(7th in my logic of competence series)
In my previous post, I explained how SKOS relationships can be used to represent the basics of competence structures. But in one of the examples cited, the QAA Subject Benchmark Statement for honours level agriculture related studies, the aspect of level of attainment is present, and this is not easily covered by the SKOS broader and narrower relations just by themselves. Let me explain in some more detail.
In this particular Subject Benchmark, the skills, knowledge and understanding are described at three levels: "threshold", "typical", and "excellent". As a first example, in one of the generic skills, (communication skills), under "threshold" one item reads "make contributions to group discussions"; under "typical" the corresponding item reads "contribute coherently to group discussions"; and under "excellent" it reads "contribute constructively to group discussions". Or take an example from the "subject specific knowledge and understanding in agriculture and horticulture" — threshold: "demonstrate some understanding of the scientific factors affecting production"; typical: "demonstrate understanding of the scientific factors limiting production"; excellent: "demonstrate understanding of the scientific factors limiting production and their interactions". Leaving aside difficulties in clarifying and assessing exactly what these mean, it is clear that there is a level structure, as illustrated in my earlier post. In both cases, the three descriptions are neither identical nor unrelated -- higher levels encompass lower ones. (But note also that benchmark statements in different subjects have different structures.)
Can one represent these attainment levels in a tree structure? One option might be to have three benchmark statements presented separately, one each for threshold, typical and excellent. However this would miss the obvious connections between the elements within each level. A more helpful approach might be to describe the common headings with the finest reasonable granularity, and then distinguish the descriptors for different attainment levels at this granularity. This would need a slight restructuring of this statement, because finer-grained common headings are possible than the ones given. For instance, "subject specific knowledge and understanding in agriculture and horticulture" could easily be subdivided into something like these, (using words that appear in each level):
- "science and management of sustainable production systems"
- "social, economic, legal, scientific and technological principles underlying the business management of farm or horticultural enterprises"
- "range of concepts, theories and methods drawn from the constituent disciplines"
At a still finer level, the descriptors mostly share many words, with just the detail differing to reflect the different levels, as exemplified above. In the example above, the common wording is "understanding of the scientific factors affecting production". Headings could be created from common wording. Then there is still the issue of relating the three described levels into the structure as a whole. Threshold, typical, and excellent are not three components of one higher level entity, they are different levels of the same entity. These levels are one kind of variant.
Variants more generally are not always easy to see in common definitions, perhaps because part of the point of having standards is to reduce variability. For a clearer example from a broader perspective, we may consider areas not documented by occupational or educational standards. Consider skill and competence at management. The literature suggests several distinct styles of management: autocratic, democratic, laissez-faire, paternalistic, etc. It is probably obvious that to be an effective manager, one does not have to be able to manage according to all these styles. Perhaps just one may be good enough for any particular management position, though different ones may be needed in different contexts. Having chosen a management style, each will have a different range of component skills. If one wished to create a tree structure to represent management competences, what would the relationship be between a reasonable topmost node, perhaps called just "management", and the four or more styles? It is rather similar to the issue with the levels we saw above, but at a different granularity. As another alternative example, look at the broader issue of developing competence in agriculture or horticulture. Probably no one is an expert in growing everything. Anyone wanting to be a farmer or grower will at some point need to decide what to specialise in, if not in academic study, then at least in terms of practical experience and expertise. There are clear choices, and the range of skills and competence needed for different specialisms will of course differ. Being a competent farmer does not mean being competent at growing all crops in the world. You have to choose.
The basic structures mentioned in my previous post start out with the idea of "broader" and "narrower" concepts. It is reasonable to say that management competence in general is a broader concept than competence as a democratic manager? Or can one say that graduate level competence in agriculture is a broader concept than being assessed as threshold, typical or excellent? Does it really help to say simply that horticulture is a broader concept than growing grapes?
What seems to emerge on thinking this through is that there are at least two kinds of "broader" (and equally two kinds of "narrower") with different logic. One type is like whole-part relationships. We saw this in the National Occupational Standards units, which were composed of things that a person needs to be able to do, alongside things that the person needs to know. In principle all parts are needed to constitute the whole. If we imagine say a personal development or learning tracking system that helps you with your learning, and you were working towards the unit of competence, then the system could keep track of which ones you say you have done, and perhaps remind you to complete the remaining ones.
On the other hand, the other type of relationship (illustrated above) is "style" or "variant" rather than "part". If we imagine a system to help with professional development, and you wanted to develop your management skill, it is at least plausible that you could be asked at the outset which style of management you would like to improve your skill in. Having chosen one (or more) the rest would be put aside. You would work towards the constituent knowledge and skills for the chosen ones, and the system would not bother you with the knowledge and skills needed for the styles you had chosen not to learn more about. Similarly, a general horticulture skill aid would have to start by getting you to select the kind of crops you wanted to grow. And for the other example, with the attainment standards of the Subject Benchmark, we can imagine selecting a topic and then being asked what level you believe you have attained on this topic, so again there is a selection process instead of simply the combination of parts.
One could indeed imagine all of these features together in a tool that helped with personal development. The system could ask you what level you believe you have attained already, and what level you are working towards, for fine grained knowledge and skills, and then reminding you to work at the identified gaps. At the same time, which fine grained areas you work at will depend on your more course-grained choices, like which styles of the competence you want to acquire, and which options you will specialise in.
It may help to compare these two kinds of relationship with ones that are very common elsewhere. UML distinguishes various relationships within class diagrams by graphical symbols, and two of the most common are called "composition" and "generalization". Composition is very close to the kind of basic relationship in competence where component skills and knowledge are required to make up a wider competence, or that various competences are required to qualify as a certain grade of professional. On the other hand, the broad concept of management competence could be seen as a generalisation of the more specific competences in various styles of management. A word of caution, however: UML is designed specifically for use in systems analysis and design, or software engineering, so it should not be suprising if the match with representing competence is not exact.
Even though the two kinds of relationship I have been talking about are well known in many fields, SKOS does not make an explicit distinction between them. Logic seems to lead to the idea (which I have heard SKOS experts suggesting) that it is up to others to define more specific relationships than (specialisations of) SKOS's "broader" and "narrower" to represent these two kinds of relationship. We don't want to deprive SKOS of the right to be called "Simple".
However we represent these two kinds of relationship, if we are going to represent them in a way which is useful for tools to help people manage their competence, their learning towards competence, and their self-assessment of competence (perhaps leading to external assessment) then it does seem entirely appropriate to represent them differently. Very simply, there are times when you need all of a set of components, and there are times when you need to choose which of a group of options you are going to choose: "and" and "or"; and both kinds of relationship are of great practical use.
Addition, 2011-06-22 On the other hand, people seem to easily mix compulsory and optional parts in the same structure. This is extremely widespread in the definition of qualifications, which are still a very important proxy for or indicator of abilities and competence. So, rather than needing necessarily to separate out the two kinds of structural relationship, we can simply be liberal about accepting whatever combinations people want to represent. If a certain ability has both necessary and optional parts, it is still very easy to understand what that means in practice, and to follow through the implications.
2011-07-04 And I give detailed argument to the reasons for optionality in post 15 in this series.
That may be a good place to stop for defining generic structure for single framework structures of skills or competence. But what I have not covered so far is relationships between different competence structures. One thing this is needed for is reuse of common elements between definitions...