Creating Life-Long Learning Scenarios in Virtual Worlds

Virtual worlds, such as Second Life, were considered an emerging technology surrounded by hype and growing educational expectations. These immersive world applications have the potential to support multimodal (using different senses) communications between learners; they set up the potential for problem – or challenge-based learning and offer the learner control through exploratory learning experiences (Saunders, 2007). In the recent Metaverse Roadmap Report 2007, Smart et al. (2007) envisage a powerful scenario over the next 20 years when: “[virtual worlds] may become primary tools (with video and text secondary) for learning many aspects of history, for acquiring new skills, for job assessment, and for many of our most cost-effective and productive forms of collaboration. (Smart et al., 2007: 7)” However despite this, much needs to be understood about how to best convert these spaces for learning purposes such as seminars, simulations, modelling, learning activities, networked learning experiences, cybercampuses and streamed lectures (Prasolova-Forland et al., 2006). Otherwise, virtual worlds might be the next misused educational technology. Many educational technologists would agree that poor utilization of the features of a technology will enviably lead to complacency with that technology and probably lead to it being either being abandoned, or worse, massively underutilized (Rappe et. al, 2008). Derived from these statements, the purpose of this chapter is to describe the integration of metadata into Second Life to foster the growth of digital learning spaces. In an effort to assist European practitioners (individuals as well as the existing communities) that work in the field of education and training educators and are genuinely interested in using Second Life within a Lifelong learning perspective the LLL3D (Life-long Learning in Three Dimensions) project team is promoting opportunities for teachers, trainers, researchers to discover, learn about and utilise different “learning scenarios” for virtual worlds. The learning scenario approach provides access to best practice case studies across formal, non-formal and informal levels and different sectoral activities. The LLL3D project is contributing to the establishment of a European research and practice area in Lifelong Learning, paying attention to the promotion of general awareness of the potential of MUVEs, dissemination and to increasing the acceptance of MUVEs (Multi-user Virtual Environments) as a highly promising cutting-edge technology for online learning. In the framework of the LLL3D project, the working group of different partners have created a set


Introduction
Virtual worlds, such as Second Life, were considered an emerging technology surrounded by hype and growing educational expectations. These immersive world applications have the potential to support multimodal (using different senses) communications between learners; they set up the potential for problem -or challenge-based learning and offer the learner control through exploratory learning experiences (Saunders, 2007). In the recent Metaverse Roadmap Report 2007, Smart et al. (2007) envisage a powerful scenario over the next 20 years when: "[virtual worlds] may become primary tools (with video and text secondary) for learning many aspects of history, for acquiring new skills, for job assessment, and for many of our most cost-effective and productive forms of collaboration. (Smart et al., 2007: 7)" However despite this, much needs to be understood about how to best convert these spaces for learning purposes such as seminars, simulations, modelling, learning activities, networked learning experiences, cybercampuses and streamed lectures (Prasolova-Førland et al., 2006). Otherwise, virtual worlds might be the next misused educational technology. Many educational technologists would agree that poor utilization of the features of a technology will enviably lead to complacency with that technology and probably lead to it being either being abandoned, or worse, massively underutilized (Rappe et. al, 2008). Derived from these statements, the purpose of this chapter is to describe the integration of metadata into Second Life to foster the growth of digital learning spaces. In an effort to assist European practitioners (individuals as well as the existing communities) that work in the field of education and training educators and are genuinely interested in using Second Life within a Lifelong learning perspective the LLL3D (Life-long Learning in Three Dimensions) project team is promoting opportunities for teachers, trainers, researchers to discover, learn about and utilise different "learning scenarios" for virtual worlds. The learning scenario approach provides access to best practice case studies across formal, non-formal and informal levels and different sectoral activities. The LLL3D project is contributing to the establishment of a European research and practice area in Lifelong Learning, paying attention to the promotion of general awareness of the potential of MUVEs, dissemination and to increasing the acceptance of MUVEs (Multi-user Virtual Environments) as a highly promising cutting-edge technology for online learning. In the framework of the LLL3D project, the working group of different partners have created a set of services and tools in order to provide guidance to all practitioners who are interested in exploring the potential of Second Life for educational purposes, and learning how to use this environment. One of the essential aspects of the project was to integrate a metadata scheme into a grid that supplies researchers and practitioners with a system of existing and possible successful approaches to using MUVEs in learning and teaching. It is our desire that the examination of this project highlights the increasingly recognised importance of structured metadata for the development of learning scenarios and the transferability of this approach to other educational domains and eventually illustrate the possibilities, encourage you and spark ideas of your own. After providing a general overview of case-based learning, this chapter proceeds with a metadata scheme that can be used for the integration of learning scenarios into MUVE s(Multi-user Virtual Envrionments) such as Second Life (SL). Although not all of the observations are positive on their own, pinning down the details of which educationally significant characteristics pertain to which entities in learning scenarios and which relationships are important is a crucial step in understanding what information is needed to create resource descriptions that meet educational requirements, and how to go about gathering that information.

Main definitions
A useful definition of metadata is that used by NISO (2004) "structured information that describes, explains, locates, or otherwise makes it easier to retrieve, use, or manage an information resource". This definition has two important parts. Firstly, it distinguishes metadata from unstructured textual descriptions of a resource. The structuring of metadata normally takes the form of elements with defined semantics to describe specified characteristics of a resource so that machine processing of the information can ocur without a need for computational semantic analysis techniques such as text mining. Secondly, the NISO definition stresses that metadata exists to facilitate a range of activities. Resource discovery is the most visible activity facilitated by metadata, and is the one that seems most closely associated with metadata by most people; however, appropriate management and use of resources are no less important. Defining what we mean by learning scenarios is more difficult. However, we think that "anything used for teaching and learning" captures the essence of what we are interested in. At this stage, an example for a learning scenario would be "all learning settings that use collaborative learning for learning a language with adult learners only in-world". This scenario would contain all the real cases that could be found in any MUVE and also those that are only imaginable. Our learning scenario grid consists of 4 dimensions (learning/teaching approaches, discipline/subject, target audience, type of interaction) with 20 sub-values. This grid should not be developed mainly by theoretical considerations but by empirical work, by collecting concrete examples of the classes of learning activities (e.g. an example/practice case for a face-to-face learning approach for language learning).
The following sections will focus on one of the main established metadata standard most relevant to learning scenarios, the IEEE LOM (Learning Object Metadata), and will briefly describe and reflect on its characteristics and applications. This paper will also outline the current work being undertaken on this schema. Finally we look at some of the future challenges facing the field of metadata for learning scenarious regardless of which specific standard one favours.

An overview of case-based learning
New standards and specification are more oriented towards describing learning scenarios than just contents. These specifications try to describe all the aspects and the elements more related to the learning process in itself, such as role playing, that is, the second level of description as aforementioned. It seems clear that all this this information needs to be stored in a higher semantic level. Although the metadata schemes may seem too complex for learning scenarios, their flexibility and multilevel description capabilities allow the specification of any learning process ranking from simple educational itineraries to complex processes including collaborative working capabilities. Nevertheless, these metadata schemes lack from a formal description for some of the definitions required for developing learning scenarios in virtual worlds. Each case in the LLL3D is designed as one or more learning situations trying to reproduce real professional situations where students in one field need to apply practical knowledge for solving a problem, in a virtual world environment. This methodology, which tries to ensure a high quality of the learning process, takes into account all the learning activities that are designed with all the learning goals in mind, in the following structure:

Resources
This structure is also partially supported by a case study template with the description of the subject, which is human-readable, but non machine readable. Each case study template has been designed on the basis of the following premises: -A sound formulation of competences and learning goals; -Learning activities which are coherent with the competencies to be developed Within the context of the LLL3D project, we define case-based learning (CBL) as an instructional strategy that uses case study as a resource and the case method as the learning scenario description where learners and instructor interact. The case study template is a descriptive document based on a real situation or event. The case tries to facilitate a balanced relationship between the multidimensional representation of the context, its participants and the the reality of the situation. A case can be used to generate different case studies from a subset of case patterns and a collection of learning resources, following an instructional design approach. Therefore, at the bottom level, we need formal representations for case-based learning scenarios, which involve all the elements in the learning process (learners, activities, competences, resources, etc.) (Barker et al, 2006). The goal is to provide a mechanism for scenario design for learning in virtual worlds according to learner preferences and already acquired competencies and learning goals given by teachers (Barker et al, 2006, Duval et al, 2002. It is necessary to adapt the particular needs of the virtual learning scenario to the specifications available where competencies are used to describe goals and outcomes of learning activities (Barker et al, 2006). In this sense, there is a lack of standards for describing competencies at a rich semantic level because major metadata schemes such as IEEE LOM are not enough to represent all these relationships identified previously.

IEEE Learning Object Metadata (LOM)
The IEEE LOM is (currently) an open and internationally recognized two-part standard for the description of "learning objects" and is composed of a conceptual data schema (IEEE, 2002) and an XML binding of that schema (IEEE, 2005). The definition of "learning object" used in the standard is "any entity digital or non-digital that may be used for learning, education, or training", which is comparable to the working definition used above. The LOM data schema specifies which characteristics of a learning object should be described and what vocabularies may be used for these descriptions; it also defines how this data model can be amended by additions or constraints. The LOM conceptual data schema consists of a hierarchy of elements as shown in figure 1. The first level is composed of nine categories, each of which contains sub-elements; these subelements may be simple elements that contain data, or they may themselves be aggregate elements that contain further sub-elements. The semantics of LOM elements are determined by their context: they are affected by the parent or container element in the hierarchy and sometimes by other elements in the same container. For example the various description elements (1.4, 5.10, 6.3, 7.2.2, 8.3 and 9.3) each derive their meaning from their parent element: e.g. 5.10, education.description describes educational characteristics of the resource; 6.3 rights.description relates to the terms and conditions of use of the resource, and so on. In addition, description element 9.3 also derives some of its meaning from the value of element 9.1 purpose in the same instance of the classification category element. The data schema also specifies the value space and datatype for each of the simple data elements. The value space defines the restrictions, if any, on the data that can be entered for that element. For some elements the value space allows any string of Unicode characters to be entered; for other elements entries must be drawn from a declared list (i.e. a controlled vocabulary) or must be in a specified format (e.g. date and language codes). Some element datatypes simply allow a single string of characters to be entered; others comprise two parts as described below: LangString datatype: where the data entered is likely to be text that would be read directly by a human the data is of a type defined by the LOM as a LangString. LangString items comprise two parts: one providing a language code and the second the Unicode text in the language specified by the code. The same information may be conveyed in multiple languages by repetition of data within an element as several LangStrings. Vocabulary datatype: where the LOM data schema requires an element to be described by a controlled vocabulary the element will be of the vocabulary datatype. Such elements are composed of Source-Value pairs; the source should contain the name of the list of terms being used and the value should contain the chosen term. DateTime and Duration datatypes: these datatypes allow a date or period of time to be given in a machine-readable format (the value space is based on the ISO 8601:2000 standard; an example of a correctly formatted date is 2003-11-22); a human-readable description may be provided instead of or in addition to the formatted date (e.g. "late 20th century"). While implementing the LOM for the development of our learning scenarios we did not select all the elements in the conceptual data schema. The creation of of our own customized metadata schemes based on LOM allowed us to specify which elements and vocabularies we will be using mostly. While we discarded some elements from the LOM we also supplemented some of the LOM vocabularies with values that are appropriate to the teaching and learning community that we wish to support. The LOM has been widely implemented by repositories and other learning resource providers, partly as a result of its status as an international standard, and partly through its association with other influential specifications, such as those produced by the IMS Global Learning and by ADL (SCORM)(no date). Examples of repositories and initiatives that have adopted the LOM are the JORUM (no date), a JISC funded repository of teaching and learning materials for UK Further and Higher Education; the European Ariadne foundation (Ariadne, no date); various European SchoolNet projects (European SchoolNet, no date); the Global Learning Objects Brokered Exchange (GLOBE, no date) federation; and many more.

The LLL3D metadata scheme
While the influence of the LOM has been considerable in terms of the development of our learning scenarios as it has formed the basis for resource, problematic issues have been noted. To begin with, the LOM conceptual data schema (the stated aim of which is to "ensure that bindings of learning object metadata (LOM) have a high degree of semantic interoperability" IEEE, 2002, section 1.2) is not based on an abstract model shared with other metadata schema. Essentially it is impossible to import elements from other metadata schema, such as Dublin Core or schema developed to support specific resource types such as images or specific features about these learning scenarios such as rights management or preservation. This is especially problematic since it is necessary for the LOM to accommodate general and non-educational characteristics (e.g. technical, rights, accessibility, etc) within the standard data schema rather than importing solutions from other domains. We believe that IEEE LOM is mostly suitable in general for defining complex learning processes; nevertheless personalization capabilities are clearly insufficient for describing the complex requirements of ach learning scenario. Although LOM can be used for describing the learning scenarios, the description of the elements is not a simple process. Seven main levels of description can be identified. Each learning scenario is described using the following categories (See Appendix A): General: This category identifies the general information that describes this case as a whole. It poinst towards features such as description, target audience, key words and source materials about the case study.
Life Cycle: This category describes the history and current state of the case and includes information about the status and date of the case study.

Meta-Metadata:
This category describes the metadata record itself rather than the case that this record describes. This category describes how the metadata instance can be identified, how, when and with what references. This category is needed to ensure reutilization of learning resources in different contexts. The LOM standard defines a structure for interoperable descriptions of learning scenarios. Metadata for a learning scenario describes relevant characteristics of such scenarios to which applies, pursuing reusability.
Educational: This category describes the key educational or pedagogic characteristics of this case. It describes the different interaction levels and types between users.
Rights: This category describes the intellectual property rights and conditions of use for this case.
Classification: This category describes where this case falls within a particular classification system. Other: This section provides the information required for a case to be completed. Any extension to include a more comprehensive description of terms is included in the 'other' category. While LOM supplies the required information that can allow to build up a learning scenario, this extended version that we built up for the LLL3D project offers richer structures and takes into account other important features of a case study such as interests, success, failures and reflections. Furthermore, the inclusion of the category such as 'Other' into our model for educational metadata highlights the importance of some types of metadata that have, perhaps, been regarded as secondary metadata in the past, but which increasingly appear to be of primary importance to education. Indeed, it seems to follow from the working definition of learning materials as "anything used for teaching and learning" that the defining educational characteristics pertain not to the material itself but to the use of that material. We think that the relevant metadata schemas are those describing also the success and failure elements of a particular learning scenario. We believe that such an extended metadata scheme does not only make a scenario easier to read and use, but also is able to incorporate new services and functionalities when required. For instance, Web 2.0 tools and other collaborative tools can be embedded within the learning scenarios (See Appendix A).

Future challenges
As Barker (2008) claimed, metadata requirements for educational resource types and purposes are not well understood and are less well articulated. Pinning down the details of which educationally significant characteristics pertain to which entities in learning scenarios and which relationships are important is a crucial step in understanding what information is needed to create resource descriptions that meet educational requirements, and how to go about gathering that information. For example, information about how a resource is used in a particular scenario,such as what course is it used for and the subject and educational level of that course, may be gathered by course management systems such as VLEs or MLEs, but this information is rarely, if ever, passed back to the system that manages the resource descriptions, i.e. the repository or catalogue. The rationale for the latter example is that the quality of the search can be enhanced by aggregating the contents of several repositories; hence the service offered, can be enhanced www.intechopen.com by aggregating information about usage from several systems. As well as being distributed across many systems it is highly likely that the metadata will be heterogeneous: different systems will record different metadata and make it available in different formats. The concepts of the semantic web may be useful in dealing with such distributed heterogeneous metadata but this has yet to have much impact in practice, particularly in the educational domain. Another observation made during the LLL3D project on metadata requirements for virtual world learning scenarios is that when precise metadata requirements are not well articulated for a particular scenario it is often common practice to provide descriptions in the form of free text. The original rationale for creating structured metadata was to record resource descriptions that were machine readable without some form of computational semantic analysis of free text. Key to this requirement is the assumption that a computer will be taking action on the basis of information conveyed in a resource description (for example selecting an appropriate resource for a given scenario) rather than a human taking this action. However, it is quite possible that in many cases it may be sufficient to find a description of the right thing (in terms of an entity or relationship in a scenario) and to present this in human readable form to the user who can then take action. This reduces the role of metadata to the well-understood role of supporting resource discovery, i.e. allowing the user to find the human readable description. Also, we think that highly relevant to situations when precise metadata requirements are not widely agreed are approaches such as (social) tagging and folksonomies. These allow users, or groups of users, to apply descriptive keywords to resources without worrying about the details of the precise relationship between the concept expressed by the keyword and the resource. The users also do not necessarily have to agree with others about what term should be used to express the concept,though many of the systems that implement tagging approaches also include mechanisms for identifying commonly used tags for each resource, which can be useful in identifying any emerging consensus about which terms are appropriate.

Closing observations
Virtual worlds are becoming true learning scenarios for both blended and pure virtual distance education. Any learning scenario pertaining to virtual worlds should ensire a proper development for each learner, taking into account possible learning activities along with their success and failure aspects. As activities rather than content are crucial for interactive learning in virtual worlds, traditional metadata specifications need to be rethought in order to incorporate this vision. In many ways the IEEE LOM standard appears to be based on a coherent record describing all aspects of a "learning scenario" and its use, complying with a single standard. We envisaged the description of any characteristics in the learning scenarios related to virtual worlds that was not already included in the LOM conceptual data schema as being achieved by extending that schema. By adopting an approach of "mixing and matching" metadata schema we can move away from this single schema approach and towards one where metadata from different schema can be mixed if they are based on a unifying abstract model.