Model for an Adaptive Tutoring System Eugenia Kovatcheva University of Library Studies and Information Technology Tzarigradsko Shosse 119, Sofia 1784, Bulgaria +359 882 255 075 ekovatcheva@gmail.com ABSTRACT  An intelligent tutoring system is educational software Nowadays the interest to the adaptive intelligent eLearning containing an artificial intelligence component. systems increases. There are different kind of adaptations one to  ITSs are computer-based learning systems which attempt to the content, other to the learning process or to the assessment and adapt to the needs of learners.’ so on. The crucial moment for the learner motivation’s is to catch their needs and possibilities and then to act, i.e. the respond from The essence of the ITS is the humanization of the learning process the system.. The intelligent tutor (system agent(s)) has to decide [1]. The modelling and constructing of the ITSs is the hot topic in the most appropriate path through the content based on the nowadays research. Computer Based Instruction and Training via collected information for learner as: learning style, learner track e-learning platforms are the predecessors of ITSs. The ITSs has to through the topics, learner grades and offer further steps. The build a complex model of the educational process, to adapt it and intelligent agent – tutor keeps all data for every single learner, to control the interactions. Usually ITSs provide individualized analyse them and offers next learner’s actions on the system. tutoring using four models for knowledge of: the domain, This paper presents a constructive model based on the learning learners, teacher strategies and user interface. Creating of the ITSs style of the learners and their ability and how it could be is based on different artificial intelligence algorithms and implement in an intelligent eLeaning system. It should be used for computational architectures as: Bayesian Networks [2], Markov self-study in formal and informal education as well as for Models [3], Neural Networks, Higher-Order Semantic Spaces [4, representing the digitalized cultural and historical heritage for 5], Fuzzy Control Systems [6, 7], Production Rules Systems, educational purposes. Generative Grammars [8, 9], Non-Linear Dynamical Systems, External Representation [10], Concept Maps [11], Path Guidance Categories and Subject Descriptors [12], Agents Based [13, 14], Ontology Based [15, 16, 17, 18], Data Mining [19, 20, 21, 22], and etc. H.1.2 [Models and Principles]: User/Machine Systems – Human factors, Human information processing 2. CHANGE OF THE WAY OF LEARNING General Terms The school is no longer the sole and the most attractive source of information and knowledge. Quick access to unlimited sources of Design, Human Factors information is widely available due to modern technologies. The traditional concept of literacy has been gradually extended to a Keywords multimedia literacy referring to students' abilities to read, write, Intelligent tutoring system, e-learning, constructive model and communicate with digitally encoded materials - text, graphics, still and moving images, animation, sounds. 1. INTRODUCTION: THE EVOLUTION The way the people learn is changed as well. The existence of OF E-LEARNING SYSTEMS IN SHORT non-formal learning that is not provided by an education or Last 30 years are time for creation of the Information Society. training institution has been widely recognised. This type of The education is on new stage implementing Information and learning does not typically lead to an official certification. Communication Technologies in it. The human ambitions for Having in mind the characteristics of non-formal learning, as well resemblance the web-based educational systems to the face-to- as the requirements of the ICT driven educational reform, a model face education is the goal of this paper. There are different of Adaptive Tutoring System has been developed. The definitions of the Intelligent Tutoring System (ITS) some of them personalized support of learners has been identified as among the  ITSs are computer software systems that seek to mimic the most important functions of the system since the learning takes methods and dialog of natural human tutors, to generate place in an open and dynamic learning environment. instructional interactions in real time and on demand, as Nevertheless of changed way of people learning the learning style required by individual students.’ is different for the people. Last few years the design of learning systems is the very BCI’12, September 16–20, 2012, Novi Sad, Serbia. important part of developed web technologies. Copyright © 2012 by the paper’s authors. Copying permitted only for private and academic purposes. This volume is published and copyrighted by its editors. To find the most appropriate model for the learners and to make Local Proceedings also appeared in ISBN 978-86-7031-200-5, Faculty of Sciences, University of Novi Sad. them closer to them arose and the adaptive learning systems. They 16 use knowledge representation and domain models, consider the The ITS model consists five elements set for the student knowledge as a means for providing adaptivity [23]. The identification of learner where: overview identifies five main features used for maintaining adaptivity:  L – Learning style,  student goals,  E – an Evaluation vector –for learner assessment during the educational process,  student knowledge and familiarity with the domain,  T – learner Time spend on the system,  student qualification (how quickly he or she acquires knowledge),  A – the learner interaction with the system – type of Actions,  experience in the hyperspace and  P – control Points for learning process.  personal preferences. The Intelligent Tutoring System consists of few modules: Admin, Expert, Domain, Learner, Assessment, Learning Style and 3. Model Learner Interface as it is shown on figure 1. In the previous work [24, 25] the author went deep in student knowledge- the most important student characteristic in the The Expert module includes pedagogical perspectives and lessons majority of the current adaptive systems. planner – connected to the Domain module (the concept map approach in combination with ontology is used in domain The presented constructive research model deals with the learning module). The Learning Style module is based on the Homey and style, student knowledge, the domain of knowledge, the learner Mumford’s model interaction with the learning system (time on the ITS and type of activities) and the intelligent tutor. The ITS is adapted to the The learner bahviour on the system is described with these five learner style, the achieved learning results- learner ability and the elements where the learning style L could be identifying on each way of access to the learning materials. The ITS introduces and student interaction with the system. The time T spent on one responds the appropriate materials to the learner - a thorough action A and evaluation E on the control point return the style L. answer and kind of material. The intelligent tutor (system On the other hand the system keeps information for the learner agent(s)) has to makes decision on each learner action and the way through the learning materials. collected learner’s data. Assessment / Feedback Admin Learning Style Learner Interface Expert Module Learner  student goals,  student knowledge  student qualification  experience in the hyperspace  personal preferences. Domain Module Learner Module Figure 1: intelligent tutoring system. 17 The definition of set describes the learner from Honey and Mumford have been developed tests for identifying different point of view. It gives an approximation to the intelligent the learning styles. The learning is more effective when the tutoring system for further learner actions on the system. It will learners use appropriate (for their learning style) resources and improve the possibility for knowing the learner. This user profile actions and they can apply what they know in the efficient way. constructive model is an attempt at comprise all learner The implementation of the possibilities for style recognition into characteristics as knowledge and as human factor. the tutoring systems will support the learners and give the flexibility to the learners. The visualization of the module is like a sphere. The expert module, learner module, domain, learning style and learning interface are angels of one pentagram - the circumference from 4.2 Computer Adaptive Tests each angle answers for one of the modules. The Assessment Computer Adaptive Tests (CAT) base on the Item Respond module observes the learner behavior on the systems supported by Theory for identification of student ability under the knowledge the expert. The modeled intelligent system assures appropriate domain in each step of the process. CAT successively selects and ease of use learning materials and environment for each questions so as to maximize the precision of the exam based on learner. It is prerequisite for effective learning. what is known about the examinee from previous questions. From the examinee's perspective, the difficulty of the exam seems to The technology pawns the repository structured by topics, tailor itself to his or her level of ability. learning styles and difficulties as well as the assessment pool. The main problem could be in front of content providers to develop The psychometric technology that allows equitable scores to be appropriate learning material to fill in. computed across different sets of items is item response theory (IRT). IRT is also the preferred methodology for selecting Knowing of students’ ability and learning style is a premise for optimal items which are typically selected on the basis of students’ engagement and involvement in mastering of knowledge information rather than difficulty, per se. and skills. This type of systems is very suitable for self-learning and distance education. The learner could be monitored at each The learning system control points are the tests for self-evaluation moment by set . not only for the different topics and with additional items describing the learner ability and motivation. One possibility for these tests is to be constructed the Computer Adaptive Test (CAT) 4. TOOL SUPPORT FOR based on Item Response Theory, [27]. The constructive research identify that the results from the CAT better describe the user 4.1 Learning Style comparing the fixed tests. The use of CAT presumes the well Learning styles are various approaches or ways of learning. They structured metadata for describing the test items and the tool for involve educating methods, particular to an individual, that are developing and pool for store the items according Thissen, & presumed to allow that individual to learn best. Most people Mislevy [28] prefer an identifiable method of interacting with, taking in, and processing stimuli or information. Based on this concept, the idea The basic computer-adaptive testing method is an iterative of individualized "learning styles" originated in the 1970s, and algorithm with the following steps: acquired "enormous popularity". 1. The pool of available items is searched for the optimal Proponents say that teachers should assess the learning styles of item, based on the current estimate of the examinee's their students and adapt their classroom methods to best fit each ability student's learning style, which is called the 'meshing hypothesis. 2. The chosen item is presented to the examinee, who then The basis and efficacy for these proposals are extensively answers it correctly or incorrectly criticized. Although children and adults express personal preferences, there is no evidence that identifying a student's 3. The ability estimate is updated, based upon all prior learning style produces better outcomes, and there is significant answers evidence that the widespread "meshing hypothesis" (that a student 4. Steps 1–3 are repeated until a termination criterion is will learn best if taught in a method deemed appropriate for the met student's learning style) is invalid. Allegedly well-designed studies "flatly contradict the popular meshing hypothesis". One of Nothing is known about the examinee prior to the administration it is of Honey and Mumford [26]. They distinguished four of the first item, so the algorithm is generally started by selecting learning styles. It is one qualification and never appears in pure an item of medium, or medium-easy, difficulty as the first item. form: As a result of adaptive administration, different examinees receive  Reflector - prefers to learn from activities that allow quite different tests [29]. them to watch, think, and review what has happened.  Theorist - prefers to think problems through in a step- 4.3 Intelligent Agents by-step manner. Intelligent Agents serve the Learner and Tutor module. They  Pragmatist - prefers to apply new learning to actual  keep tracks of the learner practice to see if they work.  interaction user between user and system  Activist - prefers the challenges of new experiences,  tutor’s emotions return the feedback involvement with others, assimilations and role-playing. 18 4.4 Concept Map [6] Lombardi, M.M. (2007) Authentic Learning for the 21st A concept map is a way of representing relationships between Century: An Overview, ELI Paper 1: 2007 May 2007, ideas, images, or words in the same way that a sentence diagram EDUCAUSE represents the grammar of a sentence, a road map represents the [7] Suraweera, P., Mitrovic A.and Martin, B. A Knowledge locations of highways and towns, and a circuit diagram represents Acquisition System for Constraint-based Intelligent Tutoring the workings of an electrical appliance. In a concept map, each Systems word or phrase is connected to another and linked back to the [8] Contreras, W.F., Galindo, E.G., Caballero E.M., and original idea, word or phrase. Concept maps are a way to develop Caballero, G.M. (2006) An Intelligent Tutoring System for a logical thinking and study skills by revealing connections and Virtual E-learning Center, Current Developments in helping students see how individual ideas form a larger whole. Technology-Assisted Education, © FORMATEX 2006 Concept Maps ensure the categorization into the domain module. [9] Salgueiro, F., Costa, G., Cataldi, Z., Lage, F., Martínez, They in combination with Artificial Neural Networks are in used R.G. (2005) Redefinition of Basic Modules of an Intelligent for knowledge representation and organization and Knowledge Tutoring System: The Tutor Module, Proceedings VII map as self-training and adjustable tool. Workshop Argentine Computer Science Reaserchers. pp. 444–448, 2005 4.5 Bayesian Network [10] Williams, B., The Role of External Representations in Bayesian network [2] or belief network model or directed acyclic Intelligent System authoring: Supporting localised decision graphical model is a probabilistic graphical model (a type of complex and evolving global context, AI-ED 2001 statistical model) that represents a set of random variables and Workshop, External Representations in AIED: Multiple their conditional dependencies via a directed acyclic graph. the Forms and Multiple Roles, San Antonio, Texas, Sunday 20th Bayesian networks could be support the analysis of the data for May 2001 huge about of learner and return visualized feedback to the tutor [11] Martins,W, Cortez, J.P., Nalini, L.E.G. Gomes, V.M. The and administration. Use of Conceptual Maps in a Hybrid Intelligent Tutoring System 5. CONCLUSION [12] Hong, C-M., Chen, C-M ,Chang, M-H, and Chen, S-C The presented constructive research model deals with the learning (2007) Intelligent Web-based Tutoring System with style, student knowledge, the domain of knowledge, the learner Personalized Learning Path Guidance, Seventh IEEE interaction with the learning system (time on the ITS and type of International Conference on Advanced Learning activities) and the intelligent tutor. The ITS is adapted to the Technologies (ICALT 2007) learner style, the achieved learning results- learner ability and the way of access to the learning materials. The ITS introduces and [13] Hospers, M., Kroezen, E. Nijholt, A. den Akker R.& Heylen, responds the appropriate materials to the learner - a thorough D. (2003) Developing a Generic Agent-based Intelligent answer and kind of material. The intelligent tutor (system Tutoring System, Proceedings of the The 3rd IEEE agent(s)) has to makes decision on each learner action and the International Conference on Advanced Learning collected learner’s data. Technologies (ICALT’03) [14] Gulz, A.and Haake, M. (2006) Virtual pedagogical agents – 6. ACKNOWLEDGEMENTS design guidelines regarding visual appearance and The researcher’s participation in the conference was financially pedagogical roles, Current Developments in Technology- supported by the Research Foundation at Sofia University “St. K. Assisted Education, © FORMATEX 2006 Ohridksi,” project “Integral University Center for E-learning” – [15] Day, M.-Y., Lu, C-H., Yang, J.-T.D. Chiou, G.-F., Ong, C.- INZ01/0111. S., Hsu, W.-L., Designing an Ontology-based Intelligent Tutoring Agent with Instant Messaging 7. REFERENCES [16] Lu, C.-H., Wu, S.-H., Tu L., Hsu, W.-L. (2004) The Design of An Intelligent Tutoring System Based on the Ontology of [1] Samuelis L,Notes on the Components for Intelligent Tutoring Procedural Knowledge, Proceedings of the IEEE Systems, Acta Polytechnica Hungarica Vol. 4, No. 2, 2007 International Conference on Advanced Learning [2] Butz,C.J. Hua, S., Maguire, R.B. (2004) A Web-based Technologies (ICALT'04), 2004, Volume 00, pp: 525–529 Bayesian Intelligent Tutoring System for Computer [17] Oguejiofor, E., Kicinger, R., Popovici, E., Arciszewski, T. Programming, IEEE Xplore, 20–24 Sept. 2004, pp 159–165 and De Jong, K. (2004) Intellignet tutoring systems: an [3] Ueno, M. & Okamoto T, Online MDL-Markov analysis of a ontology-based approach, International Journal of IT in dischussion process in CSCL, 6th Int. Conference in Architecture, Engineering and Construction, Vol 2, Issue 2/ Advanced Learning Technologies (ICALT 06) May 2004, Millpress [4] Hansen, T.K. (2006) Computer Assisted Pronunciation [18] Passier, H. (2004) Ontology based feedback generation in Training: The four 'K's of feedback, Current Developments design-oriented eLearning systems, IADIS e-Society 2004 in Technology-Assisted Education, © FORMATEX 2006 Conference [5] Schwitter, R. and Islamm Md T (2003), S-Tutor: A Speech- [19] Nkambou, R. (2006) Towards Affective Intelligent Tutoring based Tutoring System, 11th Int. Conference of Artificial System, G. Rebolledo-Mendez, E. Martinez-Miron (Eds): Intelligence in Education, AIED 2003 Workshop on Motivational and Affective Issues in ITS. 8th International Conference on ITS 2006, pp 5–12, 2006. 19 [20] Romero C.and Ventura, S. (2007) Educational data mining: Tutoring System In proceedings of ICERI2010 Conference, A survey from 1995 to 2005, Expert Systems with 15–17 November 2010, Madrid, Spain, pp 004204–004210, Applications Volume 33, Issue 1, July 2007, pp 135–146 ISBN: 978-84-614-2439-9 [21] Romero, C., Ventura S. (2006) editors, Data Mining in E- [26] Honey, P. and Mumford, A. (2000) The Learning Styles Learning, WITpress Southampton, Boston, 2006 Questionnaire. Maidenhead: Peter Honey Publications [22] Attia, S.S., Mahdi, H.M.K., Mohammad, H.K. (2004) Data [27] Weiss, D. J. (2004), Computerized Adaptive Testing for Mining in Intelligent Tutoring Systems Using Rough Sets, Effective and Efficient Measurement in Counseling and 2004 IEEE Education, Measurement and Evaluation in Counseling and [23] Brusilovsky, P. Kobsa, A. and Vassileva, J. (1998), Adaptive Development, July 2004, Volume 37 Educational Hypermedia. Kluwer Academic Publishers. [28] Thissen, D., & Mislevy, R.J. (2000). Testing Algorithms. In [24] Kovatcheva E., Okamoto T. (2008) The Framework and Wainer, H. (Ed.) Computerized Adaptive Testing: A Primer. Prospective Design for Web-Based Intelligent Tutoring Mahwah, NJ: Lawrence Erlbaum Associates. System, Proceedings of the Sixth IASTED International [29] Green, B.F. (2000). System design and operation. In Wainer, Conference on Web-based Education, Innsbruck, Austria, H. (Ed.) Computerized Adaptive Testing: A Primer. March 2008 Mahwah, NJ: Lawrence Erlbaum Associates [25] Kovatcheva E., Nikolov R., Okamoto T (2010) The User Profile Constructive Model for a Web-Based Intelligent 20