A Systematic Review of Domains , Techniques , Delivery Modes and Validation Methods for Intelligent Tutoring Systems

An Intelligent Tutoring System (ITS) is a computer software that help students in learning educational or academics concepts in customized environment. ITSs are instructional systems that have capability to facilitate user by providing instantaneous feedback and instructions without any human intervention. The advancement of new technologies has integrated computer based learning with artificial intelligence methods with aim to develop better custom-made education systems that referred as ITS. One of the important factors that affect students learning process is self-learning; all students cannot have similar experience of learning scholastic concepts from same educational material. Because students have individual differences that make some topics difficult or easy to understand regarding taken subjects. These systems have capability to improve teaching and learning process in different educational domains while respecting individual learning needs. In this study an attempt is made to review the research in field of ITSs and highlight the educational areas or domains in which ITSs have been introduced. Techniques, delivering modes and evaluation methodologies that have been used in developed ITSs have also been discussed in this work. This work will be helpful for both academia and new comers in the field of ITSs to further strengthen basis of tutoring systems in educational domains. Keywords—ITS; intelligent tutoring system; intelligent learning; adaptive learning; intelligent tutoring; ITS review


I. INTRODUCTION
Information and communication technologies (ICT) have become necessary part of educational system in order to replace traditional teaching system with modern teaching system [1].Modern information resources like hypermedia, multimedia, internet and intranet are contributed together to provide advanced learning pedagogies [2][3][4].The new educational model is learner centered model in which flexibility is provided to learner to learn according to their ability and requirements in simple words this new model is oriented toward learner or student [5,6].To increase efficiency of these learning model factors like adaptability, interactivity, intelligence and dynamically generated web contents are added to these models that leads them towards development of intelligent tutoring system (ITS).
Many factors have been highlighted in previous studies that reflect importance of ITSs in educational domain, such as time limitations and individual meeting with students make it difficult for teacher for provide feedback to individual student in large class [7].In private tutoring system, usually tutor provide help to student as much as he/she can but providing as many tutors to facilitate number of students is not feasible option from economic point of view [8].ITSs are a solution that can solve these kind of problems in education.The most important feature of ITSs is their ability to customize instructional strategies and activities according to learner's requirements and characteristics [9,10].Many researchers contribute their efforts for refining student's ability to solve different problems in education [11][12][13][14].The first ITS named SCHOLAR tutor was introduced in 1970 [15].This system was developed to review student's knowledge about South America geography.SCHOLAR tutor began two way interaction with students by using facts of knowledge and semantic network of concepts to appraise student's knowledge in geography [16].
The traditional ITSs consist of four main modules i.e. expert module, student or learner module, pedagogical or tutor module and user interface module [17].The expert module deals with domain knowledge that learner or student wants to learn [18].The methods for problem solving and evaluating students activities in learning process are also considered in this module [19].The student module completely depends on student attributes such as learning style, activities, behavior and knowledge deficiency.All relevant information about student is gathered and updated in system to refine learning process [20].Additional information such as past learning experiences and learning preferences may also be stored in system to provide adaptability in teaching process [10].
The pedagogical module identifies students deficiencies in specific topic or subject based on learner module and focus on the strategies to provide best material that overcome the deficiencies of student in specific area or subject or topic [21].The user interface module is communication part of the ITS it provides interaction facility between system and user [22].The aim of present study is to review the developed ITSs across educational fields and gather comprehensive information about their development techniques, purpose, delivery modes and evaluation methods.This paper is organized as following; in section II methodology of review is presented, results are presented in section III.The answers of posed questions are discussed in sub sections of section III and work is concluded in section IV. www.ijacsa.thesai.orgII.METHODOLOGY One of the most common ways to evaluate and understand all available research literature related to specific research problem or question is systematic literature review.In literature, numbers of methodologies are available for systematic review.The methodology adopted for this systematic literature review was based on guidelines presented in [23].The review process consists of three phases which contain ten sub activities.The details of activities conducted in these phases are shown in Fig. 1    The questions were formulated in order to complete first sub activity of phase 1.In second sub activity of phase 1 a review protocol was developed.The review protocol includes considered time span of published papers, publication sources and keyword used to find required data.The research focus on the articles published in last five years and repositories used to find out required data include IEEE xplorer, science direct, ACM, Google scholar, springer and web of science.The final review protocol is given in Table I.The developed protocol is reviewed and validated after making some necessarily changes by researchers.
The execution of search chain on selected sources provides us a set of 82 articles.These articles were screened on basis of their title and abstract.After initial review according to inclusion criteria these 82 articles were filtered and only 39 articles were left that meet our inclusion criteria.For strengthen the search related to relevant articles, backward snowballing guidelines presented in [24] was also adopted to find out cited papers in selected articles that meet our inclusion criteria; this was done to make sure that no articles related to our research questions were left to considered.
In second phase of review a literature search was conducted in pre-defined databases by using keywords mention in Table I.These keywords were used in combination and alone for initial collection of research material related to our specified research questions.The inclusion criteria of this study is strictly based on posed research questions.Only those articles that contain material related to our research question were considered in this review.

III. RESULTS
The results of review are presented in this section.A year wise representation of results is given in Table II.These results are characterized with research questions posed earlier in this study.The variables of selected studies are presented in Table III.

A. What are the Subject/Domains for which ITSs have been
Designed?
The result of review Table IV shows educational domains for which ITS have been designed.The frequency of each educational field in selected papers is demonstrated in Fig. 2.These domains were divided into six basic categories include; computer science, health/ medical science, mathematics, physics, language and others.Computer science was leading field in which ITS have been widely used to provide help in learning process.In computer science, programming was major area in which these systems were used.The frequency of systems that were designed for programming purposes in computer science domain was 61.11 %.Majority of the users of these systems were university level students.The second major field in which ITSs were used was mathematics, majority of ITSs (66.66%) were designed for school level students in this field to provide help in basic mathematics.The medical science field is at third rank, in which most of the system provides theoretical help to the students in understanding concepts about different topics, diseases and their management.In physics these systems were designed to provide theoretical concepts about different topics with focus on electronics.In languages domain all found ITSs were designed to teach English language or grammar.

B. What Type of Techniques or Tools have been used in Development of ITSs?
The review results of question 2 are presented in Table V and frequency of techniques used in development of tutoring systems is showed in Fig. 3.The techniques used for designing ITSs are categorized into eight different categories that are listed in Table V. Rule based reasoning technique was leading approach in designed ITSs.Rule based techniques increase trace of system reasoning for specific case with background familiarity [60].This method was used in 28.21% studies to design ITSs, in computer science domain 66.66% designed ITSs were based on rule based reasoning out of which most systems were designed for programming purpose.In mathematics domain 50% ITSs used rule based reasoning method for system designing.It seems that rule based method could be a suitable choice for decision making in computer programming fields and mathematics fields to deal with structure, patterns and numbers.
Bayesian technique was second widely used technique is designing of ITSs.These techniques based on probabilistic association between set of variables [61].Bayesian method was used in 15.38% studies and majority of these studies were related to computer science, mathematics and physics domains.Bayesian technique is good method to deal with problems that deals with uncertainties [62].In 12.28% studies intelligent agent systems were used in designed ITSs, intelligent agent system used to solve problems that are difficult to solve by an individual [63].These systems were used for skill improvement, multi domain learning and problem solving in computer programming.In computing field intelligent agent techniques can be used to support its users in process of communication with system [64].Frequency of data mining and Artificial Neural Network (ANN) techniques was almost similar.

Case based reasoning technique was least used technique in developed
ITSs.This technique is usually used to solve new problems by finding most similar problem and updating the existing case according to found model [65].However this technique was only used in two works that were related to computer programming.Case based reasoning can be beneficial in medical domain for design of tutoring systems [66]

C. What Type of Delivery Modes has been used by Developed ITSs?
The results of question 3 are presented in Table VI and frequency of ITSs delivery modes is shown in Fig. 4.These results show that web based mode was leading mode in delivering by tutoring systems.51% of selected papers were relied on web based systems for delivering educational material to learner.In computer science domain 55.55% intelligent tutoring systems were web based.The frequency of web based tutoring systems in mathematics and medical domains were 50% and 45%, respectively.
Computer based application software for intelligent tutoring was used in 30.77% studies.In mathematic domain 50% developed ITSs were computer based application software while frequency of application software in computer science and medical domain were 33.33% and 40%, respectively.In 10.26% studies researchers provides a conceptual or theoretical framework for tutoring systems.While model was proposed in only one study that was related to physics domain.These results reveal the fact that web based and computer application based modes are most popular infrastructure for the development of tutoring systems.In our selected articles we do not found any ITSs which can provides mobile based content delivery to facilitate learner.However, mobile devices are considered as emerging technologies and its use in society is part of daily routine [67].These devices can facilitate the implementation of tutoring systems independent of location and time [58].

D. How the Techniques have been Validated?
The results of question 4 are presented in Table VII and percentile results of question 4 are shown in Fig. 5.The method that was used widely for validation of the developed ITSs was experiment (59%).In experimental validation different type of techniques were used out of which 74% experiments were performed on basis of user or learner performance in which pretest and posttest were conducted to evaluate the system.17% experiments were based on comparative analysis between groups of ITSs users and nonusers, while remaining experimental approaches were based on learner knowledge and skills for system validation.Experimental methods were widely used in the field of computer science and medical.The frequency of these methods in computer science and medical were 61.11% and 60%, respectively.In validation of any ITS, learner plays an important role, learner experience of using system or feedback is one of the most common method that could address the problems related to usability of specific system [68].In this review we found only 15.38% studies in which systems were evaluated on basis of user feedback.While 10.25% studies use both experimental and feedback methods for system validation.The remaining validation method includes prototype testing, simulation and observation.Only one study in medical science domain was found in which not any kind of validation method was adopted for system evaluation.Intelligent Tutoring Systems have capability to replace traditional teaching method with modern adaptive teaching method in which every student get knowledge according to his/her requirement.The main feature of these systems is their ability to customize instructional strategies and activities according to learner's requirements and characteristics.In this study ITSs educational domains, purpose of ITSs in specific domain, technique for ITSs development, delivery modes of ITSs and methods of validating designed ITSs were reviewed.Computer science was a major area in which these systems were used for teaching programming subjects.Rule based and Bayesian techniques were most frequent used techniques in design of ITSs and most of the developed systems provide web based services.In addition, learner performance was major method for validation of designed system.User feedback plays important role in evaluation of any system but we found that user feedback was not considered at high level in these studies for validation of ITSs.As use of mobile devices increase, these devices can facilitate learner in personalized learning in more customized way.We also observe in this study that none of system focus on mobile based tutoring, so on the basis of this study we recommended development of mobile based ITSs.
and discussed under.In first phase of review the following questions were posed: Q1: What are the subjects or domains for which ITSs have been designed?Q2: What types of techniques or tools have been used in development of ITSs?Q3: What type of delivery modes have been used by developed ITSs?Q4: How the techniques have been validated?

TABLE II .
YEAR WISE SEARCH RESULTS

TABLE III .
VARIABLES OF SELECTED PAPERS

TABLE VI .
CATEGORY WISE RESULTS OFQUESTION 3

TABLE VII .
CATEGORY WISE RESULT OFQUESTION 4