Automated Assembly of Learning Objects

A CONFERENCE ROOM FULL OF PEOPLE

V: It's unlike R to be late. I wonder what could have happened?

O: I'm sure she's fine. We would know otherwise.

S: (Incredulously) How?

O: We'd just know.

V: Well, maybe while we wait we can talk about a topic we've mostly ignored or, when we have talked about it, disparaged.

C: You mean standards?

(Laughter, even from S)

V: No, I mean the automated assembly of learning objects into instructionally useful experiences.

D: I've actually been doing some reading about that. A guy named David Merrill.

O: So now you not only have to be a white male to be a respected academic, you also have to be named David? Wiley, Merrill, (we haven't had to deal with Jonassen yet), what gives?

R: (As she walks into the room) Well, I guess that shuts me out of respect on several counts. Sorry to hear it. (Jumbled greetings muttered by all simultaneously) So how did we get on the topic of Dave Merrill?

D: V brought up the topic of reusing learning objects automatically, and I mentioned I was just reading some things by Merrill.

R: Well, this wasn't the plan for the night, but let's go with it. Before we get into any one person's specific views, what do you understand the phrase "automated reuse of learning objects" to mean?

C: Well, I read some about this, too. And yes, O, it was written by Wiley, but it was actually about the lamest thing I've read for this class.

D: C! Wow, I didn't know you had criticism in you!

C: Well, it was just so simple-minded. His notion has a couple of parts: first, the LOs are in a repository. Second, they're all marked up with metadata, including learning objective metadata so that the machine can very simply tell if two objects teach the same material. Third, there's an overly simplified rule for sequencing the objects.

V: So what is the rule?

C: Well, for some reason he's assuming a very old fashioned kind of instruction - very much in the sit, listen, regurgitate tradition. He uses Gagne's Nine Events of Instruction - for his sequencing rule. You see, there is a field in his metadata that also includes which of Gagne's nine events the object represents. So then the machine can do things like draw out a "Gain attention" learning object at random, and know that it will still function in the sequence the way it was intended to function.

D: Sure, and I assume the Presentation event relies on the learning objective metadata, to make sure it pulls out an object teaching the right material. If he were slightly more clever instead of pulling things out at random he would pull them out according to some other qualities expressed in metadata, like a learner preference.

C: Actually, he shows an example of that using some stuff from Martinez's model of intentional learning. He gives sample metadata like this:

Educational.InstructionalArchitecture = Gagne9

Educational.InstructionalArchitecture.Level = 3

Educational.IndividualDifferences.Style = Visual

Educational.IndividualDifferences.Intention = Conforming

S: I know the piece you're talking about now. It's the one where he calls learning objects-based instruction the "new CAI" - clip art instruction. I suppose it is sort of a goofy piece.

O: Simple as it sounds, it would be nice if we could at least get to the point where we could do the sorts of things he's talking about.

V: Merrill's stuff, of course, is quite a bit more sophisticated.

R: Before we get into Merrill's view, it might be useful to look at one more approach that uses the lego metaphor of "snap the objects directly together." Is anyone familiar with the Cisco model of creating and using "reusable learning objects" or RLOs?

S: I remember reading some criticisms of early versions of the model, particularly around the lack of integration across modules.

R: Yes, many of those are addressed in this most recent offering. The model goes into some depth describing a set of Design Rules that must be followed in creating materials for use in the model. They have clearly specified the types of things you can teach with their model, including Facts, Concepts, Processes, Procedures, and Principles, and the types of learning objects you can build and how they fit together, in a Course - Module - Lesson - Topic manner. It's *highly* structured, just like LEGOs themselves. The reason they snap together so well, is that each one follows the design rules precisely.

S: You might even call them standards.

C: You might if you...

R: (Over top of C). So V, tell us about Merrill.

V: He has a really different view of things. He calls his flavor of learning objects "knowledge objects." The primary difference seems to be that metadata is the prime content for Merrill.

D: I think its more fair to say that while most people think of learning objects as bits of content stored as html, pdf, or powerpoint files, Merrill tries to store the content in a representation-independent manner. He actually stores it like data to be pushed through an algorithm. He uses the term "instructional transaction," but what he really means is an algorithm that "knows how" to present materials, facilitate practice, and give feedback.

O: Just how is a machine supposed to do all those things?

D: It's actually rather clever. He treats the instructional system as a giant state machine. In fact, from some perspectives, his "instructional transaction theory" is really just a way of building simulations, and his "knowledge objects" are just components that can be reused across simulations.

V: Another great paper by Merrill about this work is Knowledge objects and mental models. He actually goes on to argue that if you were to teach his knowledge object approach to students they could use the model as a kind of metacognitive scaffold for learning domain material more effectively.

C: Did you just say "knowledge object," "mental model," "metacognitive," "scaffold," and "domain" in the same sentence? What are you, a buzzword generator?

O: He's not the only corporate type here, C.

R: Ok, ok. Yes, that was quite the jargon filled sentence, but that's no reason to get excited. So that's two approaches to automatically assembling learning objects. I hope you fully appreciate the different opportunities and difficulties inherent in these approaches - the one that uses reified content plus metadata and the one that uses the representation-independent "metadata is content" approach. I'd recommend having a look at the papers we've referenced tonight.

C: You mean that's it?!? We're done already?

R: Yes, that's it. There's no way we could go into depth on the approaches in a conversation like this. However, if you have a look at the articles...

C: Ok, I get it.

R: The important take away from this conversation is that there is more than one way to store content for reuse later and, as we have said before, those decisions you make about the architecture of your objects impact the paths that remain open to you later.

O: So what are the questions for tonight?

(R Walks to the board and writes)


 * 1) What are some of the different ways that content can be stored / represented with an eye toward automating the reuse process? Can you think of new ones?
 * 2) What are some of the different ways that learning objects can be automatically reused? Can you think of new ones?

O: Strangely enough, we seem to have finished this conversation without covering any of the risks or downsides associated with automated delivery.

R: I think that will be the topic for next week. I should advise you though, C, that next week is likely to be another short meeting.

D: Works for me!