International Journal of Engineering Education Vol. 28, No. 6, pp. 1366–1372, 2012
Printed in Great Britain
0949-149X/91 $3.00+0.00
# 2012 TEMPUS Publications.
Artificial Intelligence in Special Education:
A Decade Review*
ATHANASIOS S. DRIGAS and RODI-ELENI IOANNIDOU
N.C.S.R. Demokritos, Institute of Informatics and Telecommunications, Telecoms Lab—Net Media Lab,
Ag. Paraskevi, 15310, Athens, Greece. E-mail: dr@iit.demokritos.gr, elena.ioan@hotmail.com
Artificial Intelligence (AI) technology has developed computer tools for carrying out a number of tasks, simulating the
intelligent way of problem solving by humans. AI techniques have also been identified as one of the most valuable
applications in the field of special educational needs (SEN). The goal of these tools is to enhance the way children interact
with their environment to promote learning and to enrich their daily life. Due to the implicit characteristics of special
educational needs, the diagnosis has been an issue of major importance. At the same time intervention strategies need to be
highly individualized to be effective. In this report we introduce some of the most representative studies over the last decade
(2001–2010), which use AI methods in making accurate diagnosis and prompt intervention action.
Keywords: artificial intelligence; special educational needs; diagnosis; intervention
1. Introduction
The steady progress in the area of technology has
devolved computing power into many aspects of
our daily life. Across the educational sector there
has been an increased trend to increase the accessibility of education. A large volume of research is
currently addressing the use of computers in education in order to develop learning environments,
which support the learning process in different
settings [1]. Several years ago various researchers
and specialists of computing science have started to
study the implementation of Artificial Intelligence
techniques in education.
Artificial Intelligence (AI) has been an active area
of research for over fifty years [2]. It is usually
defined as the study and development of intelligence
agents that can perceive their environment and take
actions that increase their possibilities of success [3].
Artificial intelligence agents can be either in a
physical form of the device (e.g. humanoid robots)
or in software form with ‘intellectual’ capacity (e.g.
a virtual avatar). The nature of technology has
changed since a few years later Artificial Intelligence
in Education (AIE) was conceptualized as separate
research community. AI techniques in education
were claimed to create powerful learning environments and to increase positive interactive experiences for all students. Some of the most typical AI
applications in the educational field involve knowledge representation, intelligent tutoring, natural
language processing, autonomous agents etc.
The benefits of A.I. in education have been
acknowledged for many years. However, during
the previous decade’s one of the research communities of Artificial Intelligence in Education deals
1366
with the intersection of A.I. and Special Education
[4]. The benefits of AI techniques have been gradually used to improve the life of those people with
special educational needs.
The field of ‘Special Educational Needs’ covers a
large number of difficulties which can cause problems during the learning process. Even though
various terms of special educational needs have
been presented during the past years, experts in
this field have not yet completely reach an agreement. Terms like ‘Learning Difficulties’ or ‘Learning Disabilities’ are also widely used [5]. Our scoping
study drew upon the various national and international publications and we decided to use the definition of the 2001 Special Educational Needs Code of
Practice, as a framework for organizing the literature under a manageable number of headings.
According to the 2001 SEN Code of Practice the
areas of needs are: Communication and Interaction,
Cognition and Learning, Behavior and Emotional
and Social Development, Sensory and/or Physical
[6]. Moreover, the Code of Practice highlights the
fact that not all children will progress at the same
rate and that each child is an individual, with
different strengths and needs. It is then necessary
to understand and model learners and the settings
where they interact in a way that enable us to
develop and evaluate technology to most efficiently
support learning across multiple contexts, subjects
and time [7].
Recent development in the area of Artificial
Intelligence and Special Education may enable
development of collaborative interactive environments and facilitate the life of individuals with
special educational needs and the people around
them. Our goal in this paper is to explore the
* Accepted 27 July 2012.
Artificial Intelligence in Special Education: A Decade Review
potential of the most representative of AI applications of the last decade. The studies that will be
presented in the following sections deal with diagnostic and intervention tools of some of the most
common learning difficulties. These proposed
models can be used from teachers, special educators, psychologists, therapists and parents as well.
Due to the implicit characteristics of learning difficulties, their high similarities and comorbidity of
their symptoms, AI assessment tools may be one
way to improve the teacher or parent capabilities
when evaluating the child. These tools can help
them observe the child’s academic level and if
necessary they will consider taking appropriate
decisions to advise a specialist if there is any
difficulty. AI training interventions are an important part of the education of children of special
educational needs since they are able to integrate the
freedom of action of the student with a more explicit
control and guide [8]. This paper will introduce
diagnostic and intervention tools developed in the
last decade for every one of the following categories
2. Sensory and/or physical impairments
Learners with long-term or more complex physical
impairments require educational services that will
help them to maintain their independence and wellbeing, and to lead as fulfilling a life as possible. In
most cases physical and sensory impairments are
assessed from doctors during the first years of their
lives. This is why AI applications concerning parents and teachers mostly aim at training students
rather than diagnosing their needs.
Georgopoulos et al., 2003 presented a fuzzy
cognitive map approach for differential diagnosis
of specific language impairment (SLI). Fuzzy cognitive maps are a soft computing methodology that
uses a symbolic representation for the description
and modeling of complex systems. The aim of this
tool is to provide the specialists with a differential
diagnosis of SLI from dyslexia and autism, since in
many cases SLI is difficult to be discerned due to its
similar symptoms to other disorders. The system
has been tested on four clinical cases with promising
results [9].
In the same year Schipor et al., (2003) attempted
to create a Computer Based Speech Therapy
(CBST) system using a fuzzy expert system for
helping learners with speech disorders. The aim of
this approach was to suggest optimal therapeutic
actions for every pupil based on the information
selected, so they designed an improved CBST
system, called LOGOMON (Logopedics Monitor)
and developed its classical architecture with a fuzzy
expert system based on forward chaining. The role
of the expert system was to store the precise evolu-
1367
tion and progress of each child and adapt the
exercises to each child’s current level and progress.
The validation of LOGOMON was performed by a
three month experiment which involved two equivalent children groups, taken from the Regional
Speech Therapy Center of Suceava in Romania.
The first group used LOGOMON, but the expert
system was deactivated and all therapeutically decisions were taken only by the speech therapist. The
second group used LOGOMON with inference
facilities, so that, a part of therapeutically decisions
was provided by expert system and the other one
was provided by speech therapist. The results indicated no significant difference between the two
groups. However, there were other advantages
using the expert system such as more therapy time,
predictability and the explanation of results [10].
Pavlopoulos et al., (2008) implemented a neural
network approach for the self-assessment for the
learners, optimized with the aid of Genetic Programming. The purpose of this method is to assess
the user’s answers from both single and multiple
questions in an e-learning environment. Test data of
this application evaluate the answers against the five
areas of learning: grammar/sentence structure,
reading, writing, letter recognition and alphabetical
order, spelling/vocabulary. The implementation of
the Genetic Programming Neural Network
(GPNN) methodology for e-learning purposes is
effective for all students who exhibit difficulties in
the above areas but can be specifically appropriate
for individuals with physical or sensory impairments. This platform was applied and evaluated
successfully the user’s answers, while the generalization of the assessment process could later lead to
the development of an intelligent e-tutor [11].
In 2008 Drigas et al., presented ‘Dedalos’ project
which deals with the teaching of the English language as a second language to hearing impaired
people, whose mother language is the Greek sign
language. In an educational e-content adapted to
the needs of every user, the whole procedure consists
of audits and evaluation of the linguistic abilities of
the e-learners. The system uses an intelligence
taxonomy system which is developed for the evaluation of the pupil and the setting of pedagogic
material. The approach promotes a complete support system for the education of hearing impaired
Greek students while at the same time opens the way
for their inclusion [12].
3. Learners with Autistic Spectrum
Disorders
Autism Spectrum Disorder (ASD) is a pervasive
developmental disorder characterized by the ‘triad’
of impairments. Children with ASD exhibit impair-
1368
ments in social skills, language and communication
skills and a tendency towards repetitive patterns of
interest and behavior [13]. AI techniques can facilitate early intervention and provide specialists with
robust tools indicating the person’s autism spectrum disorder level.
In 2006 Sebe et al., implemented an emotion
recognition computerized tool based on joint
visual and audio cues. This human-computer interaction application besides the 6 universal emotions
(happy, surprise, angry, disgust, fear and sad) is able
to recognize other affective states such as interest,
boredom, confusion and frustration. It can be used
from all children but it can be very affective in
children with speech problems as well as in training
children with ASD since they display difficulties
understanding other people’s emotion. This
approach analyzes 11 affective states, on 38 subjects
by applying Bayesian Networks for bimodal fusion.
In addition, a variable is integrated into a Bayesian
Network which indicates whether the user is speaking or not. Once the model is fitted, head motion and
local deformations of the facial features such as the
eyebrows, eyelids, and mouth can be tracked. The
recovered motions are shown in terms of magnitudes of some predefined motion of various facial
characteristics. This system was tested on 38 graduate and undergraduate students in various fields.
The results indicated that emotion recognition
accuracy is greatly improved when both visual and
audio information are applied in classification [14].
In 2007 Riedl et al, designed a platform which can
aid adolescents with High Functioning Autistic
Spectrum Disorders (HFASD) rehearse and learn
social skills with reduced help from parents, teachers, and therapists. A social scenario game is
presented— for example going to a movie theaterwhich challenges learners with HFASD to role-play
and complete tasks involving social situations.
Artificial Intelligence is used to assist the above
groups with the authoring of tailored social scenarios. An A.I. system automatically examines the
causal form of the narrative plan, searching for
points at which a student’s actions can undo
causal relationships. The alternative narrative scenario is a branch developed for handling the contingency of the learner’s action. This Artificial
Intelligence tool embed in this particular platform
decreases the manual authoring burden where
application of intervention strategies can be
handled by specialists. This social scenario intervention approach is complete and currently undergoing evaluation with promising results [15].
Arthi and Tamilarasi (2008) introduced a model
which helps in the diagnosis of autism in children by
applying Artificial Neural Networks (ANN) technique. The model converts the original autistic data
Athanasios S. Drigas and Rodi-Eleni Ioannidou
into suitable fuzzy membership values and these are
given as input to the neural network architecture.
Moreover, a pseudo algorithm is created for applying back propagation algorithm in predicting the
autistic disorder. This approach is proposed to
support apart from medical practitioners, psychologists and special educators who are involved is
assessing children. Experimental results indicated
an 85–90% accuracy of this method. In future the
autistic disorder could be predicted using k-nearest
neighbor algorithm for a comparative research [16].
4. Learners with reading, writing and
spelling difficulties
Within a classroom there is an incredible spread in
reading, writing and spelling abilities among the
pupils [17, 18]. Each and every student requires
support at their own level and for their own specific
needs. These kinds of problems tend to be diagnosed
when children reach scholar age. In most institutions there are not specialist staff in learning difficulties, it is then desirable that teachers have access
in some diagnostic and intervention tools to better
care of the students’ problems.
Srihari et al., (2008) presented two computational
method of automatic scoring of short handwritten
essays in reading comprehension tests. The aim of
this system is to assign to each handwritten response
a score which is comparable to that of a human
scorer. This tool has to contend with not only the
standard difficulties of recognizing handwriting but
also the writing skills of children. Assessing reading
comprehension tests will not only allow timely
feedback to students but also can provide feedback
to education researchers, parents and teachers. In
this study two systems are described. The first is
based on latent semantic analysis (LSA), which
needs a reasonable level of handwriting recognition
performance. The second developed an artificial
neural network (ANN) which is based on features
collected from the handwriting image. Both systems
were trained and evaluated using a test-bed of essays
written in response to prompts in statewide reading
comprehension tests and scored by humans. Even
though there were observed errors in word recognition during the evaluation of this platform scoring
performance still remains a promising tool [19].
Jain et al., (2009) proposed a model called Perceptron based Learning Disability Detector
(PLEDDOR). It is an artificial neural network
model for identifying difficulties in reading (dyslexia), in writing (dysgraphia) and in mathematics
(dyscalculia) using curriculum based test conducted
by special educators. This computational diagnostic
tool consists of a single input layer with eleven units
that correspond to different sections of a conven-
Artificial Intelligence in Special Education: A Decade Review
1369
tional test and one output unit. The system was
tested on 240 children collected from schools and
hospitals in India and was evaluated as simple and
easy to replicate in huge volumes, but provides
comparable results based on accepted detection
measures [20].
Hernández et al., (2009) introduced SEDA (‘Sistema Experto de Dificultades para el Aprendizaje’
or ‘Expert System for Learning Difficulties’ in
English) a diagnostic tool for Learning Difficulties
in children’s basic education. It is developed using
the Expert Systems design methodologies which
include a knowledge base consisting of a series of
strategies for Psycopedagogy evaluation; trying to
identify the relationships between input variables
(e.g. age, sex, educational level, reading, perception,
understanding) and the output systems (psychomotor aspect, intellectual aspect, perceptual aspect,
language aspect, personal aspect). All of the above
provides the expert system’s users the possibility of
acknowledge the psychological profile of the pupil.
80% of the evaluators rated the system as Efficient
using an estimation scale of: Poor, Moderately
Efficient and Efficient [21, 22].
In 2010 Baschera and Gross introduced an adaptive spelling training system which can be used from
all students who exhibit spelling difficulties. This
platform is based on an inference algorithm
designed to manage unclassified input with multiple
errors defined by independent mal-rules. The inference algorithm based on a Poisson regression with a
linear link function, estimates the pupil’s difficulties
with each individual mal-rule, based on the
observed error behavior. This knowledge representation was implemented in a student model for
spelling training such as optimized word selection
and lessons for individual mal-rules to pupil
adjusted repetition of erroneously spelled words.
This system was tested on a two large-scale user
studies and showed an important increase in the
learner’s performance, induced by the student
adapted training actions [23].
based classifier for the diagnosis of dyslexia with low
quality data with genetic fuzzy systems in early
childhood. It can be used by parents and school
staff for detecting those symptoms that will suggest
taking the child to a specialist for a more thorough
examination. This application consists of a fuzzy
rule based system (FRBS), whose knowledge base
is to be obtained from a sample of data by means
of a genetic cooperative-competitive algorithm
(GCCL). The FRBS includes collected data from
65 schoolchildren in Spain whose answers were
comprised to graphical tests (e.g. BENDER,
ABC) while the output variable for each dataset is
a subset of labels; no dyslexic, control and revision,
dyslexic, inattention/hyperactivity or other problems. This genetic fuzzy system can operate with
the above low quality data and provide us the
appropriate results for determining whether the
student should visit an expert. The experimental
results indicated that the FRBS from low quality
data can provide the unqualified personnel with a
diagnosis. However the percentage of misclassifications was high and future improvements need to be
made [25, 26].
Kohli et al., (2010) introduced a systematic
approach for identification of dyslexia at an early
stage by using artificial neural networks (ANN).
This approach is amongst the first attempts which
have been made for addressing the dyslexia identification problems with the use of ANN. Moreover,
it can be distinguished from other platforms of its
kind because it is based on test data, covering the
evaluation results of potential dyslexic pupils,
between the years 2003– 2007. These test data
consist of the input data of the system while the
output results classify the students in two categories
(dyslexic and non-dyslexic). An error back-propagation algorithm is responsible for mapping college
performance to the underlying characteristics. The
initial results obtained using test data were fairly
accurate and suggest the application of this platform to real data as well [27].
5. Learners with dyslexia
6. Learners with difficulties in mathematics
Dyslexia is one of the most common and most
studied cases of Developmental Disorders causing
troubles in literacy and especially in reading, writing
and spelling. It is neurologically based and lifelong
condition [24]. Diagnosing dyslexia is a complex
process that depends on many different indicators.
Even though applying artificial intelligence techniques for identifying dyslexia can be a complex
procedure, the preliminary results of recent studies
are satisfactory and open new ways in the field of
diagnosis.
In the same year Palacios et al., presented a rule-
Mathematical skills are essential to all students and
they are also a subject where many students display
various difficulties. During the latest years methods
and techniques of artificial intelligence were developed to assess the mathematical level of pupils and
to also help them acquire specifics skills [28].
Melis et al., (2001) introduced ActiveMath, a
web-based intelligent tutoring system for mathematics. ActiveMath is an Intelligence Tutoring
System (ITS), which allows the students to learn in
their own environment whenever it is convenient for
them. It uses a number of Artificial Intelligence
1370
techniques to realize adaptive course generation,
student modeling, feedback, interactive exercises
and a knowledge representation which is appropriate for the semantic Web. In ActiveMath the user
starts his/her own student model by self-assessment
of his/her mastering level of concepts and later
chooses learning goals and scenario, for instance,
the preparation for an exam. The capabilities of the
student are adapted in course generation and in the
suggestion mechanism as well. Moreover, a ‘poor
man’s eye-tracker’ is designed which is able to trace
child’s attention and reading time in detail. This
application has reported many positive outcomes in
the following years by a large number of studies, all
of them supporting the effect of this ITS during the
learning process [29–31].
Livne et al., (2007) implemented an online parsing
system that automatically assesses students’ constructed responses to mathematics questions, based
on the errors in each response. A parser is the basic
element of a free college readiness website in mathematics. During learning sessions, users are asked to
provide constructed answers to mathematical questions. The parser analyzes the students’ answers,
gives immediate feedback on their errors and provides accurate partial-credit scoring as well. This
tool apart from providing a good match to human
grader scoring, it reflects the overall response and it
also distinguishes the types of errors into two types:
conceptual and computational. The parser clearly
illustrates that natural languages and artificial intelligence principles can be applied successfully to
detect student error patterns. Overall, the system’s
total scoring closely matched human scoring, but
the parser was found to surpass humans in systematically distinguishing between students’ error patterns [32].
Anthony et al., (2008) designed an Intelligent
Tutoring System (ITS) for students learning algebra
equation-solving. Algebra is one of the subjects in
which students display several difficulties. This platform aims at improving student performance via
ITSs that accept natural handwriting input. The
type of ITS used in this method is known as
‘Cognitive Tutors’, who pose authentic problems
to students and give emphasis to learn-by-doing. In
Cognitive Tutor Algebra, students represent a given
scenario, graph functions and solve equation while
the tutor provides help and feedback. A Freehand
Formula Entry System recognizer is also used,
which has been trained from data derived from
over 40 high school and middle school algebra
students. Results from this study showed benefits
for general usability and for learning. In addition,
this platform is likely to generalize to other types of
mathematics and to other levels of learners [33].
Gonzalez et al, (2010) designed an automatic
Athanasios S. Drigas and Rodi-Eleni Ioannidou
platform for the detection and analysis of errors in
mathematical problems to support the personalized
feedback of pupils. This method is referred to all
students and particularly to students with special
educational needs such as those with Down syndrome, who exhibit difficulties in the arithmetic
operations of addition and subtraction. An error
detection algorithm was developed which is able to
analyze the data gathered as a result of the interaction between the students and the platform, while
afterwards the output of the error is available to the
teachers about the specific difficulties and to allow
them to personalize the instruction. Moreover, they
designed a model which returns the set of errors
made by the pupils in the corrected exercises so as
the students can learn from their own mistakes. The
system was tested on a group of students with Down
syndrome and the results confirm that the module
exhibits the proper behavior [34].
7. Attention Deficit Hyperactivity
Disorder (ADHD) and Attention Deficit
Disorder (ADD) learners
The terms ADHD and ADD refer to a wide range of
difficulties that become apparent at some stage
during the developmental period in a child’s life.
They are usually characterized by a set of behavior
problems of inattention, hyperactivity and impulsivity or their combination. These problems usually
show up in early childhood and more specifically
they should be present before the age of seven in
order for a diagnosis to be made [35, 36]. The use of
artificial intelligence applications has offered some
improved diagnostic and intervention tools of these
behavior difficulties.
In 2004 Rebolledo-Mendez and Freitas presented
the NeuroSky MindSet (MS) which is able to detect
attention levels in an assessment exercise by combining performance data with user-generated data,
taken from interaction. NeuroSky consists of a
headset with three electrodes, which are put beneath
the ears and on the forehead. The electrical signals
read at the above locations are used as inputs by
NeuroSky’s algorithms to assess the attention
levels. An A.I. driven avatar was also designed to
pose questions and have limited conversation with
the users. It is a low-cost, non-clinical and easy to
use tool designed for leisure. This model was tested
on first-year undergraduate students in the following years and the results indicated that there is a
positive relation between measured and selfreported levels of attention [37, 38].
Aguilar et al., (2006) designed a fuzzy instructional planner, which models the tutor module in an
intelligent tutorial system (ITS). It is an interactive
instructional method which uses a combination of
Artificial Intelligence in Special Education: A Decade Review
text, graphics, sound and video in the learning
procedure. It is especially useful for students who
have with ADHD or attention difficulties as well as
in distance learning situations. The fuzzy instructional planner consists of a rule base, an inference, a
fuzzification interface and a defuzzification interface. The aim of this system is to mimic the behavior
of the teacher able to manage learning process
satisfactorily. The input information is derived
from human expert who supplied linguistic information. The ITS is a flexible system which adapts
the teacher’s rules to the student’s performance and
it has been shown useful in several applications with
promising results [39].
Anuradha et al., (2010) developed a platform for
a more accurate and less time consuming diagnosis
of Attention Deficit Hyperactivity Disorder
(ADHD). They used one well-known Artificial
Intelligence technique, the SVM algorithm. According to the authors, this is the first attempt at
identifying ADHD using SVM algorithm. Support
vector machines (SVMs) are a set supervised learning techniques suitable for classification and regression. A data-set which was verified by a doctor
including the results of a questionnaire used by the
doctors to diagnose the disorder was given to the
SVM module. After that the data-set was introduced and afterwards returned to the SVM module,
which finally provides us with the diagnosis. The
most important advantage of applying the SVM
algorithm is that it can control the complexity of the
diagnostic process. This method was tested on
children between the ages six to eleven years old
and the results indicated a percentage of 88,674%
success in diagnosing [40].
In the same year Delavarian et al., (2010) introduced a decision support system to distinguish
children with ADHD from other similar children
behavioral disorders such as depression, anxiety,
comorbid depression and anxiety and conduct disorder based on the signs and symptoms. A differential diagnosis of the above mentioned behavioral
disorders is of major importance and practically
difficult due to their similarities and comorbidity of
their symptoms. This tool was initially developed in
assisting psychiatrists but it can also be used in
schools for a more specific examination of highrisk students. For designing the decision support
system two types of neural networks were compared: radial basis function (RBF) and multilayer
perceptron (MLP) neural networks. The system was
trained and validated to assist the diagnosis of the
disorders. The system was tested on 294 children of
12 elementary schools. The classification by MLP
networks achieved 95.50% while the RBF classifier
reached 96.62%. The limited number of diagnostic
errors compared to the errors done by specialists
1371
indicated a system that can work as a reliable and
valid tool for ADHD assessment [41].
8. Conclusions
During the last decade an important number of
studies are currently addressing the use of Artificial
Intelligence systems in the education of students
with special educational needs. This paper drew
upon the most representative studies that try to
solve major issues in diagnosis and intervention of
specific difficulties. AI application tools have successfully been applied to solve problems in the field
of special education. Based on the studies presented
in this work, it was concluded that there is a need to
support teachers, parents and therapists in the
appropriate care to students with special educational needs, particularly in assessment and treatment methods. Saving time and cost, gaining more
therapy time, increasing the early diagnosis and
intervention efficiency by creating more efficient
learning environments are some major advantages
that AI computational tools offers us. However, the
issues to be covered in special education are still
plenty due to the wide range of difficulties and the
various needs of every individual. Further research
across all types of learning difficulties and nationally
regulated adaptations of diagnostic tools have to be
resolved in order to relieve teachers and parents
workload. Nevertheless, artificial intelligence has
been considered as a promising educational aiding
tool for all children who call for an embracing and
cooperative approach to service delivery.
References
1. L. Nanni and A. Lumini, Ensemble generation and feature
selection for the identification of students with learning
disabilities, Expert Systems with Applications, 36, 2008, pp.
3896–3900.
2. B. G. Buchanan, A very brief history of artificial intelligence,
Artificial Intelligence Magazine, 25th anniversary issue, 2005,
pp. 53–60.
3. S. J. Russell and P. Norvig, Artificial Intelligence: A Modern
Approach (2nd ed.), New Jersey, 2003.
4. T. K. Wu, Y. R. Meng and S. C. Huang, Application of
Artificial Neural Network to the Identification of Students
with Learning Disabilities, International Conference on Artificial Intelligence, 2006, pp. 162–168.
5. J. M. Fletcher, W. Coulter, D. J. Reschly and S. Vaughn,
Alternative approach to the definition and identification of
learning disabilities: some questions and answers, Annals of
dyslexia, 54(2), 2004, pp. 304–331.
6. Department for Education and Skills, Special Educational
Needs Code of Practice, London: DfES, 2001.
7. J. Brodin, Can ICT give children with disabilities equal
opportunities in school? Improving Schools, 13(1), 2010,
pp. 99–112.
8. R. Lanzilotti and T. Roselli, An Experimental Evaluation of
Logiocando, an Intelligent Tutoring Hypermedia System,
International Journal of Artificial Intelligence in Education,
17, 2007, pp. 41–56.
9. V. C. Georgopoulos, G. A. Malandraki, C. D. Stylios, A
fuzzy cognitive map approach to differential diagnosis of
1372
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
specific language impairment. Artificial Intelligence in Medicine, 29, 2003, pp. 261–278.
O. A. Schipor, S. G. Pentiuc and M. D. Schipor, Improving
computer based speech therapy using a fuzzy expert system.
Computing and Informatics, 22, 2003, pp. 1001–1016.
J. Pavlopoulos, J. Vrettaros, G. Vouros and A. S. Drigas, The
Development of a Self-assessment System for the Learners
Answers with the Use of GPNN. WSKS, (1), 2008, pp. 332–340.
A. S. Drigas, D. Kouremenos and J. Vrettaros, Teaching of
English to Hearing Impaired Individuals Whose Mother
Language is the Sigh Language. In: M. D. Lytras et al.
(eds), 2008, pp. 263–270.
I. Rapin and R. F. Tuchman, Autism: Definition, neurobiology, screening, diagnosis. Pediatric Clinics of North America,
55(5), 2008, pp. 1129–1146.
N. Sebe, I. Cohen, T. Gevers and T. S. Huang, Emotion
Recognition Based on Joint Visual and Audio Cues, Proceedings of the 18th International Conference on Pattern Recognition, 2006.
M. Riedl, R. Arriaga, F. Boujarwah, H. Hong, J. Isbell and
L. J. Heflin, Graphical Social Scenarios: Toward Intervention and Authoring for Adolescents with High Functioning
Autism, Virtual Healthcare Interaction, Papers from the
AAAI Fall Symposium, 2007.
K. Arthi and A. Tamilarasi, Prediction of autistic disorder
using neuro fuzzy system by applying ANN technique,
International Journal of Developmental Neuroscience, 26,
2008, pp. 699–704.
B. M. Macmillan, Rhyme and reading: A critical review of
the research methodology, Journal of Research in Reading,
25(1), 2002, pp. 4–42.
L. Roberts and S. McDougall, Explaining Performance in
the Orthographic Analogy Task: Phonological Priming,
Phonological Skills and Letter-sound Knowledge. Journal
of Experimental Child Psychology, 84(4), 2003, pp. 310–337.
S. Srihari, J. Collins, R. Srihari, H. Srinivasan, S. Shetty and
J. Brutt-Griffler, Automated scoring of handwritten essays in
reading comprehension tests, Artificial Intelligence, 172,
2008, pp. 300–324.
K. Jain, P. Manghirmalani, J. Dongardive and S. Abraham,
Computational Diagnosis of Learning Disability, International Journal of Recent Trends in Engineering, 2(3), 2009.
J. Hernadez, G. Mousalli and F. Rivas, Expert System for the
Diagnosis of Learning Difficulties in Children’s Basic Education, Proceedings of the 8th WSEAS International Conference on Applied Computer and Applied Computational
Science, Italy, 2008.
J. Hernadez, G. Mousalli and F. Rivas, Learning Difficulties
Diagnosis for Children’s Basic Education using Expert
Systems. WSEAS Transactions on Information Science and
Applications, 7(6), 2009.
G. M. Baschera and M. Gross, Poisson-Based Inference for
Perturbation Models in Adaptive Spelling Training. International Journal of Artificial Intelligence in Education, 20,
2010, pp. 1–31.
P. Thomson and P. Gilchrist, Dyslexia: A Multidisciplinary
Approach, Nelson Thornes, 1996.
A. M. Palacios, L. Sanchez and I. Couso, Extending a simple
genetic cooperative-competitive learning fuzzy classifier to
low quality datasets, Evolutionary Intelligence, 2(1), 2009, pp.
73–90.
A. M. Palacios, L. Sanchez and I. Couso, Diagnosis of
dyslexia with low quality data with genetic fuzzy systems.
Athanasios S. Drigas and Rodi-Eleni Ioannidou
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
International Journal of Approximate Reasoning, 51, 2010,
pp. 993–1009.
M. Kohli and T. V. Prasad, Identifying Dyslexic Students by
Using Artificial Neural Networks, Proceedings of the World
Congress on Engineering, London, U.K., 1, 2010.
R. Abbott, Automated expert modeling for automated
student evaluation. In Intelligent tutoring systems, eighth
international conference, ITS 2006, Jhongli, Taiwan, June
26–30, 2006, proceedings, pp. 1–10. Lecture Notes in Computer Science 4053, Springer.
E. Melis, E. Andres, J. Budenbender, A. Frischauf, G.
Goguadze, P. Libbrecht, M. Pollet and C. Ullrich, ACTIVEMATH, A Generic and Adaptive Web-Based Learning
Environment, International Journal of Artificial Intelligence,
24(4), 2001, pp. 1–25.
E. Melis and J. Siekmann, ActiveMath: An intelligent
tutoring system for mathematics. In L. Rutkowski, J. Siekmann, R. Tadeusiewicz, & L. A. Zadeh (Eds.), Seventh
international conference artificial intelligence and soft computing (ICAISC), pp. 91–101. Springer-Verlag, 2004.
P. Libbrecht and E. Melis, Methods to Access and Retrieve
Mathematical Content in ActiveMath. Lecture Notes in
Computer Science, 2006, 4151, pp. 331–342.
N. Livne, L. Nava, O. Livne and C. Wight, Can automated
scoring surpass hand grading of students’ constructed
responses and error patterns in mathematics? MERLOT
Journal of Online Learning and Teaching, 3(3), 2007.
L. Anthony, J. Yang and K. R. Koedinger, Toward NextGeneration Intelligent Tutors: Adding Natural Handwriting
Input, IEEE Multimedia, 15(3), 2008, pp. 64–68.
C. S. Gonzalez, D. Guerra, H. Sanabria, L. Moreno, M. A.
Noda and A. Bruno, Automatic system for the detection and
analysis of errors to support the personalized feedback.
Expert Systems with Applications, 37, 2010, pp. 140–148.
World Health Organisation, International Classification of
Diseases, Tenth Revision (ICD-10). Chapter V: Mental and
Behavioural Disorders. Geneva: World Health Organisation, 1993.
American Psychiatric Association, Diagnostic and Statistical
Manual of Mental Disorders (DSM 4th edn), Washington,
DC: APA, 1994.
G. Rebolledo-Mendez and S. De Freitas, Attention modeling using inputs from a Brain Computer Interface and usergenerated data in Second Life. In: The Tenth International
Conference on Multimodal Interfaces (ICMI 2008), Crete,
Greece, 2008.
G. Rebolledo-Mendez, I. Dunwell, E. A. Martı́nez-Mirón,
M. D. Vargas-Cerdán, S. de Freitas, F. Liarokapis, A. R.
Garcı́a-Gaona, Assessing Neurosky’s Usability to Detect
Attention Levels in an Assessment Exercise, Human Computer Interaction, 5610(1), 2009, pp. 149–158.
R. M. Aguilar, V. Munoz, M. Noda, A. Bruno, L. Moreno,
Teacher Strategies Simulation by Using Fuzzy Systems,
Computer Applications in Engineering Education, 18(1),
2010, pp. 183–192.
J. Anuradha, Tisha, V, Ramachandran, K. V. Arulalan, B.
K. Tripathy, Diagnosis of ADHD using SVM algorithm, In
Proceedings of the Third Annual ACM Bangalore Conference, 2010.
M. Delavarian, F. Towhidkhah, P. Dibajnia, S. Gharibzadeh, Designing a decision support system for distinguishing
ADHD from similar children behavioral disorders. Journal
of Medical Systems, 2010 [Epub ahead of print].
Athanasios Drigas is a Senior Researcher at N.C.S.R. Demokritos. He is the Coordinator of Telecoms Lab and founder of
Net Media Lab since 1996. From 1985 to 1999 he was the Operational manager of the Greek Academic network. He has
been the Coordinator of Several International Projects, in the fields of ICTs, and e-services (e-learning, e-psychology,
e-government, e-inclusion, e-culture etc). He has published more than 200 articles, 7 books, 25 educational CD-ROMs and
several patents. He has been a member of several International committees for the design and coordination of Network and
ICT activities and of international conferences and journals.
Rodi-Eleni Ioannidou is a special education teacher. She has participated in various research projects regarding the use of
Information and Communication Technologies (ICTs) in special education.