(4th in my logic of competence series)
Specifications, to gain acceptance, have to reflect common usage, at least to a reasonable degree. The reason is not hard to see. If a specification fails to map common usage in an understandable way, people using it will be confused, and could try to represent common usage in unpredictable ways, defeating interoperability. The abstractions that are most important to formalise clearly are thus those in common usage.
It does seem to be very common practice that competence in many fields comes to be described as having levels. The logic of competence levels is very simple: a higher level of competence subsumes -- that is, includes -- lower levels of the same competence. In any field where competence has levels, in principle this allows graded claims, where there may be a career progression from lower to higher level, along with increasing knowledge, practice, and experience. Individuals can claim competence at a level appropriate to them; if a search system represents levels of competence effectively, employers or others seeking competent people will not miss people whose level of competence is greater than the one they give as the minimum.
For example, the Skills Framework for the Information Age (SFIA) is a UK-originated framework for the IT sector, founded in 2003 by a partnership including the British Computer Society. This gives 7 "levels or responsibility", and different roles in the industry are represented at one or more levels. The levels labels are: 1 Follow; 2 Assist; 3 Apply; 4 Enable; 5 Ensure, advise; 6 Initiate, influence; 7 Set strategy, inspire, mobilise. These levels are given fuller general definitions in terms of degress of autonomy, influence, complexity, and business skills. There are around 87 separate skills defined, and for each skill, there is a description of what is expected of this skill at each defined level -- of which there are between 1 and 6.
The European e-Competency Framework (e-CF), on which work began in 2007, was influenced by SFIA, but has just 5 "proficiency levels" simply termed e-1 to e-5. The meaning of each level is given within each e-Competence. There are 36 e-competences, grouped into 5 areas.
The e-CF refers to the cross-subject European Qualifications Framework, which has 8 levels. Level e-1 corresponds to EQF level 3; e-2 to EQF 4 and 5; e-3 to EQF 6; e-4 to EQF 7; and e-5 to EQF 8. However, the relationships between e-CF and SFIA, and between SFIA and EQF, are not as clear cut. The EQF gives descriptors for each of three categories at each level: "Knowledge", "Skills", and "Competence": that is, 24 descriptors in all.
This small selection of well-developed frameworks is enough to show conclusively that there is no universally agreed set of levels. In the absence of such agreement, levels only make sense in terms of the framework that they belong to. All these frameworks give descriptors of what is expected at each level, and the process of assigning a level will essentially be a process of gauging which descriptor best fits a particular person's performance in a relevant setting. While this is not a precise science, the kind of descriptors used suggest that there might be a reasonable degree of agreement between assessors about the level of a particular individual in a particular area.
For comparison, it is worth mentioning some other frameworks. (Here are just two more to broaden the scope of the examples; but there are very many others throughout the professions, and in learning education and training.)
In the UK, the National Health Service has a Knowledge and Skills Framework (NHS KSF) published in 2004. It is quite like the e-CF in structure, in that there are 30 areas of knowledge and skill (called, perhaps confusingly, "dimensions"), and for each "dimension" there are descriptors at four levels, from the lowest 1 to the highest 4. As with all level structures, higher level competence in one particular "dimension" seems to imply coverage of the lower levels, though a level on one "dimension" has no obvious implication about levels in other "dimensions".
A completely different application of levels is seen in the Europass Language Passport. This offers 6 levels for each of 5 linguistic areas, as a way of self-assessing the levels of one's linguistic abilities. The areas are: listening; reading; spoken interaction; spoken production; and writing. The levels are in three groups of two: basic user A1 and A2; independent user B1 and B2; proficient user C1 and C2. At each level, for each area, there is a descriptor of the ability in that area at that level. That is 30 different descriptors. All of this applies equally to any language, so the particular languages do not need to appear in the framework.
Overall, there is a great deal of consistency in the kind of ways in which levels are described and used. Given that they have been in use now for many years, it makes clear sense for any competence structure to take account of levels, by allowing a competence claim, or a requirement, to specify a level as a qualifier to the area of competence, with that level tied to the framework to which it belongs, and where it is defined in terms of a descriptor. This use of level will at least make processing of competence information a little easier.
But beyond level it seems to get harder. The next topic to be covered will be other attributes including conditions or context of competence.