Content-type: text/html ~ Stephen's Web ~ Granularity in Learning Objects With Analogies From Engineering and Language

Stephen Downes

Knowledge, Learning, Community

Sept 01, 2004

Interesting talk by Jacques du Plessis drawing a comparison between how language uses basic objects, such as sounds and morphemes. Summary of his talk from ITI in Utah.

Forget about learning objects - think physical systems. What's the difference between blocks and geos - blocks rely on gravity, which creates limits in how you can join them. Consider meccano - youcan join them in more creative ways. Consider architecture - you can see cabling, etc. You have much more structure. Here's another physical system, DNA, with only four components, A,C,T and G.

So, why don't we build this room with Lego blocks. Then we could just change walls the way we wanted. Simple answer: cost. We ask the question: do we want to change this wall, really? It will cost $80,000 more. With a natural system, there is no such impediment.

Now we look at virtual systems. Digital. Language. We can draw from physical systems to see granularity, but in language we have many ways to see granularity: we have sounds, morphemes, semantics, syntax, inflection, pragmantics. In language, we have a limited set of sounds. We have more morphemes, but still a limited set. And in semantics, it keeps expanding. All these layers have to be handled naturally. And it becomes reusable. So language uses reusable objects.

Decontextualized sound - useless. Sound, morpehemes, semantics, inflection - when these are in place, we have a deliverable. When words just stand independently, much is lost. The purpose of langiage is the whole that is delivered. That is the same objective with learning objects. There is a whole that has to be delivered, the learning that is to be done. Consider the many meanings of 'uh', given different inflection. But to interepret these, we (often) need to be given the meaning.

In language, we have sounds, in digital systems we have bits and bytes. In language, we have morphemes, and in digital systems, the corresponding entities are objects (such as vectors, bitmaps, characters, digitalsounds). A sound doesn't have meaning, but morphemes begin to add meaning. Same with the digital objects. The meanings are added, layer by layer. In a digital object. the same is the case. Now, take these, add them together, and you get a digital movie. But is a movie a digital object? Of course.

What you need to do is establish your basic objects. With morpholophy, you establish an identity of meaning. In digital media, you need your objects. So in language, you have semantics. In digital media, you have index. You have simpe systems, and you have complex systems. By 'index' - you've got to know what you've got.

There are two ways to do it. Wrap everything up in one bundle, call it a folder, and you just drag and drop the folder. It's pre-packaged. The other way to do it is that it's not prepackaged, it's just there. There is an index layer. And then it brings you to a way to programatically manipulate what you've got there. Programs are to digital media what syntax is to language. With one, it's a package, signed, sealed and delivered. With the other, it's open. The conceptual way in which you deal with it in an instructiuonal sense is open. In the prepackaged way, the decision has been made. In the open mode, I deal with it as it arises.

When we come down to deliverables, in a language system, the concept is 'pgragmatics', which in a digital system, the context is 'program & objects & index'. [Good way of laying it out - SD].

Comment - what is most reusable is down at the asset level. Yes. The same with language - sounds are the most reusable. If I just have objects combined - is that an object? Indeed. But you have complex objects and simple objects.

Comment - one of the problems with reusability, as the contextualization goes up, value goes up. But reusability goes down. There's some useful level of granuality that allows us to maximize the functionality. I'm looking at the prepackaged environment versus the program environment. In language, we actually do it. As you take your stuff and you put it together, you can think of, say, a map as being useful in a lot of contexts. At that level it's pretty obvious that it's reusable. So it gets at the retrievability of the various objects.

Comment - language systems are limited, you approach the ineffable. And if you look at the deconstruction, you see that language if formed only by referentiality. There's no 'catness' about the word 'cat'. So that raises questions about reusability - how far we can push the system. Right. Look at machine translation. Everybody tries, but it creates very funny translations. But if you restrict translations to a very specific environment, like weather reports, it's very clear. So we stand at a very early stage.

Comment - we have the capacity to use syntax, but not program. I agree with that. That's why we look at machine translation. It's working out the sophistication of details, working all the variables. So I see us for now working with the prepackaged environments. But the broader we go, the more difficult it will become.

Comment - the notion of prepackaging is complex. Learning is prepacaging. You are talking about giving the people the materials, to be repackaged, or just giving them the package, and say, here, just consume it. There's a tension.

And it will always be there. You can never create your own package if you don't know it. But you learn by doing - you can be given a package, but still it was done, I didn't do it.

Comment - you can get too primitive, if you look just at the sounds it's too basic. That's why I focus on the morphemes, the first unit of meaning. You say, you stay within the digital system, you have little sound bites. But that's too primitive.

Comment - let's consider minerals, we take a rock, we melt it and get aluminum, and then we make a car. Its value goes up, its worth, but over time, there's a certain obselescence. So we might melt down the car. But we never go back to the rock. We wouldn't put it back together as a rock. Right. The analogy of the building. It's a question of cost. The cost factor of development forces us into a certain model. But in the long term, we shouldn't overlook the development of a dynamic system. We camn imagine a library, in which the dynamic creation of all these objects is possible.

Comment - thye cost issue is a good one. We have a multimedia database. We are struggling with how to partition it.Do we separate those, do we bundle thm? We face the same problems. For example, I can capture a movie of what you do on the screen with a voice over. Boom, all captured in one. Or, I can have an audio screen, and say 'go to slide two'. Then I can go back, shuffle, thinsg, because of the editability. But that is the cost problem, it's not worth it. It takes a lot of skill. It's like sword fighting versus a gun. A gun is simple.

Comment - I found a midi file on the internet. I listened to it on my laptop and it was OK. The point there is fidelity of the rendering that you get. But th point is to have the element in some standardized way. So standards have to be a part of this. To recreate the piece, the point is, the standard of midi says, ok, this instrument is on this track. There again is the issue of accessibility in a standard way. Yes. Same with vectors.

Tomorrow Jeremy Brown will be talking about MPEG 7. To render time-based video. This allows a hierarchical representation, wehere you have both, you have the high level film, but you have the multi-level version of the film, scenes, subscenes, utterances, based on their time codes. So it's an indexing scheme of the whole thing. So it's all in the context of the others. Kind of like what Lessig did this morning with Bush and Blair.

Comment - it seems like we're defining indexing as a layer that allows referencing. But the pros and cons - do you have a single source, it's changed, you change it everywhere, or do you have multiple sources. Do we toss out morphemes. It's a living system - we assume a fixed environment, do the index is really solid, but if you mess with the index... Also, we need to talk about how to index. Yes, that will be a massive work in the future. We had HTML, but content was not address, then we had XML.

My thoughts: To push the language analogy - you do not learn without speaking. But digital media is more like Chinese than English - the 'words' are not code, they are symbols. How can we work through this? How can we teach people to 'speak' in digital media?

Stephen Downes Stephen Downes, Casselman, Canada

Copyright 2024
Last Updated: Jul 16, 2024 09:06 a.m.

Canadian Flag Creative Commons License.