An Interoperability Framework

E-Learning has many aspects including learning related, content related, data and metadata, infrastructure, systems and environments, administration, teacher and student communities. Each of these aspects is subject to standardization one way or another and the effective and efficient operation of any educational system is greatly affected by the degree of interoperability reached, i.e. the degree to which components of the system can interoperate. Indeed interoperability is an enabler for sharing pedagogy, infrastructures, tools and resources (digital and non-digital including human resources) as well as administrative data. It has been well documented that interoperability and the use of standards has a number of benefits in terms of lower costs, greater supply, higher accessibility and availability, higher quality, and a shorter delivery time.

Different Layers of Interoperability

Interoperability is the ability of two systems to interoperate. Whereas the term 'system' is in this context often understood as a technical system, it applies to systems in the wider sense, i.e. including all actors in the educational system. Hence, interoperability can, be examined in a semiotic framework [Fri1998, Sta1973] that can help us to understand different aspects of interoperability :

Physical layer:
the physical appearance, the media and amount of contact available

Empirical layer:
the entropy, variety and equivocation encountered

Syntactical layer:
the language, the structure and the logic used

Semantical layer:
the meaning and validity of what is expressed

Pragmatic layer:
the intentions, responsibilities and consequences behind the expressed statements

Social layer:
the interests, beliefs and commitments shared as a result

These layers can be divided into two groups in order to reveal the technical versus the social aspect division. Physics, empirics and syntactics, taken together, constitute a domain where technical and formal methods are adequate. However, semantics plus pragmatics plus the social domain can hardly be explored if those methods are used exclusively and without modification.

The physical and empirical layers are today well covered by achievements in the ICT industry on which any educational system making use of ICT can build. At the syntactic layer we are concerned with the language, structure and logic used in order to have systems, subsystems and modules interoperate. The semantic layer addresses the interoperability of meaning (semantic interoperability); i.e. will information given by one actor in an educational system be understood correctly by another actor. This might involve terminology aspects (homonyms, synonyms, scope) as well as human language aspects. The pragmatic layer is concerned with common intentions such as a common pedagogical goal, and with responsibility aspects such as trust. For example, for any Digital Rights Management system to work a certain amount of trust is always required. Likewise, when an educational institution is issuing a certificate then people will have to trust the validity of the certificate. Finally at the social layer, interoperability is concerned with the compatibility of beliefs and values of different educational systems. Whereas beliefs and values of education in Europe may vary from one country or region to another, in general we can say that they are compatible, grounded in a common European tradition. As such it is not really an issue in the context of this paper.
Therefore this section focuses on the following layers:

the syntactic layer which as part of the technical layers deals with technical interoperability
the semantic layer which deals with semantic interoperability
the pragmatic layer which deals with pragmatic interoperability. For pragmatic interoperability the focus is on infrastructural concerns such as trust, rather then the pedagogical aspects which is outside the scope of this paper.

Technical Interoperability of Services

Complex ICT systems are today often built following a service oriented approach where each of the services knows technically how to interoperate with the other services by means of a well defined interface. The major advantage is that system builders can make use of services from different service providers given that they obey the service interface specifications. Just like printers are interchangeable, given that they have a Centronics or USB plug, a learning object repository could be accessed easily if it has implemented for example the Simple Query Interface (SQI) [CWA2005b]. Service developers from their part can develop their service the way they want as long as they obey the interface specifications. Obviously, the more these interface specifications are shared among service developers, the greater the interoperability.

Semantic Interoperability

Semantic interoperability is achieved to the extent that users of interoperable services give the same or compatible meaning to information exchanged between the services. An objective criterion is hard to establish since the interpretation of data, resulting in meaning, is a very subjective matter. Hence, the same or compatible meaning should be understood in operational terms. For example the term 'special education' might be understood slightly differently by teachers from different countries. However when a teacher is using this term in a search for learning objects and she obtains some results indeed related to 'special education', the results will be accurate enough as far as the term 'special education' is concerned.

Semantic interoperability relates to information being exchanged between services and is achieved through several means. First, it requires a common conceptual model. Standards such as the IEEE LOM and specifications such as various IMS specifications typically make use of a conceptual model or an information model and separate the what from the how; i.e. the conceptual model describes what information is exchanged in terms of concepts, their properties, and relationships between these concepts while a so-called binding expresses how this information is exchanged.

Second, the concept properties may have values that require a common understanding. The values being exchanged are on the lexical level while semantics is at the conceptual level. Semantic interoperability is therefore also concerned with questions such as do different terms (possibly from different languages) express the same concept and does a specific term used by different users induce the same semantics? Therefore in order to achieve a higher degree of semantic interoperability, controlled vocabularies are often used. The term vocabulary is used in this paper in the broad sense, referring to value lists, classifications, taxonomies, glossaries, dictionaries, ontologies, and thesauri [CWA2005a]. The meaning of a term is not only determined by its definition but also by its relationships to other terms such as in a taxonomy or a thesaurus.

Pragmatic Interoperability

Pragmatic interoperability is achieved to the extent that users of interoperable services have compatible intentions, responsibilities and consequences concerning the interoperable services and information exchange. Whereas common intentions are very pertinent, for example in the context of a common pedagogical goal in the context of a European project, this paper will only deal with pragmatic interoperability to the extent that it is essential for the good working of the services envisaged. In this context pragmatic interoperability is especially concerned the fact that all parties involved in providing and using a service also take their responsibility. Examples are:

A provider of educational material is trusted to provide indeed educational material
Persons that provide metadata and other descriptions are trusted to provide correct data including the data on accessibility etc.
LMS and LCMS providers have the responsibility to implement digital rights enforcement correctly
Educational institutions are trusted that the learner information, for example related to a certificate, is correct.

Clearly without pragmatic interoperability, no educational system will work. As can be seen from the examples, some responsibility support can be translated into software, but some others can not. Where responsibility support can be implemented into software, it will inevitably involve semantic and technical interoperability. If responsibilities are assumed without software support, these responsibilities should be made explicit (e.g. as pre- and post conditions to a service) and assumed by the appropriate parties.

Standards and specifications for interoperability

Interoperability relies on agreements and the more these agreements are shared the greater the interoperability. This is where standards and specifications come into play. In this paper we use the term 'standard' for de jure standards, agreed by national bodies. For other written agreements concerning interoperability the term 'specification' is used. This may include so-called industry standards.
Following the reasoning that an educational system consists of a set of interoperable services, standards and specifications in the field of learning technologies relate to the interoperability of these services. The following elements can be distinguished:

The request identifier; for example a string such as 'Synchronous Query' identifies the requested service.

The parameters sent with the request; for example a query statement
The result; for example a set of learning object metadata (LOM)

Service interface specifications can be defined in abstract terms (such as in the e-Framework initiative) and can be bound to a specific expression format such as an application programming interface (API) or a web service description (e.g. in the Web Services Description Language - WSDL). The parameters and results may consist of complex information structures which themselves are subject to standards and specifications. For example the result of a query to a LOM repository may be a result set of LOM instances following the IEEE LOM standard.

Standards for information structures typically consist of a set of assertion containers. For example the IEEE LOM standard v1.0 contains assertions concerning:

General Information
Life Cycle
Meta-metadata
Technical
Educational
Rights
Relation with other material
Annotation
Classification

Learner information could have assertions about

Competency
Demographic Information
Preferences
Accessibility
Performance and Achievements
Plans/Goals/Reflections
Activity
Map of Relationships

Typically a standard or specification will consist of rules on how to express such assertions; i.e. what assertions, by whom, when, etc. For example it will be important to know who made the claim about a performance or achievement.

A specification will have an information model (sometimes called conceptual model) and one or more expression formats called a binding. For example the IEEE LOM has several XML bindings. Sometimes the conceptual model and the expression format are integrated into a single specification; i.e. the conceptual model can be expressed only in a single way.

Each assertion will have an information structure. Dependent on what is asserted it might be more or less elaborated. For example data element 6.3 of the IEEE LOM 'Rights.description' allows a free text while section 9 is fully elaborated in order to express taxonomies. Standards and specifications builders typically will have to choose between a relatively free way of expressing the assertion, or a prescribed format such as a formal language, or something in between where a prescribed top structure is provided but some details are in free format. In addition standards and specification builders will have to choose whether to allow more than one format or on a single one. For example the rights applicable to a learning object might be described in more than one Digital Rights Expression Language. The major criterion of choice will be the extent to which agreement can be reached among stakeholders. Obviously, an agreement on the details and an agreement on a single information structure facilitates interoperability. However this is not always possible.

As such building standards and specifications has also a timing aspect. Whereas an agreement is not possible today an agreement may be possible tomorrow. On the other hand an agreement reached today may be premature and preclude more interesting and better future specifications.

From specification to implementation

<<Here comes the adoption model, adapted from CETIS' adoption model>>

References

CWA2005a CEN Workshop Agreement 15453, Harmonisation of vocabularies for eLearning, Brussels, November 2005
CWA2005b CEN Workshop Agreement 15454, A Simple Query Interface Specification for Learning Repositories, Brussels, November 2005
Fri1998 E. Falkenberg, F. J. M. Van Assche, et al. A Framework for information system concepts. IFIP, 1998. ISBN 3-901882-01-4
Sta1973 R.K. Stamper: Information in Business and Administrative Systems, Batsford, London, 1973, and Wiley, New York, 1973