Europe focuses on EML
Delegates from around Europe converged on Turin, northern Italy last week to discuss educational modelling languages (EMLs): technologies that attempt to model the learning process and not just learning content.
The EML Workshop in Turin formed part of an effort by CEN/ISSS to conduct a survey of current EML approaches; a process that may culminate in an EML standard for Europe. Such a move would bring EMLs – until recently on the fringes of learning technology specifications – to the heart of the learning technology field.
One reason to be interested in EMLs is the element of process. Process is one of the things that differentiates EMLs from other learning specifications; in an EML, an educational "unit of study" may not involve any content at all, but may instead consist of a series of activities, perhaps involving conversations between learners and a tutor. Or content may be the outcome of the unit; for example, students working together to create a report after conducting research.
According to Koper, this distinction between learning content and learning process is at the heart of the EML philosophy, "Learning is different from consuming content [...] learning is active; students who are not active don't learn anything".
Using the EML approach, the learning content is a commodity used by learners and tutors while performing a learning activity; the focus of learning is the activity, not the content.
Learning methods – none, one, or many?
Another fundamental for EMLs is pedagogy - the methods used for teaching and learning - and the "teaching objects" in a course, such as study tasks, objectives, prerequisites and so on.
There are three options for any learning technology when it comes to pedagogy:
Most current tools and technologies aim to be "pedagogy neutral". However, according to Koper this makes them "semantically poor solutions" – and as CETIS' Bill Olivier points out, "by not supporting a single pedagogy, you may in fact preclude some pedagogies".
- You can support no pedagogy at all ("pedagogy-neutral")
- You can support a single pedagogy and enforce it as the standard
- You can support multiple pedagogies
For example, units of study may consist entirely of activities without any predefined content. Such a course consists entirely of "teaching objects" and would be difficult if not impossible to work with in a "pedagogy-neutral" system.
Koper also believes the single-pedagogy approach is an untenable position for an EML, "there are hundreds of different pedagogical models; implementing one model is not the right direction for general use & standardisation."
Instead, Koper favours using a "pedagogical meta-model" that allows templates to be created for different methods. With this approach, the key is finding a way to express any pedagogy at a sufficiently high level as to allow a diversity of approaches, while retaining usability.
Not everyone agrees, however, as Eddy Forte of the Swiss Federal Institute of Technology asserts: "there is always an underlying pedagogical model for any EML" regardless of attempts at abstraction or meta-modelling. So "a standard must remain pedagogically neutral as far as possible."
Pedagogic neutrality is an essential part of the learning object reusability model - if an object is already purposed and contextualised, how can you reuse it for another purpose or in another context?
One approach is to provide a pedagogically driven high-level description of a learning process that makes use of pedagogically neutral learning objects. The issue then shifts to one of granularity rather than fundamental approach.
So why EMLs?
The real driving force behind EMLs is a belief that current content-driven approaches are too reductionist and just aren't specific enough to education ("Content Packaging could package elephants or ferraris" as one speaker put it).
By reducing study to learning objects and ignoring the teaching and learning methods, the argument is that current approaches will never achieve real interoperability.
"EMLs contain all the logic and content of a learning unit, so it allows for real interoperability and collaboration between institutions and industry" explains Koper, "the logic of interpreting the content is not fully specified. WebCT can't interpret Blackboard and vice versa [...] for real interoperability we need an expression of the complete course regardless of the tools used to make or run the course."
Atoms and molecules
Looked at in this way, EMLs and learning objects represent opposite approaches to the interoperability question. For the learning objects approach, interoperability for content is hampered by the contextual "baggage" of the object – only by stripping content of all its outside referents such as teaching methods, student roles, and activities can an object be truly reusable. EMLs, on the other hand, take the approach that if you make the units larger – whole lessons, practicals, or courses – and include all the contextual information, then you have large, useful components that are truly shareable and reusable.
The argument is perhaps analogous to the one that dominated reusability in computer programming in the early years of Object-Oriented Programming. The original notion was that you could re-use individual Objects (Classes) in other applications; however, it became apparent that reuse could also be achieved by grouping objects into Components, together with a functional context for use. Reusing components made building new applications from existing parts much easier than by reusing objects.
The relative complexity of an EML unit of study concerns Forte, "[the] real practitioners [in this area] are teachers and they will only accept a simple basic model". According to Forte, teachers are not themselves experts in pedagogy, with a largely empirical training. "They need something simple, that they should be able to customise if they want." On a more positive note, Forte observes that two of the EML approaches described at the workshop understand this need for customisation by practitioners.
The CEN/ISSS EML survey
Given the degree of interest that EMLs are generating, its timely of CEN/ISSS, the European standardisation body concerned with IT, to undertake a survey of the EMLs out there so we can examine the different approaches.
There are a number of EML-like projects around; some better supported than others. At Turin, the workshop concentrated on two key projects:
- PALO, produced by UNED Madrid
- EML, produced by the Open University of the Netherlands
There was also time to look at some "non-EML" but interesting projects, including
- TargetTeam, produced by the University of the Armed Forces, Germany,
- ARIADNE's Course Description Format (CDF), and
- the 3DE Personalized Learning system produced as part of the European Commission's 5th Framework programme
Other systems, such as LMML and TML, are also being considered as part of the EML Survey, but were not covered in depth by the workshop. There may be 20 or 30 (or more) EMLs "out there", most of them domain-specific, and the survey will try to identify and describe as many as possible.
If you know of any examples, you can send them to Adrian Rawlings (Adrian.Rawlings@ou.nl). A report will be presented at the CEN/ISSS meeting in Berlin on the 17th of November 2001.
You can also find out more on the CEN/ISSS EML Survey website
Below, we take a brief look at some of the EML projects discussed in the workshop.
Miguel Artacho of Spain's National Distance Learning University describes PALO as "a cognitive approach to an educational modelling language". Like the EML of the Open University of the Netherlands, PALO tries to describe larger units of study rather than individual learning objects, and incorporates the instructional methods for the unit.
PALO provides a declaration of the structure of the unit (which can be rendered as a table of contents for print, or as navigation elements for use on the web), the activities students and tutors undertake, and the scheduling of activities and content.
Interestingly, rather than encompassing a set of local or distributed "knowledge objects", the PALO approach instead involves creating a course-specific repository of semantically linked material.
So, for example, you could assemble "concepts", "problems" and "solutions" for the domain of study and then link them together, so that you can provide "problems associated with concept x" or "solutions for problem x" in the unit of study.
Using the PALO approach, the construction of this themed knowledge base is a core aspect of course development.
Another crucial difference of approach is that, whereas EML uses a meta-model approach to explicitly describe the pedagogical approach used, with PALO the pedagogy is implicit in the particular PALO template used.
This carries the risk that you could end up with a proliferation of templates to cover diverse pedagogical methods, and that practitioners could be landed with the task of building DTDs themselves (and, as Eddy Forte points out, teachers and lecturers often receive empirical training without any particular expertise in learning theories).
Of course, this situation could be alleviated if suitable tools were available; currently PALO courses are SGML documents edited in EMACS. "Tools are a real problem" Artacho admits, however there is a current drive at the PALO project to build new tools in Java and migrate the information model from SGML to XML.
PALO is certainly an interesting approach to building learning materials; take a look at the online demonstration to learn more.
Perhaps the most well known Educational Modelling Language - and rather confusingly synonymous with it - is EML, developed by the Open University of the Netherlands (OUNL).
EML's approach to pedagogy is to provide a high-level abstraction of learning methods: roles (such as tutors and students) undertake activities with an environment. As Jocelyn Manderveld of the OUNL explains, "all pedagogies contain this axiom because learning is an active process".
The "environment" can contain learning content, tools, communication, and other elements usable by learners and others in an activity; this is a key difference between an EML and content-centric learning models, and is explicit in the EML schema.
Activities are then structured using a "learning flow" that includes decision-points (so that, for example, performance in one activity determines the next), sequences and choices.
This high level of abstraction and flexibility makes EML a very powerful tool for expressing very different learning scenarios, including personalized learning.
We'll be covering EML in more detail in a future article, in the meantime you can visit the EML website or view EML related articles on this site.
TargeTeam and TeachML
TargeTeam is not actually an EML as such, as it focuses on content and not process in learning. Gunnar Teege, of Germany's University of the Armed Forces, describes TargetTeam as an "author-centric" approach intended to allow content reuse.
However, it does feature a system of semantic relationships between learning objects (called "issues" in TeachML) similar to that employed by PALO: in addition to Dublin-Core style metadata, TeachML "Issues" have specific educational metadata to describe semantics.
For example, an issue can be an argument for (or against) another, an example of another, and so on.
Other relationships include problem, solution, hint, explanation, and proof. Providing these semantics gives some direction for presentation of materials without the need to provide explicit formatting or structuring.
For more information on the TargeTeam system, visit the TargetTeam website