In my previous post in this series, I focused on elearning, and its shift into the mainstream at the end of the 90s. This was accompanied by new approaches, often derived from computer science. One of these that gained prominence was learning objects. The concept can be seen as arising from programming – object oriented programming had demonstrated the benefits of reusable, clearly defined pieces of functional code that could be implemented across multiple programmes.
Learning objects seemed like a logical step in applying this model to elearning. As Stephen Downes argued:
“there are thousands of colleges and universities, each of which teaches, for example, a course in introductory trigonometry. Each such trigonometry course in each of these institutions describes, for example, the sine wave function. Moreover, because the properties of sine wave functions remains constant from institution to institution, we can assume that each institution’s description of sine wave functions is more or less the same as other institutions’. What we have, then, are thousands of similar descriptions of sine wave functions. …
Now for the premise: the world does not need thousands of similar descriptions of sine wave functions available online. Rather, what the world needs is one, or maybe a dozen at most, descriptions of sine wave functions available online. The reasons are manifest. If some educational content, such as a description of sine wave functions, is available online, then it is available worldwide.
This made a lot of sense then, and it still makes a lot of sense today. Step forward then, the idea of learning objects, with a rough definition of “a digitized entity which can be used, reused or referenced during technology supported learning” (more on definitions later). A lot of work accompanied the learning object gold rush: standards were developed to make them reusable, platforms were built to deploy them, content was produced in their style, and papers were written about them.
But they never really took off, despite the compelling rationale for their existence, that Downes and others set out. Their (or our) failure to make them a reality is instructive for all ed tech I feel, and they are something I frequently reference when we’re discussing new technologies. So, here is my list for why learning objects failed (although, to be honest, this video interview with Brian Lamb is a better account):
Overengineering – I’ll cover standards in another post, so won’t say much here, but in order for LOs to work like software objects, they needed to be tightly standardised. This version of the LO dream went beyond Downes’ sine wave simulation, and had as its dream courses that were automatically assembled on the fly from a pool of LOs for a personalised, just in time learning experience. For this to be reality you really needed to make those LOs machine friendly, and so they became so overengineered and full of accompanying metadata, that no-one would create them, and they lost any sense of being an interesting subject for educators to engage with.
Definition debates – related to the above, the ed tech field debated endlessly what a learning object was. I mean, every paper started with their own definition. It was exhausting. For some it was ‘anything that could be used in a learning context’. This could be a photo, but it didn’t even have to be digital, it could be a stone. Which is fine, but doesn’t really get you anywhere. Other definitions were more general but specific to digital, and others had tight definitions around having a learning objective or meeting a specific standard. The problem this highlighted was twofold: Firstly, it highlighted the academic obsession with definitions to the point where most discussions degenerated into two men (it nearly always was men) shouting definitions at each other across a conference hall until everyone left and went to look for doughnuts. Secondly, the more specific definitions helped you decide what an LO was but ended up excluding too much, while the general ones included too much. The definition problem hinted at a more fundamental issue with LOs, which is next on the list.
The reusability paradox – David Wiley (it was through learning objects that I first encountered David, so they’re not all bad) got to the heart of the problem with LOs, and particularly the vision of automated assembly with the reusability paradox. He argued that context is what makes learning meaningful for people, so the more context a learning object has, the more useful it is for a learner. But while learners want context, machines don’t – in order for them to be reusable, learning objects should have as little context as possible, as this reduces the opportunities for their reuse. This leads to Wiley’s paradox, which he summarises as, ‘It turns out that reusability and pedagogical effectiveness are completely orthogonal to each other. Therefore, pedagogical effectiveness and potential for reuse are completely at odds with one another.’
An unfamiliarity threshold – we wanted LOs to be like reusable code, but the concept of sharing chunks of code was already familiar before it got formalised in object-oriented programming. And even then you learnt the concept as part of the language. LOs never achieved this for education, so the very idea seemed quite alien to many teachers, and particularly in terms of digital content. It began to look less like an ed thing and more like a tech thing. And you’ll never reach critical mass if that is the case.
The world wasn’t ready – you could argue, that like so many things, it takes more than one go at these concepts, each one building a bit on the momentum of the previous one. LOs didn’t take off, but OER did (to a greater extent anyway), and open textbooks more so. It’s possible LOs are ripe for a revival (or because ed tech only does year zero, rediscovery).
Education is too messy – this is probably just reiterating Wiley’s point about reusability, but in coding the boundaries are fairly well delineated (cue laughter from software developer friends). But education doesn’t break down so neatly. Particularly once you get beyond neatly defined concepts. To take Downes’s example, a sine wave LO might be easily reusable, but pretty soon the way I describe and illustrate even a shared concept will differ for PhD psychology students to first year undergrad engineers, partly because you know what they want to do with it (Wiley’s context again).
Reluctance from educators – as well as being unfamiliar, there was also a reluctance to share their carefully crafted material. This persists with OER – there simply isn’t the same culture of sharing for teaching as there is for research. This is largely to do with reward structures – you get promoted for getting your research paper cited by 1000 people, you get sacked for giving away intellectual copyright relating to teaching (I’m overstating, but you get the point).
They didn’t fail – while LO repositories may not be competing with Google for web traffic, you could make the argument that they didn’t fail. As mentioned above, they sort of morphed into OER, which sort of gave rise to MOOCs, and a lot of the LO work fed into standardisation around platforms, assessment, and content transfer. Publishers (shhhh) probably took the LO idea to heart more than others and have a large number of subscribers who pay for elearning content that can be redeployed in their context. LOs may be a successful failure after all.
PS – I tweeted that I was going to post on this, and Brian Lamb pointed me to a recent post of his, which sets out the LO lessons better than I managed, but I can’t abandon this post now.