I first discovered the notion of learning object granularity when I was tasked with creating my first learning object. I was using an authoring tool that allowed you to describe (or tag) absolutely anything. This was a revelation! My tool allowed me to assign descriptions to individual photographs and sets of navigable pages that could contain any type of digital media you could imagine. You could also assign descriptions to an entire learning object. Not only was I struggling with how to use the fields (title, description, keywords) that I had to complete, it was also difficult to know where I should stop!
Terms of reference
There are a significant number of terms here that beg further explanation. The idea of a learning object (wikipedia) is one that is slippery: it varies depending upon who you speak to. I see a learning object as one or more digital resources that have the potential to provide useful information to, or serve a useful function for, a consumer. Consumers, I argue, are both the end-users (learners), and those who might use learning objects within a course of study.
An alternative definition might be: a set of learning resources that can be used together to help a learner achieve a defined set of learning objectives. I think I prefer this second definition. It feels a little more precise, but there are few words that allude to how large a learning object might be.
Benefits of learning objects
One of the often cited benefits of learning objects is that they have the potential to be reused. A digital resource could be taken from one learning situation could be reused (or repurposed) to another situation. The benefits could include an increase in the quality of the resulting material and possible savings in terms of time and money.
Learning objects are sometimes held within mysterious instruments called repositories. If existing materials are taken and modified, they could then be later returned to a repository and placed back into circulation for other people can use and modify, thus creating a virtuous cycle. One problem with placing material in a repository is that if your repository contains tens of thousands of individual objects, finding what you want (to solve your particular teaching need) can become difficult (as well as tedious).
Metadata (wikipedia) has the ability to 'augment’ textual searching, potentially increasing the quality of search results. Metadata also has the ability to offer you additional information or guidance about what an object might contain and how it might have been used, allowing you to make judgements regarding its applicability in your own teaching context.
There is a paradox: the more granular (or mutable) a learning object is, the more easily it can be reused, but the less useful it is likely to be. The larger a learning object is, the more useful it is to an individual user (since it may attempt to satisfy a set of learning objectives), and the less likely it could be transferred or 'repurposed' to different learning and teaching contexts or situations. Furthermore, the smaller the learning object, the more moral fibre one needs to successfully create correct (and relevant) metadata.
'Repurposing' is a funny word. I understand it to mean that you take something that already exists and modify it so it becomes useful for your own situation. I think repurposing is intrinsically difficult. I don't think it's hard in the sense that it's difficult to change or manipulate different types of digital resources (providing you already have skills to use the tools to effect a change). I think it's hard because of the inherent dependencies that exist within an object. You have to remember to take care of all those little details.
I consider repurposing akin to writing an essay. To write a really good essay you have to first understand the material, secondly understand the question that you are writing about, and then finally, understand who you are writing it for. If you write an essay that consists of paragraphs which have been composed in such a way that they could be used in other essays, I sense you will end up with an essay that is somewhat unsatisfactory (and rather frustrating to read).
There is something else that can make learning object repurposing difficult. Learning objects are often built with authoring tools. Some tools begin with a source document and then spit out a learning object at the other end. The resulting object may (or may not) contain the source from which it was created. This is considered to be 'destructive' (or one way) 'authoring', where the resulting material is difficult to modify.
Even if we accept that reuse is difficult, there are other reasons why it is not readily performed. One reason is that there is no real sense of prestige in using other people materials (but you might get some credit if you find something that is particularly spectacular!). Essentially, employers don't pay people to re-purpose learning materials; they pay people to convey useful and often difficult ideas of learners in a way that is understandable. There is no reward structure or incentive to reuse existing material or build material that can be easily reused. Repurposing takes ingenuity and determination, but within the end result, much of this may be hidden.
There is a final reason why people may like to create and use their own learning resources rather than reuse the work of others. The very act of creating a resource allows one to acquire an intimate understanding of the very subject that you are intending on teaching. Creating digital resources is a creative act. Learning object construction can be constructivism used to prepare for teaching.
Considering metadata granularity
The terms 'aggregate' (or 'composite') and 'atomic' objects are sometimes used when talking about learning objects. An atomic object, quite simply, are objects that cannot be decomposed. An atomic object might well be an image or a sound file, whereas an aggregate object might be a content package or a SCORM object.
In my opinion, many aggregate objects should be considered and treated as atomic objects since it could be far too difficult, complex and expensive to treat them in any other way. I hold this view since learning objects are ultimately difficult to reuse and repurpose for the reasons presented earlier, but this should not detract from the creation and use of repositories. Repositories are useful, especially if their use is supported by organisational structures and champions.
I hold the view that metadata should match the size of a resource that it describes. There should be metadata that describes an object in terms of overall learning objectives. Lower-level metadata can be used to add additional information to a composite object (such as an image file) that cannot be directly gained from examining its properties or structure (such as using an algorithm to determine its type).
In essence, tagging operations for aggregate and atomic object types must be simple, economic and pragmatic. If you need to do some tagging to add additional information to a resource (a pragmatic decision), the tagging operation should be simple, and in turn, should be cost effective.
High and low-level metadata tagging
The purpose of high level tagging, the description of high level aggregate object, should be obvious. Consider a book. A book has metadata that describes it so it can be found within a library with relative ease (of course, things get more complicated when we consider more complex artefacts, such as journals!).
Low-level (or lower level) metadata may correspond to descriptions of individual pages or images (I should stress at this point, my experience in this area comes from the use of software tools, rather than any substantial period of formal education!). Why would one want to 'tag' these smaller items (especially if it costs time and money)? One reason is to provide additional functionality
Metadata helps you to do stuff, just in the same way that storing a book title and list of authors help you to find a book within a library. Within the EU4ALL project, metadata has the potential to allow you to say that one page (which may contain an audio file) is conceptually equivalent to another page (which contains a textual equivalent).
By describing the equivalence relationships between different resources, the users experience can be optimised to their preferences. There is also the notion of adaptability, for example, whether a resource can be dynamically changed so it can be efficiently consumed using the device from where it was accessed (this might be a mobile device, a PC, or a PC that is using assistive technology).
One of the biggest challenges within EU4ALL is to ensure that the users interface to an adaptable learning technology system is coherent, consistent and understandable. By way of addressing accessibility concerns, all users could potentially benefit. Learners could potentially be presented with an interface or a sign that indicates that different alternatives are available at certain points during a learning activity, should they be found to exist. Presenting alternatives in a way that does not cause disruption to learning, yet remains flexible by permitting users to change their preferences, is a difficult task.
Creating metadata is something that is difficult and tiresome (not to mention expensive). As a result, mistakes can be easily introduced. Some researchers (slideshare) have been attempting to explore whether it is possible to automatically generate metadata using information about the context in which they are deployed. In fact, it appears to be a subject of a recent research tender. But ultimately, humans will be the final consumer of metadata, although metadata languages are intended to be used by machines.
The notion of a learning object is something that is difficult to define. Speak to ten different people and you are likely to get ten different answers. I hold the view that the most useful learning object is an aggregate (or composite) learning object.
Just as the idea of a learning object can be defined in different ways, the notion of granularity can also have different definitions. The IEEE LOM standard offers four different levels of 'aggregation', ranging from 1, which refer to 'raw media data' (or media objects), through to individual lessons (level 2), a set of lessons (i.e. a course, level 3), to finally a set of courses which could lead to a formal qualification or certificate (level 4). I hold the opinion that metadata should match the size of a 'learning object'. Otherwise, you might end up in a situation where you have to tag everything 'in case it might be used'. This is likely to be expensive.
High level metadata (in my opinion) is great for storing larger objects within repositories, whereas low-level metadata can be used to describe the adaptability and similarity properties of smaller resources which opens up the possibility of delivering learning resources that match individual user needs and preferences.
Posting image: from cocoi_m, from Flick. Thanks go to an anonymous reviewer whose comments have been very instructive, and all those on the EU4ALL project. The opinions that are presented here are my own rather than those of the project (or the Open University).