H1 (LANA CITS can
support learning in autistic children).
A. First participant Results and discussion
Mariam: The first participant interviewed in this study, was
Mariam. She is eleven years old in Grade 5 and she has an early
community clinical diagnosis of Asperger’s, with no problem
with language and intellectual impairments. Based on the
teacher’s interview, Mariam made an improvement in her
behaviour during her study in the school from Grade 1 to 5. Her
education level is very good especially in Maths and English.
She is interested in using technology such as a computer, IPad,
and mobile devices. The study begun by detecting Mariam’s
learning style using the questionnaire. The results from the
questionnaire showed that her learning style was Visual.
Mariam completed the pre-test and then interacted with the
tutorial and answered the questions. During the tutorial,
videos/images appeared for each information in the tutorial (see
Figure 3). After she completed the tutorial, a post-test interface
started and she answered the questions in the post-test. Mariam
enjoyed using LANA CITS as indicated in her interview
answers. She said that she especially liked watching the videos
and the pictures during the tutorial. She was able to use LANA
CITS independently with ease and required no additional
support once told in how to use it. Her learning gain was
improved when comparing the pre-test and the post-test scores.
Table 3 shows the pre-test and the post test scores for all three
participants. She was able to remember what she had learned
from LANA, when she was asked in the interview. The rest of
the interview questions were positive but when she was asked
if she like learning from LANA CITS more than learning from
her teacher, she answered ‘Maybe’.
