One of Simon Grant's publications, substantially presented at the TEN Competence workshop, Manchester, 11th - 12th January 2007
Appears in Service Oriented Approaches and Lifelong Competence Development Infrastructures: Proceedings of the 2nd TENCompetence Open Workshop, pages 167-173.
Simon Grant 2007-03-09
Independent Consultant, working with the JISC CETIS Portfolio SIG.
Abstract: This paper outlines a distributed top-level approach to representing relationships between competence-related definitions, when held by several parties. The suggested approach is to use existing specifications to represent the competence definitions themselves, and Semantic Web technology to represent the relationships between these competences. Five concepts are suggested for these relationships: equivalence; non-equivalence; satisfaction; contribution; and overlap. This will allow much greater automatic processing of competence-related information while avoiding the pitfalls of a centralised approach. It is stressed that theoretical perfection is here less important than laying the foundations of a practical solution acceptable to existing competence authorities as well as developers.
Keywords: competence, interoperability, skills frameworks, distributed architecture.
Competence-related definitions can be used in several ways, and in several contexts. In the field of education and training, they can be used in the definition of the outcomes of a course, and for setting out the requirements for entry into a course or educational programme. Related to employment, individuals can define their skills and competences not only in their own terms but also in terms that are recognised as being in common use; employers can specify the requirements for a post in terms of competences. Professional and workforce development can be carried out referring to common definitions of the competences that are required in the workplace.
Definitions that are used in these ways could be regarded as common in two ways. Firstly, experienced people could recognise common terms that are in current use: this would rely on human judgement and appreciation of the terms used, but would not be affected by slight variations. Synonyms would be most likely recognised by the people concerned. Secondly, the terms could have identifiers that were defined as meaning the same thing: this could be either because only one identifier is used for one term, or through the explicit mapping of equivalent identifier. This second approach is necessary if there is to be any reliable automatic processing involving the competence definitions.
The first approach is what has happened in most cases for a long time. The problems are relatively well-known. In contrast, the challenge taken up in this paper is to point towards a workable system using the second approach.
As Grant (2006), here the term “competence” is used inclusively to cover any of a range of closely-related terms. This is also the pragmatic approach taken by the websites of HR-XML and IMS RDCEO, referred to below, but using the term “competency”. The distinctions, though significant in some contexts, are not considered significant in this work.
In practice, there are several ways in which the same competence-related definitions may be wanted to be used across different systems and contexts. Within education and training, the outcomes of one course delivered by one institution or body may relate, in principle, to the entry requirements for another course given by a different institution or body. This scenario has been investigated in the XCRI (eXchanging Course-Related Information) project, where much related work has been done. The challenge is that two different institutions, without consultation, are likely to come up with different characterisations of their competence-related definitions. Without some kind of harmonisation, it would be difficult if not impossible automatically to match outcomes from one course with entry requirements of another.
There is still more of a challenge around relating such definitions between education and employment, as there is even more likelihood of definitions differing between educators and employers. On the other hand, training bodies are more likely than, say, higher education institutions to relate their courses to competence requirements in terms set out by employers, who often pay for training employees.
There are several possible responses to these requirements.
The default approach is to do nothing to aid automatic processing. Arguments for this approach could include the idea that human processing is all that is needed, and thus that there is little value in enabling automatic processing of competence-related definitions.
Another approach which is readily brought up involves some kind of central registry for competence-related definitions. Many people are sceptical about the potential for success with this approach, as it would appear to be extremely difficult to achieve agreement between all the interested parties. The natural evolution over time of the concepts and terms pose a further difficulty for this approach.
Logically, the third possible approach is to coordinate definitions in a distributed manner. That is, without having any central authority, develop an approach to relating together such definitions, so that there can be automatic processing and recognition of how variant definitions relate together.
Compare XFN, the Xhtml Friends Network. This, like FOAF ("friend of a friend"), is intended to be a way of exposing social networks without having any centrally held information about the social relationships involved. It attempts to represent the way that people relate to one another, in broad terms. Though many people might not participate in XFN or FOAF due to privacy concerns, no such concerns will attach to taking a similar approach to competence-related definitions.
This distributed approach is the one which is investigated in the current paper.
Again, FOAF and XFN provide a useful parallel. What should one include in the specification for the information which is to be held in a distributed manner? The only relationship included in the basic FOAF specification is "knows", though other relationships have been proposed as extensions. XFN, on the other hand, has 18 different possibilities for its "rel" attribute, ways of being related as humans. Many more are possible.
Just as the success of FOAF or XFN depends on people publishing information about themselves and their friendships, the suggested approach to competence-related definitions is based on the assumption that they are published, in a way that allows different people and systems to refer effectively to the same competence-related definition. The premise, here accepted, of being able to do this by using a URI (or more generally IRI) is the subject of much debate which will not be reproduced here (see, e.g., Pepper & Schwab (2003)).
Here, five relationships between competency-related definitions are suggested for consideration. This is a small vocabulary, which would be easily manageable by web-based and knowledge-based systems.
The simplest possibility is for each competence-related definition to be associated with some information stating which other similar definitions, held by other authorities, are accepted by the first definition owner as equivalent.
If this were carried out, it would mean that if two systems are used with the intention of referring to the same competence concept, using different identifiers, there could be automatic checking of whether either or both of the authorities maintain that the two concepts are equivalent.
In OWL (the Web Ontology Language) there is a relationship “sameAs” which would appear to be suited for this purpose.
2007-05 Note: my opinion has changed: owl:sameAs should be reserved for
different representations of identical information. We need another, new, predicate to represent
operational equivalence. The same applies to owl:differentFrom below.
2008-11 Note: now that SKOS is nearly established, it makes sense to consider the SKOS mapping properties. skos:exactMatch or skos:closeMatch may do here; other of the SKOS mapping properties may be seen as similar to Satisfaction Contribution and Overlap.
Simply allowing the noting of equivalence in such definitions does not help when dealing with spurious claims of equivalence. For instance, a pretentious training supplier may claim that a learning outcome for their (short, cheap) course is equivalent to a learning outcome claimed by another (longer, more expensive and better-established) course. Worse examples of misrepresentation can easily be imagined.
Allowing declarations of non-equivalence would enable clarity in this respect: although one party claimed equivalence, the other party could explicitly reject that claim, for the avoidance of doubt.
Again, OWL has a relationship “differentFrom” which seems to fit the bill.
Entry requirements or specifications can be used either for educational or training opportunities, or for person specifications for employment. These requirements or specifications can in principle be represented in the same way as other competence-related definitions. But in this case, one obviously useful facility would be to be able to indicate which other attainments would also satisfy this particular entry requirement.
Thus, for example, a person specification for a certain job may require a mathematics qualification at school secondary level. Alongside that definition, it could be indicated that a degree-level qualification in mathematics, or physical sciences, would also satisfy that requirement.
A satisfaction relationship would be transitive. The fact of being a transitive relationship could be represented formally in OWL, but OWL does not appear to have predefined relationships in this or the following cases.
Looking from the other end, when devising course modules, educators can be asked to frame the intended learning outcomes from that module. Perhaps the course, overall, is intended to relate to a broader, discipline-wide set of competences. In this case, the educator building the module might want to indicate the intention that a learning outcomes from the module is intended to contribute towards a greater outcome at the level of the course as a whole, or the educational programme as a whole.
An alternative way of achieving a similar result would be to detail the broader programme-related outcomes, creating a framework of skills or competence (Grant, 2006), whereby skills or competences that are regarded as greater or of broader scope were related to the ones regarded as lesser or of narrower scope. A module developer might then be asked to match their module outcomes to definitions in that framework, rather than writing their own.
This alternative approach relies on the representation of the same kind of relationship - that of narrower or lesser skills or competences contributing towards broader or greater ones - but this time within the framework itself, rather than devolving the responsibility for relating greater to lesser onto a module developer.
"Credit" frameworks could also use a similar approach. A greater objective could be defined as being satisfied by a certain number of credits (within a specified credit scheme), and lesser courses or modules could "contribute" defined values within that credit scheme. Not much extra architecture would be necessary to implement this kind of framework.
Contribution and satisfaction are not necessarily inverse relationships in all cases. Communication competence may be held to contribute to ICT (Information and Communication Technology) competence, and indeed vice versa, but neither would satisfy a competence definition of the other.
It may also be useful to note overlap between two competence definitions. This would be little machine processing enabled by the noting of this relationship, except as a kind of “see also” link, but it would serve to alert people of an incomplete mapping. Any cases where overlap is noted could in principle be broken down to finer-grained units where the other relationships can be applied.
An issue which has not been dealt with here is the application of levels to competence. Very often, when people define skill or competence, they are inclined to define an area of competence and a related set of levels of that competence that are relevant to distinguish for their purposes. If levels are not treated explicitly, competence-related definitions are constrained to be free standing, as discussed above: relationships of satisfaction and contribution have to be defined explicitly.
The HR-XML competency specification is one where level can be represented. Introducing levels explicitly could be regarded as a further step in this process, to be undertaken later.
If such a distributed approach is to be implemented, decisions have to be taken on how to represent the chosen information for interoperability purposes. There are in principle many options for holding competence-related definitions, including HR-XML mentioned above, and the IEEE Reusable Competency Definitions (RCD) specification, which is based on the older IMS RDCEO (Reusable Definition of Competence or Educational Objective) specification.
Relationship information could be represented within the competency definition files themselves. The advantage of this would be that during checking no extra files would have to be fetched apart from the RCD files themselves, which would, or would not, refer to each other. The disadvantage would be in having to mix specifications, and the consequent complexity of the file format.
An alternative approach would be hold all the relationships in a separate file. Each definition would have a single extra element pointing to the relevant file of relationships. The file of relationships could be held using a general-purpose knowledge representation format such as XTM (XML Topic Maps) or RDF (Resource Description Framework) The definition URIs could be used in the relationships file as identifiers. An XTM file in this situation would be largely self-contained, and it would remain to specify the constraints on that file for it to qualify as a proper instance of a competence-related relationships file. For an RDF file in a similar role, there may be the need to have associated ontology information (e.g. OWL) and to specify the possible relationships.
Typically, some relationships would be internal to a relationships file. Thus, an examinations board could maintain firstly one file for each qualification, and secondly a relationships file where they noted at least which qualifications could be considered to be implied by higher-level qualifications. A higher-level qualification could "satisfy" a lower-level one, while the lower-level one could "contribute" to the higher-level one.
Other relationships could connect local definitions to definitions maintained by other authorities. This is when the relationships of equivalence and non-equivalence will be most useful: there is little point in maintaining two local equivalent definitions.
This approach would appear to be more pragmatically feasible than alternative approaches. People seem unlikely to use centralised registries, or even to use other authorities' definitions of competences. While there is no documented evidence of this, such pragmatic considerations are vital to the effective implementation of these architectures.
The provision of the equivalence and other relationships allows people to build their own competence frameworks, and then still to relate them to other ones. Existing frameworks can thus be related with other ones with a minimum of extra effort.
If such an approach can be established, and if relevant people can be persuaded to represent their related information in these ways, many possibilities open up.
The XCRI work could be effectively extended to make chaining together course outcomes and entry requirements more possible.
More generally, the approach would open up the possibility of more usefully representing competence in personal electronic portfolios, which could then be used as part of the process of application for opportunities, whether in education or employment.
This short paper extends only to exploring the overall top-level design of a system for competence-related interoperability. Future work will extend to describing a more detailed specification of this, along with indications of possible "bindings" to RDF and/or XTM, and how these would fit in with existing specifications in this field.
Further work remains to be done to integrate levels into this approach: however it might be wise first to see whether a basic approach without levels is likely both to work and to be popular enough to be widely adopted.
Some of the concepts underlying this paper were developed while undertaking work for the XCRI project – the relevant XCRI Competence Report is available on the XCRI web site. The author has written this paper partly as a contribution toward his work for JISC CETIS.
Grant, S. (2006) Frameworks of competence: common or specific? Proceedings of International Workshop in Learning Networks for Lifelong Competence Development, TENCompetence Conference. September 12th, Sofia, Bulgaria: TENCompetence. Retrieved November 2006, from http://dspace.learningnetworks.org/handle/1820/746
Pepper, S. & Schwab, S. (2003) Curing the Web's Identity Crisis. Retrieved November 2006 from http://www.ontopia.net/topicmaps/materials/identitycrisis.html
FOAF: http://en.wikipedia.org/wiki/FOAF; also http://rdfweb.org/, http://xmlns.com/foaf/0.1/ and http://www.foaf-project.org/
IEEE RCD: http://ltsc.ieee.org/wg20/ is the site of the responsible working group. The specification itself is not freely available.
IMS RDCEO: http://www.imsglobal.org/competencies/
XCRI: http://www.elframework.org/projects/xcri. http://xcri.org/
Copyright © 2006.