Nov 21, 2002
what about the learner? Instead of having a course that flows and progresses and engages their interest, they're going to get a set of sterile stand-alone modular objects. They will have had any vaguely interesting idiosyncratic features carefully formatted out of them so they can be (hopefully) seamlessly attached to the next personality-less RTO. And I look at it all and think, but surely for learners to learn, we want to tickle their intellects, and reward them with intriguing new thoughts. How can we expect them to engage with personality-free disjointed nuggets? And aren't we going to try to surprise them any more?
1.
When I was a student what I learned to hate was that professor who was more personality than content. Every day was a surprise, which meant I never had any idea what I was going to learn and had no way to prepare for it. Oh, certainly, I had a lot of fun (it is very easy to distract a professor like that) but at the end of the course I did not know a lot more than I knew going in - and I didn't know what it was that I had learned.
The fact is that it is not true that a sequence of, say, molecules, results in a whole with no continuity. If this were so, we could not have continuous surfaces, such as, say, desks, composed out of molecules. Or our language, composed as it is out of atomic words, could never achieve the elegance and rhythm of a Shakespeare sonnet.
Hollet's error here (and it is a very common error) lies in presupposing that the whole must be somehow contained in the parts. That unless there is some manifestation in each molecule of a surface, that a collection of molecules could never create a surface. But the whole is the emergent property of the set of the parts. Though no brick could ever aspire to be six feet tall, the collection of them, in some sort of non-random order, can build a wall.
There is a great deal of evidence to suppose that elegance and even beauty can be assembled from a set of inelegant and ugly parts, that coherence may be obtained from a set of unrelated entities. I have mentioned molecules a few times. Perhaps also consider the function of the brain, in which a set of connected but autonomous neurons managed to create the Sistine Chapel. Or perhaps consider Minsky's theory of the society of mind, in which cognition arises from the set of autonomous agents.
It will take a great deal more argumentation to show that this phenomenon, which is repeated throughout nature and human endeavour, cannot succeed here as well. And if some of the first creations have been bland and uninspiring, like a child's pile of blocks, well perhaps that is to be expected in the first few years of effort. I see no reason why elegant learning objects each dedicated to explaining a particular concept cannot, when assembled, engage and even surprise.
I would like as well to address some comments offered by Diana Carl, who notes that "the approach described by Ms. Hollet works very well for training and situations in which observable, measurable changes in work performance is the desired outcomes. However, in an academic setting where the learner's use of content is less defined and [intentionally] less predictable, it harnesses the experience."
Now I think that Carl has a point, but it is a point that should be drawn out a bit. This concerns the question of how the content of a learning object should be defined. What is the taxonomy by which a given subject area should be described?
As Carl suggests, it really does depend on what you are up to. In some environments, a performance based taxonomy is appropriate because we have specific performances in mind as outcomes. This is often the case in the military or in business, where the learner is expected to be able to perform a specific function as a result of learning. In other areas, however, this is not the case. But this does not mean no taxonomy is possible, it means simply that it is defined differently. In a university class, for example, the division of a course of study into components may be based on the subject matter. First we'll look at Act I and then we'll look at Act II. First we'll examine character as the struggle between good and evil, then we'll look at the theme of portent and omen.
It is arguable - and I would argue - that any academic approach can be broken down in this way (or conversely, that any instructor who is not able to identify subtopics in the material does not, in fact, understand the material). For otherwise it would be impossible to divide programmes into courses, courses into classes, books into chapters, philosophers into schools, phenomena into categories.
The key understanding here is that there is not, and never will be, one correct approach to the identification of different parts of a field of enquiry. And that in some cases one means of identification will be inappropriate in a given circumstance. An academic study of a language would find a list of practical phrases absurd and useless, but for someone travelling to Italy for the first time, practical phrases - and not an overview of Italian grammar - are what is needed. The taxonomy depends on the circumstance.
And as Carl suggests, a blend is often the best approach. "Like any professional, I have to choose my tools and interventions according to the situation that is going to best help the client for the learning get what will truly help."
She continues, "understanding the cognitive mapping issue in constructivism is important in developing SCOs. Whether it is discipline-specific such as is nurtured within a university, or whether it is culture-specific, as nurtured within a foreign culture, or whether it is mapping associated with the unique way an individual has come to view the world, employing this map when SCOs are developed helps make SCOs reusuable in a constructivist way so that in the end the tools available to the learner help her or him "construct" their own new knowledge."
I think this is exactly right. And as a consequence it is important not to confuse the intent of SCORM with the implementation, particularly as it has occurred in the military and the corporate context. Organizing content the SCORM way (whatever way that is) might be inappropriate in an academic context. But even if it is inappropriate, it does not follow that there is no appropriate way to organize content. And indeed, there must be, for otherwise all academic learning would be formless and shapeless. Just as a military manual would be a little out of place in a university English class (and a book of Sonnets out of place in a tactical exercise), we need to understand that different approaches to learning objects are appropriate in different settings.
This is why, when I write about learning objects, one of my over-riding concerns is directed toward multiple semantics. That is, we need to be able to use various flavours of learning object metadata to describe learning objects. Indeed, the choice of taxonomy itself will help professors and students choose between performance related content, as described using SCORM, and concept related content, as described in, say, UECML (University English Content Markup Language (it doesn't exist, I made that up).
SCORM is most appropriately described as an "application profile". That is, it starts with a basic vocabulary (a standard, known as IEEE-LOM) and tailors it for a specific use. It is, in my opinion, the first of many application profiles that will be developed, each tailored toward a particular application. What we need to learn is how to manage different vocabularies in metadata as easily as we do the same thing in English, to be able to determine by context, say, whether what we mean by a 'calf' is a farm animal or part of the human anatomy.
2.
Please let me respond to Brad Jensen's brief challenge to the comments above and to some of Steve Eskow's comments.
Jensen writes, "Okay Stephen, you are generating a ton of content day by day. How much of it have you structured and tagged as RLOs (Reusable Learning Objects) and where is it and how do we use it?"
Since 1998, my content has been tagged using a metadata system known as RSS (specifically, RSS 0.91). RSS (it stands for 'Rich Site Summary') is intended for syndicated news content. RSS is more appropriate for my purposes because my content is more like news and articles than online courses. But the method and the purpose is the same.
For example, I generate an RSS version of my newsletter every day. You can find this RSS file at http://www.downes.ca/news/OLDaily.rss (you may need to right-click on this file to download it). This RSS file is generated automatically by the smae system that generates my newsletter in text, HTML and Javascript versions.
The easiest way to use my XML content is to download a product called 'Amphetadesk' and install it. Amphetadesk is what is generically called an RSS headline viewer. It is in essence what is missing in the world of learning objects: a quick means to locate and view learning objects. Once you have installed Amphetadesk go to the web version of my newsletter at http://www.downes.ca/news/OLDaily.htm and click on the orange icon with the pill at the bottom of the page. This will cause Amphetadesk to automatically retrieve my content (along with any of many other syndicated news sites).
Amphetadesk is one of several tools for viewing RSS-syndicated content. You can also view directories of RSS content at newsisfree.com (or just Google for RSS viewers and RSS diretcories). They vary in their utility, and none of them provides everything I would like, but the essential point is that they work and they are currently in use by tens of thousands of news sites and millions of users.
Much of what I have had to write about learning objects has been informed by my work over the last few years on syndicated news content. It is my opinion, and I have expressed this in numerous cases, that the structure of a learning object network should resemble that of an RSS network. Once again, let me refer you to the paper I delivered in Milan http://www.downes.ca/files.milan.doc, where the principles and structure I propose echo and improve upon the RSS network structure.
It may be argued that, because I have elected to use RSS instead of, say, SCORM or AICC, that I have not infact created learning objects (we will leave aside the fact that it would be trivial to do so, since having already created RSS files I need only create a new template to create these other sorts of metadata).
But part of what I asserted in my previous message stresses that it would be absurd to adopt one and only one metadata schema for the tagging of content, even educational content. SCORM was created for a very specific sort of object; my work is a different sort of object. Many other educational objects (some people call them knowledge objects) are in a similar category. It would be absurd, for example, to assign many of the SCORM tags to an image; an image is not the sort of thing that has (innate) educational outcomes. Much more appropriate for images would be an image markup language.
The question, then, is whether learning content management systems (LCMSs) and learning management systems (LMSs) should be able to locate and use these alternative sources of content. I believe strongly that they should. It is likely, for example, that some of my articles may be of use in online courses. The articles should not be excluded from view because they have been tagged in RSS. A proper LCMS (or learning object network) should allow for the selection of wide varieties of content, depending on preferences exercised by the user.
The other part of Jensen's question - beyond whether my work is properly tagged - is whether I have included appropriate 'wrappers' around my content so that they can interact with learning management systems when in use. The short answer is no, and the reason for this is that I don't believe it is necessary. It is my opinion that if the LMS requires special coding (even if only simple Javascript calls) in order to display an HTML paged, then it is broken. An LMS, when presented with learning material without appropriate wrappers, should provide default wrappers.
3.
Jensen continues, "I'm not tryoing to put you on the spot. I'm trying to ppoint out that creating RLOs is a pain, and until there is some universal place to go and find them, a a reminder that a person has placed content there, they just won't have any critical mass. I suggest you create such an archive, if you really think they are useful, then have a link at the trailer of each of your emails that says: here are my free/paid RLOs, here's my major subject area, click here for the search engine. Then get everyone you know to do the same."
That is in fact exactly what I am doing. In my work with the eduSource project in Canada (and withy a number of smaller and parallel initiatives) I am working toward the establishment of what I call a 'distributed learning object repository network' (or DLORN). The purpose of the DLORN will be to create an open marketplace for the exchange of learning objects (and other educational resources) using a wider variety of metadata. Part of DLORN essentially involves the creation of what I call a 'learning object browser'.
Other projects are directed toward the same objective, and I would not be surprised if some entity with greater software resources that I have access to will be able to create, say, a learning object browser before I can. That's all right with me, particularly if I am able to work with such people to establish compatibility between their software and ours. Our intent is to build an open source royalty-free version of the network with simple and usable components. The mnore versions of each component, the better.
To change subjects, now, both Jensen and Eskow have difficulties with my use of emergent properties to explain how order may be obtained out of learning object chaos. The phenomenon of emergence is well recognized in other fields, which is why I sought to explain what I meant by analogy.
Jensen suggests, "You cannot tag a sonnet as RLOs and expect to use a phrase here and a couplet there, and come out with more sonnets."
This is simply false, but the falsity is obscured by a misunderstanding of the granularity of the parts and the syntax used for tagging. At the lowest level of granularity, Shakespeare is working with a set of 26 letters. In the use of letters, ordering is important (reminds me of IMS's simple sequencing). Letters are not tagged per se, but as physical entities their nature is clearly indicated by physical shape.
At the next level of granularity, Shakespeare is using words (and indeed, for the most part he does not invent words in the manner of, say Carroll, but uses the set of 50,000 or so previously created words available to him at the time). Words are tagged by the use of a single space and some basic and simple punctuation: periods, commas and the like. Again, the order of the words is important.
But note: the tagging, the ordering, even the meaning of the words is not contained in the word itself. The meaning of the sentence cannot be found by studying individual words in isolation. It is only when looking at the sentence as a whole (or if you're Chomsky, by looking at the syntactic structures embodied in the sentence) that an understanding of the meaning may be obtained. The meaning of the sentence is an emergent property of the sentence: it may be found only by considering the entire sentence, and even then, only when viewed in a context of reference or representation.
This is why a word such as 'calf' is a useful example of a learning object. By itself, it is of extremely limited semantical value. If I simply went to someone and uttered the word 'calf' then (unless I am in a fine restaurant with a tolerant waiter) my intent and signification would be opaque. The meaning and relevance of my utterance is significant only in the wider context. In this way, the meaning of the word 'calf' combines with that of the other words in the sentence, and the contextual surround, in order to create a meaning that is quite over and above what is contained in the original component.
This isn't Zen or magic or anything mystical. It is a commonly understood and utterly repeatable property of networks (in the case of a sentence, a semantical linear network). Learning objects, like words, are nothing over and above nodes in this network. What makes them interesting, what makes them pedagogically useful, is the manner in which they are combined. A single image is of limited value; a series of images linked together can create a (silent) movie. A single instant of perception is by itself meaningless; a series of perceptions ordered creates a lifetime of experiences.
Jensen continues, "I dont follow this as an abstraction of her argument. It seems to me that it is the RLO argument to think if you chop up Shakespeare and reuses parts, you might get Yeats."
On the account I have just given, this is not only possible but likely. Reduce all of Shakespeare to a set of words and bundle these words into a repository (for simplicity, I will simply call this a 'dictionary'). From this dictionary one can obtain all the elements needed in order to create Yeats.
People should stop thinking of learning objects as though they were classes or lessons or some such thing with built-in intent, at least from the point of view of thinking about how they are used. It is much preferable to think of them as a greatly enhanced vocabulary that can be used in a multi-dimensional (as opposed to merely linear) network.
Criticisms such as Jensen's and Hollet's seem to me to be like somebody holding up an instance of the letter 'T' and saying, "I just don't get it." But there is nothing in the letter 'T' that suggests that it could be a part of the word 'To' just as there is nothing in the word 'To' that suggests that it could be part of the sentence 'To be or not to be.' In a similar fashion, there is nothing in one of my objects, one of my 100 word essays (of which I produce a half dozen each day), to suggest a wider philosophy, but ordered in a certain way, themes and concepts emerge (I try to capture this with the 'Research' feature of my website, using regular expressions to create sequences of objects).
Jensen's postscript (perhaps he has had second thoughts) comes much closer to the truth: "You want an metadata with the granularity of English that is not English itself. Maybe the content is its own metadata, and what you need is a search engine that is thesaurized. Another possibility might be for a content analyzer that is vocabulary driven, and rates documents for simularities to other documents (similar relatively unique words, content semaphores such as 'Lewinsky', etc.)"
4.
I would like now to turn to Steve Eskow because I think he demonstrates some extrordinary insight in his comments, even if he does say that my argument is based on a fundamental misunderstanding.
Eskow gets my point exactly: "Steve wants us to understand that letters and words do not of themselves combine and cohere into a sonnet that begins 'When in disgrace with fortune and men's eyes," or "Shall I compare thee to a summer's day?'"
He then moves to the next step of the inference: "Shakespeare designed beauty out of growls and grunts and vowels and consonants, and Steve is wanting to lead us to accept the notion that the 'instructional system designer' similarly takes inelegant and ugly parts and creates a 'course' that is a thing of elegance and beauty."
Then he attributes to me the following error: "The error to which Steve's elegantly constructed argument is leading us to is, if the critics of ISD are at all right, is this: It is not the instructor who has to create learning out of bits and pieces, often crude and inelegant, but the student."
Here Steve leads the reader elegantly in three paragraphs to a conclusion I would whole-heartedly endorse, and indeed, have explicitly stated on numerous occasions (as for example, in my essay In Practice... and in various presentations about online learning environments).
He continues, "A course ought not to be, or need not be, a thing of beauty for students and accrediting bodies to look at with awe and reverence, judged like we judge the Sistine Chapel or a Shakespearian sonnet. It often needs to be a thing of shreds and patches that requires the student to do the hard work of organizing into form: of connecting the dots, as the current clichÈ would have it."
Eskow endorses, with a good reference to Arthur Chickering's "Education and Identity," a "junkyard curriculum made up of pieces of stuff found by instructors and students and assembled into coherence mostly by students."
I completely concur. I have long felt that the attempt to organize bits of instruction into neatly packaged courses is a mistake. That does not mean that the presentation of materials to students is completely without order: rather, it is to argue that the grammar of such presentation is no, should not be, a linear sequencing of prepackaged learning events, a presentation in which the student is nothing more than a mere spectator. Such a presentation is effective for the sequencing or letters and words, but just as the set of rules for ordering letters is different from the set of rules for ordering words, so also is the set of rules for ordering higher order constructions (aka and now almost meaninglessly called 'learning objects').
My belief is that the grammar for organizing higher order constructions is the grammar governing semantical networks. I believe that examples of this grammar exist, and that the organization and presentation of layers of increasingly complex concepts should (in the first instance, at least, until and unless we find a better way) emulate this grammar.
This grammar is instantiated in the design of the human brain and has been modelled in neural networks. What I am after here is not simply a mere mimicking, though, but an understanding of the design parameters: an inderstanding of why such a network functions as it does.
A network is essentially a collection of points in a multi-dimensional space. Each one of these points is a representation of a higher order construction - an image, a 100-word essay, a concept. In the literacture there are various attempts to define what these must be (I don't have the references at hand, but I'm thinking of people like Land and Kosslyn).
The grammar of the network is therefore the set of rules describing the connectivity of the points in the network. A very simple (and incorrect) grammar would be to take a sequence of points (a sequence of concepts, a sequence of learning objects) and join them in a linear fashion. This, unfortunately, is an extremely limited and inflexible grammar.
But nore would it be of any use to create a system whereby every concept is joined to every other. As Francisco varela shows, the idea point of connectivity in a network is somewhere in the middle. You get meaningful emergent patterns if networks are only partially connected. Intuitively, this makes sense. If you sought meaning from the world wide web, just as a list of zero pages would not help you, so also would a list of every page. But neither would a list of one page, which is what you get in a linear sequence. You need a mechanism whereby connections between the pages are limited in number, therefore restricting your view, but not overly limited.
In neural networks (such as, say, image processing), this limited connectedness is obtained by organizing the neurons into layers. One layer of neurons can connect to the next layer of neurons, but no further; to get to the third layer of nurons you must proceed first through the second layer. The act of moving from one layer to the next changes your state; the concept you are working with has changed. It has become a combination of concepts, or it has been an abstraction of the concept. A sequence of movements through the layers (both forward and backward) creates a series of patterns of activation, and through this, a more complex construct emerges, just as the meaning of a sentence emerges by means of a movement through the sequence of words (or some other form of perceiving the words).
So the organization of learning content - of learning objects, as it were, though we find now that this designation refers only to one layer in a much more complex ecosystem - in a learning environment ought to proceed by defining layers of interconnected resources. The art of instructional design, therefore, comes in the placement of a student somewhere within that ecosystem and the identification of relevant connections. The student, by navigating (perhaps in some goal-directed way, perhaps out of curiosity and interest - this is part of the design) therefore creates a new entity, perhaps by identifying new connections, perhaps by deducing abstractions already present (but not revealed) in the network, perhaps by an analysis of the nature and role of component parts.
5.
Some of Gilly Salmon's recent work gives us a first approximation of what such an environment would look like. Salmon identifies four major types of educational ecosystems (or 'worlds' as she puts it). To briefly recapituate Gilly Salmon's Four 'Worlds':
Contenteous
- technology as a delivery system for content
- instructors as content experts
- libraries, repositories, databases
- institutionally centered
Instantia
- based on the idea of learning objects
- instructors as content assemblers
- context-senstitive, just-in-time learning
- repository or aggregator centered
Nomadict
- learning travels with the learner
- mobile learning, alternative delivery systems
- learner centered
- instructors as 24-hour help desks
Cafélattia
- learning communities, communities of practice
- asynchronous and synchronous communication
- focus on professional development and tacit knowledge
- instructors as moderators
The entities in each of Salmon's four worlds are very different. They differ in granularity, they differ in use, they differ in meaning and function. As Salmon, with a cynical wit, observes, it would make no sense to design an educational strategy in anyone of these worlds. An education based on (sequences) of content only would be meaningless.
Salmon presents these as four alternatives, from which we must make a choice (as a literary device, she tells me). But of course they are not four separate alternatives: they are layers of interrelated types of content. They form a single ecosystem. Salmon also argues in favour of education at the level called Cafélattia. This creates what I would call a lattecentric view of education. But learning communities do not exist in isolation; a lattecentric ecosystem is at the center (or the bottom, depending on your perspective) of an atmosphere made up of the other three layers.
Now the sort of conversation that we envision happening in a café consists of sentence and words. This is what confuses us. It is difficult to see how any of the otrher layers could be relevant when they are made up of the different sort of discourse that occurs in a café. But as I tried to argue in The New Literacy we have a much wider range of semantical entities to choose from when conducting these conversations. Though words and sentences could be used, they allow a far narrower a range of conversation than could occur in a fully developed ecosystem.
We would expect students to converse using a wide range of complex objects, composed perhaps of words and images and other multimedia, arranged in some cases in linear fgashion but more commonly as a network of related concepts, as parallel and multi-tacked constructs, thus employing a vocaublary far wider than our meagure collection of 50,000 words may employ. This vocabulary is the set of objects available in the ecosystem, a set of objects selected from (using Jensen's 'thesaurus' of learning- or concept-objects) the environment, the next layer of the atmosphere.
Educational design, in the first instance, therefore consists in the creation of these environments, of the writing of a (modified and perhaps restructed) version of the thesarus, making a certain relevant set of objects available for discourse, or providing the tools for and motivation for interaction and the construction, either individually or in groups, of new entities, of the creation of mechanisms and guidelines for pointing, suggesting, even cajoling, of having a broader understanding of the wider ecosystem and of creating a microcosm of that ecosystem.
Educational design, in the second instance, lies in the creation of these objects and mechanisms through which they would be delivered. At greater levels of complexity, such objects may have the specific intent to teach a concept, and may resemble a traditional design of learning objectives, content, exercises and even assessment. But may other designs are possible: simple simulations designed to animate a concept, a collection of film clips designed to illustrate a sequence of events, summaries and analyses of articles and papers and more.
At yet a further layer, educational design involves the creation of the raw materials from which these more complex entities are constructed, the creation of single clips from which a sequence is derived, the creation of expository works out of which a lesson is designed. There is no attemtp at this level to recreate the entire educational experience; the point is to create, if you will, a large and complex vocabulary of individual concepts.
And the nature of instructional systems design, properly so-called, lies in the creation of networks of interactions between these layers, in the manner in which raw materials will be presented for use by people creating learning objects, and in the manner in which learning objects are presented to students working within a lattecentric educational environment.
The vast confusion in the field of learning objects arises when we conflate all of these tasks as being a single task, when we try to compress all the levels of educational design into a single level. It occurs when we presuppose that one single sort of semantics (such as, say, SCORM) could adequately describe the myriad entities and connections available in and between the different layers. Each sort of entity, each sort of connection, requires its own semantics, and the set of semantics taken collectively represents the grammar that governs the network.
So people like Jensen ask, where is this network and why aren't you using it. And of course the network as a whole does not yet exist, though component parts are coming into play. Ands just so, I use RSS because RSS is the appropriate semantics for my little bit of a fairly simple (and non-educational) layer, in anticipation of the time when the creation of the appropriate layers enabling semantic interaction may take my work and, though aggregation, abstraction and interpretation, make it the sort of entity that may be exchanged and used productively in a learning envrionment.
And it is also with this model in mind that I am working on the design of a distributed learning object repository network. What I am working toward, through both the identification and creation of transport protocols and semantical networks, is the creatioon of a mechanism that instantiates the transmission of content (very widely conceived) one layer to the next. And I am working to identify simple instances of learning environments, instanced by such tools as Amphetadesk, where a multi-layer transmission of information occurs.
