Internet-Zeitschrift für Kulturwissenschaften	17. Nr.	März 2010
Sektion 8.15.	New Approaches, Innovations and Research in Education | Neuigkeiten, Innovationen und Forschungen in der Erziehung SektionsleiterInnen | Section Chairs: Leyla Esentürk-Ercan and Melek Çakmak (Gazi University, Ankara, Turkey) Dokumentation | Documentation | Documentation

Adaptation of e-Learning Environments:
Determining National Differences through Context Metadata

Thomas Richter [BIO] | Jan M. Pawlowski [BIO] (KGIT, Seoul)

Email: thomas.richter@icb.uni-due.de | jan.pawlowski@icb.uni-due.de

Abstract:

The paper shows how existing e-learning modules can be internationalized using structured information on the context and specifically culture. Reusing e-Learning contents is a promising concept for the internationalization or cross-cultural purposes. However, most adaptation efforts are often limited to pure language translation.

As the only alternative is rewriting, reusability allows a massive cost reduction by implementing and adapting already established courses, for example into developing countries on a low-cost level. Our approach provides a basis for international and cross-cultural adaptation. In the approach, we identify, collect and store as many parameters about the source and target context and culture as available. After comparing contexts, we determine changing needs by analyzing the impacting differences.

To implement this approach on a large-scale, we plan a public database containing the necessary information for the comparison process. In our research, we have identified a set of around 170 parameters describing national and, more specifically, cultural attributes related to various situations. Utilizing those in an adequate way will lead to an easier and efficient adaptation process.

 

Introduction

In this paper, we present an approach to describe and compare different learning contexts on the basis of context metadata. Transferring e-learning contents and scenarios from one context to another, e.g., from a higher education context in Germany to management training in Korea, requires a high adaptation effort. How can existing learning scenarios be re-used? For this, we propose structured descriptions of the context, e.g., in a database for public use, containing as many context descriptions as necessary. This shall lead to a procedure to compare contexts, provide adaptation guidelines for the transformation process, and to efficiently assist in developing solutions based on existing materials.

In the following, we first explain the general concepts on the adaptation of e-learning and re-use. We deduce the need for a common description of contexts using a metadata approach. Afterwards, we show the concrete context metadata classes and an example of how the comparison process will work. Finally, we show difficulties within the evaluation on what we can deduce out of and expect from the data, gathered through the comparison process.

The Context of e-Learning and the Adaptation Process

The context of e-Learning denotes influence factors on learning scenarios and, more specifically, e-learning environments. This means that we consider only factors which have an influence on the learning process but cannot be influenced by the designer. As an example, the time in which a course is held is part of the defined e-Learning environment and directly related to the course. But the country’s time zone(s) in which the course takes place is part of the context. An e-Learning course, which is for example designed for the synchronous transmission in the whole Russia over the Internet or television, has to deal with 11 time zones at once. The huge number of time zones can be a reason not to use a synchronous communication approach in this case. The counterpart to the context of e-Learning can be seen in the e-Learning environment, which for this paper shall be defined as all components which are designed on purpose within the development process, e.g., the course itself, defined preconditions concerning the age or knowledge of the human actors, the recommended literature, etc.

The adaptation of learning to different or specific situations and contexts is a widely discussed field. Edmundson [Edmu07] discusses cultural adaptation on a general level, defining relevant dimension for cultural adaptation. Guetl, Garcìa-Barros and Moedritscher [Gue+04] discuss the adaptation on certain (cultural) user needs. Other approaches focus on specific systems, such as Adaptive Hypermedia Systems [Brus97] or intelligent software solutions [HoHa98, CrDe02]. All approaches are relevant, however, there is no holistic approach, taking into account the most relevant influence factors for different classes of e-learning scenarios and systems. A general approach has to take a lot of further influence factors into consideration so that an adaptation process for e-Learning can be considered as complete and successful. An adaptation process is considered to be successful if a course enables students within the new context to experience at least the same knowledge gain than those within the original one. Therefore, we need to define the main influence factors.

The adaptation process for e-Learning consists of five steps as shown in the Figure 1: Firstly, suitable course modules are to be found. In the second and most significant step, the reusability as well as the changing needs are being determined and evaluated. Re-use aims at the efficient development of courses to avoid “re-inventing the wheel”. Re-use is discussed in various contexts, specifically in the field of software engineering and development [Jaco+97]. In the educational context, the discussion focuses on learning objects [Wile00] and learning activities [KoMa04, MaSa05]. The variety of aspects on re-use is discussed in [LiBu03]  The main outcome is whether it is useful or efficient to re-use existing materials / scenarios. If the course module can not be reused for the targeted context, other modules have to be found which fit the requirements better or the course module has to be rewritten. In the third step, the adaptation process starts by considering context information. Adaptation in this concern means changing the course module Cx (A) and its requirements in a way that it can be reused in the targeted context Cx (B). After having finished the adaptation process, the solution has to be validated to find out if the result Cx (B) corresponds to the needs within context B. This validation should utilize common processes, e.g., using the ISO/IEC validation criteria [ISOI05]. If it does not fit the requirements, the adaptation process has to be redone and maybe also the results of the comparison process in step 2 have to be reevaluated. After the validation, the course can be republished.

Figure 1: The comparison within the adaptation process

Context Metadata

Context information can be described in many different ways. However, it is necessary to describe contexts in a format which can be used to communicate those contexts amongst people or to be used for adaptive information systems. To achieve such a solution, we use context metadata to represent the variety of information in a structured, machine-usable common language.

According to the W3C group, Metadata are considered as “data about data” [W3C98]. The less abstract definition of the IEEE group describes metadata as “information about an object, be it physical or digital” [IEEE02].

Therefore, context metadata are defined specifying influence factors on learning environments in a clearly defined data structure. These data can for example be stored within a public database and addressed for search, direct download by people, and applications such as expert systems.

In the following, the classification model of the e-Learning context metadata is discussed. We focus specifically on cultural issues in the adaptation process. Carrel & Eisterhold show the significance of cultural aspects for the adaptation of learning environments as they write that “one of the most obvious reasons why a particular content schema may fail to exist for a reader is that the schema is culturally specific and is not part of a particular reader’s cultural background” [CaEi83]. Most found influence factors are related to culture in some way. However, modeling our metadata approach, a class “culture” has to be separated from other fields. Therefore, a clear definition of the term culture and its differentiation from other fields is necessary. Following the definition by Hofstede [Hofs91], we define culture as a mental coding which every member of a society, organisation or group experiences and acts corresponding to this. This definition allows the conclusions that in our case not everything is to be considered as being called an aspect of culture and further on that cultural aspects are only to be considered, when they correspond to the behavior and thinking of a group (majority) of people (and not to individuals as Trompenaar suggests [TrHa06]).

Another found modeling problem is that one culture per nation can not be implied [HoHo05, Gann04, Pog05]. This would not fit the reality. Therefore, we design our metadata and related database so that a single country needs to be relatable to more than a single cultural dataset.

Different approaches have already defined dimensions of cultures [HoHo05, HaHa90, Hend96]. Since concrete attributes are needed for the comparison process, the dimension models do not go far enough. Nevertheless, during the research, they allowed the definition of some (e.g.) cultural context metadata. Based on those models and approaches, we designed context blocks representing classified aspects of context. Around 170 parameters were identified. Designing the context blocks in a way that no relations in between remain has not been possible, but there have been found attributes within every single class, which are not related to other classes and on an intuitive level, every class contains only such elements which internally are more related to each other or the class than to others  The data behind the single context metadata vary from a single numerical value to complex collections of documents in different formats and languages. In the following, a list of the single context blocks is shown followed by some references, which led to them.

• Demographical development [UN06], [ChWe04]
• Religion [Asch06], [Pau80]
• Technical infrastructure [Edmu07], [Seli04]
• Rights [Lean05], [Marc04]
• History
• Politics [Kear90], [Davi06]
• State of development
• Media richness [Wils02], [Gul+06]
• Financial aspects [Guna05]
• Human actors [Zimm01], [Niel93]
• Rules [Wood04]
• Companies [DaSe04], [DIN04a]
• Geography and educational infrastructure [SaFo02], [LaZh03]
• Learner satisfaction – Known demands [John00], [DaJo02]
• Internet security [Kno+03]

Culture as the still largest class of influence factors has not been included above, because it can further be divided into the following sections:

• Language [Leon02], [DeMa06]
• Ways of communication [Cak+02], [Ting88]
• Humor [Kend06], [LiLe07]
• Acceptance of media types [Wils02]
• Gender differences [Hasl00], [Hend07]
• Social capital [Mer+02], [Stone01]
• Habits and preferences [Noce02], [DuMa07]
• Acceptance of technology [HeNk06]
• General opinion [Bea+06]
• Taste [Loh+06], [Gans99]
• Indigenous cultures [Dys+06], [HuDa07]
• Hints for technology transfer [LaDy06], [LeSm94]
• Pedagogical approach [Mcca07], [DuMa07]

The dependencies between the context blocks are shown in Figure 2, as they are to be modeled within the database and correspond to our recent state of research  As an example, the classes “Technical Infrastructure”, “Rights” and “Human Actors” also are strongly related to a lot of other context blocks.

Figure 2: Dependencies between Context Blocks

By our approach, we have outlined a discussion base to describe context in a structured way. It is clear that a lot of research remains: For specific purposes, extensions are needed, for other purposes only a few categories are needed. Building those application profiles will be the task in the user communities.

The Comparison Process

The comparison process shall yield the actual differences between two contexts as a result. For the comparison process and in special to find a solution on the decision which of the found differences are significant for changing needs, various methods can be used, as similarity comparisons (similar situations have been handled in a certain way) and recommender systems [MaSa04] (decision support systems). They also can base on former experiences [PaBi06], e.g., by using knowledge databases. The final decision on changing needs at least at the moment has to be done manually: Until now, the number of documented experiences is very low and focused on certain aspects and situations (e.g., user satisfaction after changing the level of activity, a.o.). Another reason is the fact that a monitored difference between two contexts does not necessarily mean that a course module (or course) actually is impacted by this attribute. Therefore, we need a collection of experiences, to be made after and during the adaptation process. Such documentations will be crucial to determine cross-impacting attributes and the kind and level of their impact.

Before the data can be compared, they have to be gathered. One data set is partly attached to the course module which is to be adapted. The attached data are those, which describe the individual skills and attributes of the author. All the rest of the data concerning the course are attached as references to the database. This shall reduce the overhead. The other dataset related to the targeted context completely has to be picked up at the database. The gathering of data as far as the data are available (stored within the database) can automatically be collected.

Figure 3: Data-gathering and -comparison procedures

In Figure 3, the necessary steps to realize the comparison process are visualized. The arrows with dashed lines symbolize data requests while those with full lines represent actions in which data are transported (unless if tracked or pushed)  The process is initialized after the data collection function has been started. The function gets the name of the targeted context Cx(B) and the storing location of the course module as input information. In the following the single actions are named by number: Firstly, the data concerning context Cx(A) which are attached to the course are collected: (1 + 2) the personal data representing the author’s individual attributes are tracked out of the course module (1 access to the document, 2 reading and storing the data). The general information which are relevant to describe the author’s context represented by pointers on the database are identified (3), tracked (4) and the data taken (5) out of the database. (6) The complete dataset concerning the targeted context Cx(B) are requested at the database by name (pushed or pulled). The data are given to the comparison function (8), which is divided in a part which realizes automatic comparisons (output 10a) and pre-structures (9) the information for the manual comparison process (output) as far as they cannot be compared automatically and the manual comparison process which for the further automated use again has to lead to a computable input (10.b). The result of those two processes is a list (10a, 10b) which shows all differences between the two contexts. Data, which only are available within a single context, are to be handled as differences  This list finally is handled to the evaluation process in which later on the adaptability will be checked and recommendations on changing needs can be given.

The evaluation of the differences deducing changing needs provides difficulties because there are not only impacts resulting of pair-wise differences but also because of combinations of different attributes, which are not yet researched. At least it will have to base on documented experiences  Furthermore, the differences - if only pair-wise taken into consideration - do not lead to a trivial decision, because depending of the targeted context the acceptance level concerning unused situations sites the users play a significant role. Also here in most cases at least in the first time, manual decisions are unavailable. What the system can do is providing recommendations (as a recommender system) but taking the full decision is not possible because of expectable cross-effects

The comparison process together with the evaluation phase will exemplarily be shown in the following on sample data sets of South Korea and Germany. First of all, selected data are to be shown as stored within the database. The attributes have consciously been chosen in a way that different data structures which need different methods for comparison are shown.

Table 1: Sample comparison between Germany and the Republic of South Korea

The data in table 1 are chosen in a way to demonstrate general differences between collectable data and their structures on the one hand and the different ways how to compare and at least evaluate them. There are further fields in the database, such as the related contextual elements, attribute-definition-time, date of the last update, information source, data-structure-type (type) and a defining description (kind) of the contained information.

In the following, the attributes are pair-wise discussed first on a general level. We try to contrast the differences intentionally as much as possible, as it is clear that there are nuances in between depending on individuals. Afterwards, the comparison takes place, followed by an evaluation on how the data can be interpreted. First representative concrete and interpretable experiences within the described context are expected after finalizing an evaluation in October 2007.

1 – Teachers Role: The teacher’s role in Germany is the one of an assisting adult [HoHo05] who basically shows and helps the learner to understand the learning contents [Hend96]. He is allowed to present wrong statements as a method to provoke discussions or irritations within the group of learners. In Korea, the teacher is seen as person of authority, rarely to be questioned  Spreading wrong information would irritate the students and in the worst case could result in the loss of respect.

As a consequence, this means that tasks within the learning content, which ask for the critical discussion of content or methods, have to be presented in a different way. In the opposite way (adapting a course from South Korea to Germany) the reaction sites the students may be irritated, they could refuse their sympathy if the author’s writing style or the tutor’s behavior is too dictatorial but finally they would be able to deal with the situation.

2 – Value of Errors: Depending on the culture, the value of errors can be seen very differently [Hend96]. While errors in Germany mostly are seen as a chance to learn, in South Korea they show a lack of knowledge and expertise and can cause a face-loss. In the consequence for e-Learning courses and in special for tutorials and evaluated practices, the way how to point on errors has to be chosen very differently: While in Germany it is usual to directly tell a student that (and why) an answer or implication is wrong, in Korea this has to be done more cautiously, e.g., by stating that the chosen solution has been already good but could be better while taking an other way or different conclusions.

3 – Context Type of Society: The division of societies into the two types, high context and low context, has been defined by Hall [HaHa90]. While in high context cultures a lot of contextual elements help people to find their place within the society and to understand rules, low context cultures do not provide this feeling of security but define everything which is seen as necessary information explicitly. In this definition Germany is a low context society and Korea a high context society. Regarding e-Learning, the specific shape of this dimension is significant to know from both, the originator’s society and the targeted society. Typical attributes which impact learning for high context societies are for example a strong developed nonverbal communication, a strong distinction between in-group and out-groups, a high use of metaphors and implicit messages (a lot is interpreted between the lines), a higher preference to relationships than to reaching goals and a process oriented working style (instead of product oriented). As a consequence, contents, texts and pictures have to be modified as well as tasks and group work need to be newly defined.

15 – Cultural Variable, Language: Not only the language itself is significant but also the cultural component which is integrated in the way how language (and silence [DaJo02]) locally is used [Leon02]. Taking the upper example with the high and low context societies it is obviously that even if both contexts have to deal with their first foreign language the interpretation of the contents may be very different. But even when both countries use the same mother-tongue English irritations can appear [DeMa06]

45 – Education Achievement [Ram+07]: The university achievement is interesting from more than one perspective; there are cross-effects to other context metadata  We only show some obvious examples and consequences. First of all, knowing about the education structure allows conclusions on different curricula and the necessity to pay fees for education in certain levels, combined with a possible duty going to school (context metadata No. 80, RM50013) until a certain minimum level gives hints on the (generally) expectable knowledge and skills of the students within a country.

Comparing South Korea with Germany, in Germany the basic education is completely free and education is defined as a basic human right. The first 10 levels of school (Primary 4, Middle 6) are duty by law. Children can, if they fulfill the preconditions additionally (free of charge) go to the high school (2-3 years). The university in Germany is also free of charge as far as it is the first higher education. In South Korea there is a duty going to school until end of the Middle School (9 years). Primary School (6) and Middle School (3) are free of charge. High School and university are to be paid. The aim of the learning process is the best possible education so that there is a chance to reach a higher social position.

These metadata can not be used for explicitly defining actual changing needs but gives hints if other metadata in special in combination crucially are to be taken into consideration. At least, the experience will show if such kinds of metadata provide the expected use.

64 – Mobile Technology Infrastructure: This attribute in special should be used when the implementation of mobile learning technologies is faced. Depending on the used type [ArCl04] of mobile technology (context metadata No. 65 TIM40010) and the divide within the country (also attachable to regions [Guna05], if defined) different technologies can be used [TiLa04]. The mobile standard in Germany is the 3G-Standard, including GSM, GPRS and UMTS. While GSM (low bandwidth) is nearly everywhere reachable, UMTS as a high-speed mobile standard is reduced on the urban environments. Location based services in a useful way are only possible within certain regions. In South Korea, CDMS is the commonly used standard and it is accessible in all habited regions, so that nearly everywhere a high-speed access to the internet via mobile technologies is possible. UMTS recently is going to be implemented. The difference between both countries in the kind of technology and its divide for example shows that education via mobile television or multimedia content can easily be implemented in South Korea but right now not within Germany.

129 – Self Set Educational Goals: As already discussed, this attribute represents the motivation which causes students to learn [MoKe96]. While the German learn because they want to gather knowledge, be able to manage their live by getting the ability to solve problems and reach social competence, the South Korean students want to fulfill the society’s (in special their family’s) expectations and find the highest possible position within the society. The kind of learning motivation alone is no indicator on changing needs and does not need to be evaluated in this meaning. It is not useful to change a constructivist course into a behaviorist one because the learning purpose is different. However, this attribute can point out that it could be useful introducing on how to deal with open tasks when they are to be implemented. The gathering and comparison of the data can be done automatically but the evaluation is to be done manually.

As we have shown by those examples, the comparison process based on our context metadata can provide useful hints for the adaptation process. There are further effects which can be expected, such as a better awareness of problems and changes in intercultural settings or new systems using our adaptation guidelines. However, at the current state our framework will enable users to take many aspects into account which have been neglected in many adaptation projects.

Conclusion

The context-based internationalization of e-Learning situations is related to a lot of different aspects to be considered  Cross-effects between those fields as well as the necessary differentiation of regions and societies within single countries and the non-static character of the context lead to a complex adaptation process. Not even the comparison of two contexts can be realized automatically at the current stage. The act of changing the content and aspects of the situation will stay a manual procedure based on human decisions. The result of this research until now is the knowledge about significant factors which now can be evaluated and extended.

8.15. New Approaches, Innovations and Research in Education | Neuigkeiten, Innovationen und Forschungen in der Erziehung

Sektionsgruppen | Section Groups | Groupes de sections

Inhalt | Table of Contents | Contenu  17 Nr.

Webmeister: Gerald Mach     last change: 2010-03-04