Anything Or Automating

A CONFERENCE ROOM FULL OF PEOPLE

R: Here we are again.

C: I'm actually finding myself looking forward to these get-togethers... (Hurriedly) but don't tell my Product Manager... He still thinks I hate them and that he's really going to owe me one by the time they're all over.

(General laughter)

D: To be honest, I'm hoping that we're going to get to something involving development soon.

(S rolls his eyes)

You're always on and on about standards, so don't act surprised that I'm looking forward to talking about the topic that interests me, too.

R: Actually, D is in luck tonight. Our topic tonight is the internal structure of learning objects. You might think about it as the architecture of learning objects.

D: (Excitedly) Architectures, huh? Now we're getting somewhere!

V: I thought we agreed that the definition of learning objects depended a great deal upon the context in which we find ourselves.

O: That's right.

V: Then how can we have a conversation about what a learning object's internal structure should be like? I can speak about our model at some length. Our objects are structured around an instructional objective, teaching materials, and assessments. It's a completely self-contained package delivering an effective instructional experience...

R: (Cutting in) Thank you for the advertisement, V.

O: Well that's not what I think of as the way a learning object should be structured. I'm much more interested in the quality of the content than in how its structured.

V: How can you guarantee a level of quality without a standardized process?

O: (Incredulously) How can you guarantee a level of quality with a standardized process?

R: Okay, okay. This is the core of the argument we're going to have tonight, but let's try to make it a little less personal. V seems to be arguing in favor of a specific structure produced by a specific process. Let's spend the first part of our time discussing the benefits of such an approach. (To V) And V, if you're going to participate in this conversation, please try to sound a little less like a radio ad.

V: It's a simple matter of quality improvement. If each object is something completely different - a product of a different process or a product almost unrelated to the product produced before - how can they be compared? How can we tell if one is better than another? If we're getting better from one generation of our product to another generation? Rigorous, scientific study involves controlling the environment, changing only one variable at a time, and making empirical comparisons of the differences. When multiple variable are changing value from one generation to the next, there is no way to ask meaningful questions about what makes one generation of product better than another. In other words, we place ourselves right back in the mess that educational research has been in for the last century.

R: That's quite the accusation.

S: But he's mostly right, isn't he? I mean, he's wrong about the beginning of his argument. We *can* carry out studies to determine the relavtive effectiveness of any two instructional products, whether they're learning objects or anything else. But his more important point is that we have no rigorous way of determining what is was that accounted for the difference in the two experiences if we don't carefully control the artifacts as well as the environment of use. And in that point he's correct.

O: (Confusedly) What does any of this have to do with learning objects?

D: How do you mean?

O: I mean, how are learning objects any different from textbooks or manipulatives or flash cards in what you're saying?

V: They're not different at all. And that's the point. Effective instruction is effective regardless of the medium the instruction is communcated over. Why can't we learn that lesson?

R: Are people interested in learning objects because they think they will be more effective? We've just heard an argument that the medium of communication has nothing to do with effectiveness. So I don't think that can be it.

V: They can be more effective, but not because of the direction you guys are taking the argument. They can be more effective because they can be more consistent.

C: Nuuuh, consistency doesn't guarantee quality. You can have consistent garbage.

V: Yes, but. When the learning objects all share a common architecture - when we know exactly what we can expect to find in a learning object - it becomes possible to write all sorts of systems that can automatically use and reuse learning objects. And then all the arguments in favor of intelligent tutoring systems apply to the learning objects approach. The automated system can keep perfect track of every learner's progress. It patiently provides practice and never forgets feedback. And, most importantly, it is completely consistent. It always follows the principles of best practice.

O: Whoa, chief! That's quite the leap! How did we get from sharing a common architecture to having intelligent tutoring systems?

V: Because when I know what kind of data I will find inside an object, and how it will be formatted, then I can write algorithms that consume the data and do useful things with them.

D: And there's more to it than that. If we are sensible in the way we architect all our learning objects, we will store the content in a representation-independent manner, like in XML. This not only allows us to automatically use all manner of instructional approaches with the content, but to render the content in all kinds of formats for all kinds of devices - HTML and PDF for web browsers, sms messages for mobile devices, text-to-audio streams for iPods and the visually impaired... it goes on and on. That kind of architecture also allows the presentation to "degrade gracefully."

I: Right. Not everyone can afford the latest and greatest hardware. We need to make the content usable in the inner city and the Third World.

T: My classroom, in other words.

C: That's a good point. Lots of times we can specify the delivery environment, because we're working with our clients' IT departments. Of course, they often constrain what we can do, especially on the web. Firewalls that block plugins, ancient browsers... gettin' off topic. Sorry. Go ahead, V.

V: When you combine the perfect consistency of the intelligent tutor with the device independence of a format like XML with the ability to carry out controlled research aimed at incremental improvement, there can be little doubt left that a well-planned, (nodding to S) standard learning object architecture is exactly what we need.

O: (Looking not quite as sheepish as she should) Well, I will express that "little doubt" then.

(All heads turn suddenly to look at O)

O: How many learning objects are there in the world today that have been developed according to this standard architecture?

S: None, of course, since no such standard architecture exists, strictly speaking.

C: What is that supposed to mean?

S: Well, the ADL's SCORM describes a minimal architecture an object must conform with in order to be considered a sharable content object (SCO). The SCO must have an interface by which it can communicate with a learning management system, in order to report things like time spent on the object and scores made on exams.

D: Right, a SCORM wrapper. But that has almost nothing to do with the internal structure of the content, and much less to do with the development process.

R: What about V's architecture? Or the learning object architectures developed and adopted by other corporations?

D: Yes, those would describe internal structure. But we need to go deeper - a representation-independent structure and markup of content. Something using XML, RDF, or even the emerging microformats.

S: Don't be deceived, though. While XML and RDF have been through consensus-based standards processes, no way of expressing instructional materials using them has. SCORM is the closest standard we have to the learning object architecture standard you're talking about... unless we're going to talk about IMS-LD.

O: Talk about putting the LD in learning design... (C chuckels) Anyway, S's answer is the correct one. If one takes this "standard architecture" view of learning objects, there are no learning objects in the world. Not only are there no learning objects in the world, there won't be any for a very long time.

C: How on earth do you figure that?

O: Well, if there is no standard architecture, there can't be any learning objects that conform to it. And since the consensus-based standards process is famous for being glacially slow there is very little chance that such a standard will exist any time soon. And since the people who work on this standard will, in fact, be trying to solve the same problem the semantic web / ontology folks have been trying to solve for years now, I would say that the odds of it *ever* getting done are near zero.

D: The alternative to a consensus-derived standard is one that's just imposed.

O: The Redmond Borg certainly would like that! They've done it for slideshows and documents by and large, but they haven't succeeded with learning objects.

R: Probably because, as we've seen, learning objects are a lot more complicated than slideshows or documents.

I: Slideshows and documents can BE learning objects. Or not... (grins)

V: So this is the "little doubt"? That there is a lot of hard work ahead of us? Of course there is. You don't think we are equal to the task?

O: I don't want to seem to lack faith in humanity. Let me express another little concern then. Let us assume that such a standard architecture existed today. That it was approved by an internationally recognized body like the IEEE LTSC this very morning. How many learning objects are there now in the world?

C: None. But now it is only a matter of time.

O: Will NBC, CBS, ABC, the BBC, the AP, Yahoo, Google, Rush Limbaugh, Howard Stern, Wikipedia, and all the bloggers in the world begin marking up their content in this way?

R: I should say not.

O: Then if I wish to use some BBC or Wikipedia or other content as a learning object, how can I?

D: You just need to transform the content from whatever format you find it in to a format conforming to the standard architecture. Or wrap it in an interface layer.

O: (Attempting to sound like R) Ding! We have a winner!

(Everyone seems supremely unimpressed)

O: We've spent a few weeks now talking about how reusability enables all kinds of desirable outcomes, like shortened development times and lower costs of production. Let me ask you this: how much content is there on the internet today?

V: Google claims to have indexed something on the order of six billion objects.

D: The Internet Archive has a petabyte of data and gets 20 terabytes bigger every month, much of which is publicly avaialble. But what's the point? It sounds to me like you're arguing that plenty of learning objects already exist.

O: I'm arguing that it is queer for a development methodology so heavily reliant on *reuse* to adopt a philosophy which precludes the reuse of six billion existing objects, and a petabyte of images, animations, and video. I suppose you could reuse them, but not without costly retooling and modification. Why not work with a philosophy that embraces existing resources as they are? Why not develop an approach that can reuse any resource?

V: What could you possibly do with resources whose structure is completely arbitrary? What kind of automated systems could you build? What kind of research could you carry out?

O: (Firing up) Many more systems and much more research than I could carry out with a collection of nonexistent resources.

C: How would these learning objects be assembled?

O: Just like all the other resources in the history of education have been assembled. By hand; by human beings. Personally, I have little faith in the ability of an automated system to choose the right learning objects, let alone sequence them effectively.

R: Don't confuse the issue, O. We're not arguing about the viability of AI. I believe your point is an interesting one. For all the emphasis the proponents of learning objects put on reuse, it is interesting that so many would take this approach that makes it impossible to reuse any material you find without changing its format.

D: Saying you can't reuse the materials without changing them first raises some interesting questions about what we mean when we say "reuse."

I: That's what Stephen Downes is on about. Remix, feed forward. Blogging-as-a-verb, to quote Will Richardson. What's the big deal about having automated systems assemble content? Make the content AVAILABLE. Let people - individuals - take what they need and leave the rest. They'll provide the context and the sequencing. Who needs a robot to assemble content and spoon-feed it?

R: That assumes that learners can in fact decide what they need to learn and choose the most appropriate materials. The research shows that while high-ability, self-directed learners like us can do that most of the time, there are a lot of folks that need to be spoon-fed. But that's a topic for another week. So even though we haven't gone quite as long as we normally do, why don't we pause here for the night. What should I add to the board for tonight?

S: How should learning objects be structured?

V: Perhaps something more like what should the structure of learning objects be?

O: How about should learning objects to adhere to a common structure?

R: Let's go with that:


 * 1) Should learning objects adhere to a common structure?

R: Thanks everyone! More next time.