CCIS 111 - A Module for Adaptive Course ... - Springer Link

A Module for Adaptive Course Configuration and Assessment in Moodle Carla Limongelli1 , Filippo Sciarrone2 , Marco Temperini3 , and Giulia Vaste1 1

Department of Computer Science and Automation, Roma Tre University Via della Vasca Navale, 79 00146 Roma, Italy {limongel,vaste}@dia.uniroma3.it 2 Open Informatica s.r.l., E-learning Division Via dei Castelli Romani, 12A - 00040 Pomezia, Italy [email protected] 3 Department of Computer and System Sciences, Sapienza University Via Ariosto, 25 00184 Roma, Italy [email protected]

Abstract. Personalization and Adaptation are among the main challenges in the field of e-learning, where currently just few Learning Management Systems, mostly experimental ones, support such features. In this work we present an architecture that allows moodle to interact with the Lecomps system, an adaptive learning system developed earlier by our research group, that has been working in a stand-alone modality so far. In particular, the Lecomps responsibilities are circumscribed to the sole production of personalized learning objects sequences and to the management of the student model, leaving to moodle all the rest of the activities for course delivery. The Lecomps system supports the “dynamic” adaptation of learning objects sequences, basing on the student model, i.e., learner’s Cognitive State and Learning Style. Basically, this work integrates two main Lecomps tasks into moodle, to be directly managed by it: Authentication and Quizzes. All in all, and so far, the advantage of the presented integration is in the real possibility to deliver and take personalized courses, residing and basically remaining into the moodle environment.

1

Introduction

Personalization and Adaptation are among the main challenges in the field of e-learning. However, just few Learning Management Systems (LMSs) support such features, mostly as experimental ones. As a matter of facts, the integration of personalization aspects into state-of-the-art and widely used LMSs is a complex task and it is taken into consideration from the scientific community. For instance, the Grapple Project1 is a three years European project involving 14 partners from 9 European countries, and aims at delivering to learners a technology-enhanced learning environment that guides them through a life-long 1

http://www.grapple-project.org/

M.D. Lytras et al. (Eds.): WSKS 2010, Part I, CCIS 111, pp. 267–276, 2010. c Springer-Verlag Berlin Heidelberg 2010

268

C. Limongelli et al.

learning experience, automatically adapting to their personal preferences. In [10] an extended discussion is presented about personalization and adaptation in elearning and on the development of suitable tools for the authoring of learning material and its use in adaptive courses. Two main aspects of the work in this area of adaptive systems are related to the ways the learning material (usually embedded and specified into learning objects) is automatically sequenced in a course, and the ways teachers and content developers can define such material and express learning (or teaching) strategies to sequence and deliver it. In several systems, such as AHA! [3] and ELM-ART [9] the learning objects sequence is actually produced and delivered step by step, for example through the link-annotation technique [2]; in other systems the sequence is completely set initially, and then maintained, modified and possibly reproduced on occurrence [1,7]. The Lecomps system, presented in [8], follows an approach of the latter type. Here we cope with the complexities of the integration of personalization aspects into the moodle LMS. We present an architecture that allows moodle to interact with an adaptive system, the Lecomps system, that we developed earlier, and that has been working stand-alone so far. The Lecomps system supports the “static” configuration and the “dynamic” adaptation of sequences of learning objects, basing on a student model constituted by learner’s Cognitive State (CS) and Learning Style (LS). Basically, the work we present here is the integration of the following two Lecomps tasks into the moodle LMS. The first task is the Authentication management: the learner is a “moodle student” and interacts mainly with it. The second task is Quizzes management, that, thanks to Lecomps, are personalized basing on the learning objects taken by the learner: in the Lecomps environment they are first designed and then imported-in and delivered-through moodle; then the results of quizzes are sent back to Lecomps, to allow for student model updating. Lecomps updates the student model by updating its CS an LS [8]. In particular the CS is updated by either adding new knowledge (“certified” by right answers to quizzes), or increasing/decreasing the degree of certainty for knowledge that were already in the CS, or canceling knowledge from CS (when it drops below a stated degree of certainty). The delivery of a course to the particular learner it was configured for, is performed completely into the moodle environment. Conversely, the thorough import of the questions defined in Lecomps into the moodle databases, allows for the delivery of quizzes through moodle, also allowing teacher and students to access a complete history of the quizzes taken and maintained to support item and performance analysis. The advantage of the presented integration is in the real possibility to deliver and take personalized courses by the moodle environment only. This integration brings two main advantages: moodle is enriched with personalization, allowing for a different sequencing of learning materials for different students; the Lecomps personalization process could be enriched using both the moodle quiz management and logging, characteristics that make moodle one of the most used LMS. Our work is still to enhance: Lecomps is still needed for certain activities that should be better managed in an LMS such as moodle such as authoring and management of learning objects repositories.

A Module for Adaptive Course Configuration and Assessment in Moodle

269

However, the final aim in this line of work is to circumscribe the Lecomps responsibilities to the sole production of personalized sequences of learning objects together with the management of the student model. In Section 2 we describe the Lecomps system. In Section 3 the integrated architecture of the resulting system is presented with some example of quizzes management. Finally, in Section 4 conclusions are drawn.

2

The Lecomps Learning Management System

Lecomps is a web-based e-learning environment, in which learners and teachers get support to the following activities: – authoring of learning material, i.e. learning component, as described in the following definition 2 – creation of learning environments (the framework in which students enroll to get personalized courses about a stated subject matter, and teachers state the general aims of such courses, i.e. the target knowledge common to all students) – enrollment of learners and monitoring (from both teachers and learners) of the learning activities and results – automated construction and adaptive delivery of the personalized courses, tailored over learning goals, student’s knowledge and learning styles. A personalized course is a sequence of learning objects, picked up from a repository. The sequence is defined specifically over the present state of the student model. For experimental reasons the system is available in different versions, each one basing on a particular engine, responsible for learning object selection and sequencing; the first engine issued (which is also the one running into the version of Lecomps referred to by this article) used a graph traversing algorithm, specifically designed; a second version [6] was redesigned to exploit a planning-based approach; a third version is being studied, in parallel, using a constraint-logicprogramming-approach [8]. The concept of learning object is implemented in Lecomps by the structure of Learning Component (LC). The representation of knowledge, occurring in the definition and management of LCs and in other areas of the system, is expressed by knowledge items (kis), which are basically nouns for concepts. To manage Learning Styles we adopt the Felder and Silverman’s model [4,5], which allows us to qualify a learner by a combination of two possible values for each one of four dimensions. Definition 1. Student Model The Learner’s Cognitive State (CS) is the set of pairs < ki, certainty >, listing the kis presently owned by the learner, each one qualified by its level of certainty (as determined by the system basing on learners answers to tests). The Learner’s Learning Style (LS) is represented by a 4-tuple of couples, < d1 , v1 ; d2 , v2 ; d3 , v3 ; d4 , v4 >, where each di is one of the basic values in a

270


dimension of the Felder-Silverman model (d1 = active/reflective, d2 = sensing/intuitive, d3 = visual/verbal, d4 = sequential/global) and the vi ∈ [0, 11]. Definition 2. Learning Component A Learning Component (LC) comprises: Required Knowledge. A set of kis denoting the knowledge that is supposed to be needed in order to take the learning content of the LC. Acquired Knowledge. A set of kis denoting the knowledge that is supposed to be gained after taking the LC. Learning Content. Defined as an XHTML resource, possibly in four versions (one for each combination of the LS dimensions sensing/intuitive and visual/verbal, each one weighted by values e.g. to state how much sensing and how much verbal is the fourth version). Annotations. Notes (one for each value in the LS dimensions active/reflective and sequential/global) devised to be submitted to the learner during the taking of the LC. Except the global annotation, they are shown together with the learning material of the LC. The global annotations, instead, are used to provide the learner with a summary of each lesson (i.e. subset of the LCs in the course). Questions. A set of quizzes, deemed to be used during the verification process of a course. Each question can be labeled, if reasonable, by a ki (so right answer to such question will give possession of the related ki). Questions are embedded in the LC because this allows to contextualize them to the actual learning material of the LC, which we suppose makes them more effective. Index and Effort. A name for the LC and an informal quantification of the effort needed to take the LC content (effort is used to balance the partition of the course in lessons). The system builds a personalized course as a selection of LCs; the selection (which is called configuration) is based on the Target Knowledge (TK) and on the individual Starting Knowledge (initial state of CS). TK is a set of kis, as well. The set of LCs (a course configuration) is linearized in lessons, according to appropriate constraints on lesson length. Then the learning path is administered, by presenting the material in different ways, depending on the personal learner’s learning styles. In Figure 2 the functional schema of Lecomps is depicted. A Learning Environment defines the part of the system from where the courses about a certain subject matter will be constructed and delivered, under the administration of a certain teacher. It comprises: Domain of knowledge. All kis used in the definition of the LCs (the corpus of knowledge - or, better, the names of the concepts in the corpus - possibly managed by the courses). LC Pool. The repository of LCs, managed by the (authors and) teachers, through the authoring tools available in Lecomps. Basic TK. A set of kis, designing the target knowledge that is common to all the learners’ personalized courses; the TK of each individual course might be varied by the teacher, at configuration time.


271

Fig. 1. Functional schema of the Lecomps system

Approaches. A set of approaches designed by the teacher. An Approach is a way to express special preferences of the teacher with respect to the content of a personalized course. It can be described either as a set of kis, or as a set of LCs, which will be forced in the course configuration. An approach can be used to constrain a course in such a way it will use a certain terminology, or just will contain a certain set of learning material, in spite of the possibility it is not indispensable to reach the basic aims of the course. Verification Policy. This is a set of parameters that drives the management of the CS during the taking of an individual course. Such parameters say what certainty is given a ki when it enters the CS after a right answer in a test, or how much such certainty increases or decreases after answers to further questions about that same ki, or what certainty thresholds are assumed for the kis to be extracted or permanently joined to the CS (under a threshold the ki is lost by the learner; above another threshold, no further questions will be met by the learner about that concept). When a student enrols at a course, the Initial Questionnaire and the ILS Questionnaire initialize, resp., the CS and the LS in her/his model. Then, basing on the present (initial) student model and on the basic TK (possibly augmented by the teacher through some approaches), the personalized course is configured. The course is a sequence of LC, suitably partitioned in Lessons. When a component is presented, the learning content and the annotation versions are selected appropriately from the LC, with respect to the present evaluation of the LS. After each lesson an intermediate quiz is defined dynamically, by randomly selecting questions from the LCs in the lesson, so to check about the kis “apprehended”

272


during the lesson. The answers allow to update the student model: the certainty of kis already in CS may be modified; kis may be added, and others may be extracted. Moreover, the LS can be updated basing on the answers, under the assumption that a right/wrong answer witnesses LS-adequacy/inadequacy of the presented learning material (so the present LS evaluations are consequently strenghtened/weakened, according to LS weights that label the presented material). After student model update, the course might be reconfigured, either by the teacher, or automatically (that’s another parameter of the Verification Policy settings).

3

Integration

The integration of Lecomps in moodle is illustrated in Figure 2; it is performed proposing Lecomps as a moodle activity.

Fig. 2. Integration of Lecomps in moodle: functional view

The basic management of users registration and authentication is provided by moodle. The integration proceeds through a communication layer, set up between moodle and Lecomps. The first task of such communication layer is to maintain synchronized the moodle database of quizzes with the LCs repository: basically each time a question is added to a component, it can be transferred


273

to the moodle database; this is a semiautomatic process, presently. A learner, registered in moodle, can enroll in a course and be presented with an initial questionnaire. The initial questionnaire is designed by Lecomps (SK Inspector module) and communicated to moodle. moodle presents the questionnaire and collects the results. Then feedback is sent back to Lecomps, in order to initialize the Cognitive State. The Learning Style is initialized in this stage as well, yet the process is carried on entirely by Lecomps. After the configuration of the personalized course is done, it is presented to the learner.

Fig. 3. Users’ access into the integrated system: above, the registration phase is shown; below, the access to the moodle Lecomps-Activity (related to a course called “lucus3”)

When a (intra lessons) quiz has to be delivered, Lecomps produces its specification and transfers the data to moodle, which will 1) prepare the test extracting the questions from its database; 2) present it to the learner; 3) collect results. Then the results are transferred back again to Lecomps, and the student model is updated accordingly.The quiz will be stored, for possible further analysis from the teacher.

274


Fig. 4. A small quiz, made up by moodle, with a selection of questions decided by Lecomps and extracted from the moodle quizzes databases

Fig. 5. Report about questions and questionnaires


275

This approach exploits moodle modularity: Lecomps is installed in moodle in the directory mod as a new module. The teacher logs in moodle; (s)he can introduce a Lecomps activity in his courses, and navigate the Lecomps environment. Figure 3 shows the student’s functionalities performed in moodle: the student is registered as a moodle user, (s)he enrolls the course through moodle authentication and moodle course registration, and (s)he obtains her/his personalized course through the Lecomps personalization engine, seen as an activity of the course (”lucus3” in the bottom of the figure). Figure 4 shows a quiz delivered to a learner. As mentioned above, Lecomps produces the list of questions for a quiz, by selecting randomly from the questions contained into the learning components seen in a lesson. Such specification is automatically imported in moodle, where the real quiz is composed (by accessing the quizzes database), and presented to the learner. The feedback coming from the answers to a quiz go back to Lecomps, in order to perform the necessary updates over the student model. Meanwhile the quiz is stored with its results. When a further analysis of quizzes is required, the teacher can access them, as shown in Figure 5.

4

Conclusions and Future Work

Our research is aiming at the development of a full integration of capabilities, for the support to personalized and adaptive courses, in a state-of-the-art LMS. In particular we are working on moodle, due to its well known modularity and extendibility, and on a standalone LMS that we developed earlier, Lecomps, in which we implemented our approach to personalization and adaptation in e-learning. In this paper we propose a step in such integration. We have developed a first level embedding of the Lecomps system in a moodle module (namely an activity). Yet, by no means we claim that a complete integration has been reached. The integration, so far, is limited to the management of enrolling and authentication, and to the automated import of the questions defined in the learning components, into the moodle database. As limited as it is, the simple communication implemented between moodle and Lecomps let us gain something relevant: the quizzes defined by Lecomps after each lesson of the course, are stored, with their answers, for future consultation; history of the learner’s participation in the course (through answer to questions) is recorded; the records of the answers given by all the students to a same question, are stored and can be evaluated, allowing for an analysis of the questions defined in the learning components. Presently the personalized courses are presented through the moodle interface, yet it is still the original Lecomps system acting to select the material to be actually presented (basing on the definition of the course, as a sequence of Learning Components, and on the actual evaluation of the learner’s learning style). All the other functionalities of Lecomps, such as the authoring tools,

276


the graphical interface to the repositories of learning components, and the interaction of the teacher with the configuration interface (to manage learning environments and to actually produce courses), are still visibly in Lecomps. In terms of future work, we plan to advance along the integration path. While working on the lacks mentioned in the previous paragraph, we intend to devolve to moodle several tasks, related to the presentation of the courses, the storing of the previous versions of a same adaptive course, and the collection of data relevant to the student model update, so to leave to Lecomps software only the kernel of personalization activities. Another aspect we plan to work upon, is the accomodation of the learning components in the framework of the SCORM standard for e-learning. We plan to use such result in order to allow the definition of a personalized course as a SCORM package, and let moodle receive it and present it through its implementation of the SCORM API.

References 1. Baldoni, M., Baroglio, C., Brunkhorst, I., Marengo, E., Patti, V.: Reasoning-Based Curriculum Sequencing and Validation: Integration in a Service-Oriented Architecture. In: Duval, E., Klamma, R., Wolpers, M. (eds.) EC-TEL 2007. LNCS, vol. 4753, pp. 426–431. Springer, Heidelberg (2007) 2. Brusilovsky, P.: Adaptive Hypermedia. Journal of User Modeling and UserAdapted Interaction 11, 87–110 (2001) 3. De Bra, P., Smits, D., Stash, N.: Creating and delivering adaptive courses with AHA! In: Nejdl, W., Tochtermann, K. (eds.) EC-TEL 2006. LNCS, vol. 4227, pp. 21–33. Springer, Heidelberg (2006) 4. Felder, R.M., Silverman, L.K.: Learning and teaching styles in engineering education. Engineering Education 78(7), 674–681 (1988) 5. Felder, R.M., Spurlin, J.: Application, reliability and validity of the index of learning styles. Int. Journal of Engineering Education 21(1), 103–112 (2005) 6. Limongelli, C., Sciarrone, F., Temperini, M., Vaste, G.: Lecomps5: a Framework for the Automatic Building of Personalized Learning Sequences. In: Lytras, M.D., Carroll, J.M., Damiani, E., Tennyson, R.D. (eds.) WSKS 2008. LNCS (LNAI), vol. 5288, pp. 296–303. Springer, Heidelberg (2008) ISBN/ISSN: 978-354087780-6 7. Sangineto, E., Capuano, N., Gaeta, M., Micarelli, A.: Adaptive course generation through learning styles representation. Universal Access in the Information Society (UAIS) 7(1-2), 1–23 (2008) 8. Sterbini, A., Temperini, M.: Adaptive construction and delivery of web-based learning paths. In: Proc. 39th ASEE/IEEE Frontiers in Education Conf. (2009) 9. Weber, G., Brusilovsky, P.: Elm-art: An adaptive versatile system for web-based instruction. Int. Journal of AI in Education 12(4), 351–384 (2001) 10. Wolpers, M., Grohmann, G.: PROLEARN: technology-enhanced learning and knowledge distribution for the corporate world. Int. Journal of Knowledge and Learning 1(1-2), 44–61 (2005)