Back in July and August the Talent Marketplace Signaling W3C Community Group made good progress on how to relate JobPostings to Educational and Occupational Credentials (qualifications, if you prefer) and Compentences. These seem to me to be central concepts for linking between the domain of training, education and learning and the domain of talent sourcing, employment and career progression; a common understanding of them would be key to people from one domain understanding signals from the other. I posted a sketch of how I saw these working,.. and that provoked a lot of discussion, some of which led me to evaluate what leads to misunderstandings when trying to discuss such concepts.
This post is my attempt to describe the source those misunderstandings and suggest that we try to avoid them. Finding clarity in talking about competences and credentials is certainly not “all my own work”. Jim Goodell, Alex Jackl and Stuart Sutton and many others in the Talent Signal group and beyond have all been instrumental navigating us through to what I hope will be a common understanding, documentation of which is currently being editted by Alex. This is, however, my own take on some of the factors feeding in to that discussion, I wouldn’t want to laden anyone else with any blame for what’s described below. Also, I have simplified some of the issues raised in the discussion, and for that reason do not want to suggest that they represent the views of specific people. If you want to look at who said what, it’s best you read it in their own words on the email list.
So what are the factors that make talking about competences and credentials difficult?
We might think that we know what a credential/qualification is: the University of Bristol offers a BSc in Physics; I have a BSc in physics from the University of Bristol–I could show you the website describing that qualification and a photo of my certificate. But they are not the same thing, there’s a difference between the abstract credential being offered and the specific certificate that I have, just as the story “Of Mice and Men” is not the same as the thing with the ISBN 978-0582461468, and the thing identified by that ISBN is not the physical copy of the book that I have on my shelf. We’re all familiar with distinguishing between abstract classes (think of Platonic archetypes) and specific instances and we’re pretty good at it, but let’s just acknowledge that it’s difficult. It is difficult to know what level of generality to give to the abstraction (or abstractions, it’s often not a simple as instance and class); it’s difficult to know the words that you can use to make clear which you’re talking about, and it’s easy to talk at cross-purposes by incorrectly assuming that we have made that clear.
Naming things is hard, especially abstract things. One way that we try to deal with this is to refer to things by relation to something more concrete (in our experience): thus we call programmes of study after the credential they lead to (“Jamie is doing an MA in Film Studies”), or we call parts of a course a skill “Phil has completed 20 skills in Duolingo Greek”, or we refer to people after the credential they hold (“Google hires lots of PhDs”). This may help in narrow contexts: if all you ever talk about is programmes, courses and their components it doesn’t matter if you call them after credentials and skills; but when you start doing that when you talk to someone who usually only deals with credentials and skills then you’ll cause confusion.
Another way to deal with naming abstraction is to coin terms that have specific meanings in context. The only problem with this is that in the Talent Signal work the point is that we are trying to talk across contexts, and the odds are that the jargon used is either meaningless out of context or, worse, means something different. (This latter is especially likely if it is a metonym. Seriously, don’t do metonyms.)
As I wrote while I was thinking about this, different technical modelling approaches talk about different things, specifically in RDF we refer to things in the outside world: a description of a person is about that person, the identifier used to say what it is about is the identifier of an instance who is an actual person. When building data objects we might create a class of Person and have instances of that class to describe individual people. So for RDF the instance of Person is out there in the world, for the data object modellers the instance of Person only exists in an information system. This matters if you want to get a copy of the instance. Of course the reality is that what is in the information system is a description of a person, and we have just hit metonymy again.
fallacy of the beard
Bearing all that in mind we can come up with a set of definitions for Compentences and Credentials (the abstract things), Competence and Credential Definitions (descriptions of generic competences and credentials that may exist in information systems or elsewhere), Competence Assertions and Credential Awards (associating instances of competences and credentials with other things).
One use of a credential award is to assert that an individual has acheived a certain competence, so is a credential just a competence assertion? Here we hit some of the issues raised in discussion: it was important to some people that institutions should be able to “offer competences” as distinct from “issuing credentials”. I would rephrase that as make Competence Assertions without there being a Credential Award. This seemed tied to the idea that a credential was related to a complete course or programme whereas a competence was related to a part of the course.
In RDF we make assertions, so is the statement <Phil> <hasAbility> <ShoelaceTying> a competency assertion? If so, what more (if anything) do you need to have a Credential Award? We seems agreed that a Credential is somewhat more substantial and more formal, but the problem with that is that it is a judgement based on graduations along a continuum, not a clear disinction. That’s not to say that credentials and competence assertions are not distinct. I am clean shaven. If don’t shave tomorrow I won’t have beard; if I don’t shave for one more day or another after that I still won’t have one; so at what point would I have a beard? How many days I have to go unshaven before my stubble becomes a beard is impossible to define, but that does not mean that beards don’t exist as distinct category.
I hope that if we acknoweldge and identify the right abstract concepts, recognise that people in different contexts will understand jargon differently and take different approaches to what they consider as important parts of a model, and avoid metonyms then I think we can make ourselves better understood.