Semantic structures for teaching and learning
I'm attending the JISC CETIS conference at Aston University in Birmingham over the next couple of days. One of the sessions that I've chosen to attend is on the use of the semantic Web in elearning, Semantic Structures for Teaching and Learning. A couple of days ago a copy of all the position papers by the various session speakers came thru for people to read - hey, I didn't realise I was actually going to have to do some work for this conference! :-)
The papers made interesting reading, all essentially addressing the question of why the semantic Web hasn't had as much impact on elearning as we might have hoped it would a few years back, all taking a variety of viewpoints and perspectives.
Reading them got me thinking...
Some readers will know that I have given a fair few years of my recent career to metadata and the semantic Web, and to Dublin Core in particular. I've now stepped back from that a little, partly to allow me to focus on other stuff... but partly out of frustration with the lack of impact that these kinds of developments seem to be having.
Let's consider the area of resource discovery for a moment, since that is probably what comes to mind first and foremost when people talk about semantic Web technologies. Further, let's break the world into three classes of people - those who have content to make available, those who want to discover and use the content provided by others, and those that are building tools to put the first two groups in touch with each other. Clearly the are significant overlaps between these groups and I realise that I'm simplifying things significantly but bear with me for a second.
The first group is primarily interested in the effective disclosure and use of their content. They will do whatever they need to do to ensure that their content gets discovered by people in the second group, choosing tools supplied by the third group that they deem to be most effective and balancing the costs of their exposure-related efforts against the benefits of what they are likely enable in terms of resource discovery. Clearly, one of the significant criteria in determining which tools are 'effective' has to do with critical mass (how many people in the second group are using the tool being evaluated).
It's perhaps worth noting that sometimes things go a bit haywire. People in the first group put large amounts of effort into activities related to resource discovery where there is little or no evidence of tools being provided by the third group to take advantage of it. Embedding Dublin Core metadata into HTML Web pages strikes me as an example of this - at least in some cases. I'm not quite clear why this happens, but suspect that it has something to do with policy drivers taking precedence over the natural selection of what works or doesn't.
People in the second group want to discover stuff and are therefore primarily interested in the use of tools developed by the third group that they feel are most useful. Their choices will be based on what they perceive to work best for resource discovery, balanced against other factors such as usability. Again, critical mass is important - tools need to be comprehensive (within a particular area) to be deemed effective.
The third group need users from the other groups to use their tools - they want to build up a user-base. The business drivers for why they want to do this might vary (ad revenue, subscription income, preparing for the sale of the business as a whole, kudos, etc.), but, often, that is the bottom line. They will therefore work with the first group to ensure that users in the second group get what they want.
Now, when I use the phrase 'work with' I don't mean in a formal business arrangement kind of way - as a member of the first group, I don't 'work with', say, Google in that sense. But I do work within the framework given to me by Google (or whoever) to ensure that my content gets discovered. I'll optimise my content according to agreed best-practices for search-engine optimisation. I'll add my content to del.icio.us and similar tools in order to improve its Google-juice. I'll add a Google site map to my site. And so on and so forth...
I'll do this because I know that Google has the attention of people in the second group. The benefits in terms of resource discovery of working within the Google framework outweigh the costs of what I have to do to take part. In truth, the costs are relatively small and the benefits relatively large.
Overall, one ends up with a loosely coupled cooperative system where the rules of engagement between the different parties are fairly informal, are of mutual benefit, evolve according to natural selection, and are endorsed by agreed conventions (sometimes turning into standards) around best-practice.
I've made this argument largely in terms of resource discovery tools and services but I suspect that the same can be said of technologies and other service areas. The reasons people adopt, say, RSS have to do with low cost of implementation, high benefit, critical mass and so on. Again, there is a natural selection aspect at play here.
So, what about the Semantic Web? Well, it suffers from a classic chicken and egg problem. Not enough content is exposed by members of the first group in a form suitable for members of the third group to develop effective tools for members of the second group. Because the tools don't exist, the potential benefits of 'semantic' approaches aren't fully realised. Members of the second group don't use the tools because they aren't felt to be good or comprehensive enough. As a result, members of the first group perceive the costs of exposing richer Semantic Web data to outweigh any possible benefits because of lack of critical mass.
Can we break out of this cycle? I don't know. I would hope so... and Eduserv continue to put work into Semantic Web technologies such as the Dublin Core on the basis that we will. On the other hand, I've felt that way for a number of years and it hasn't happened yet! In rounding up the position papers in her blog, Lorna Campbell quotes David Millard, University of Southampton:
the Semantic Web hasn’t failed, it just hasn’t succeeded enough.
That's one way of looking at it I suppose and it's probably a reasonable view for now. That said, I'm not convinced that it is a position that can reasonably be adopted forever and, with reference to my earlier use of the phrase "natural selection" it hardly makes one think of the survival of the fittest!?
What do I conclude from this? Nothing earth shattering I'm afraid. Simply that for semantic approaches to succeed they will need to be low cost to implement, of high value, and adopted by a critical mass of parties in all parts of the system. I suspect that means we need to focus our semantic attention on things that aren't already well catered for by the very clever but essentially brute-force approaches across large amounts of low-semantic Web data that work well for us now... i.e. there's no point in simply inventing a semantic Web version of what Google can already do for us. One of the potential problems with activities based on the Dublin Core is that one gets the impression that is what people are trying to do.
Again, I'm not trying to argue against the semantic Web, metadata, Dublin Core or other semantic approaches here... just suggesting that we need to be clearer about where their strengths lie and how they most effectively fit into the overall picture of services on the Web.