XplanaZine

Complexity and Ontologies Part 3: Complexity as the Key to Intelligent Repositories

This is the second in a multi–part series on the process of developing ontologies in education and why they matter. These articles are a continuation of a series on taxonomies in education also published in XplanaZine (Part 1, Part 2, Part 3, and Part 4). This article talks about “complexity” and how understanding its principles will help us arrive at repositories of information and teaching materials that are intelligent and can anticipate our needs.

I envisioned this series as an attempt to help people understand how technology and information strategies can work together to give us “intelligent” storage and distribution for learning objects (aka quizzes, syllabi, video, audio, images, readings, etc.). This evolution is important , primarily, because we are going to see a dramatic shift in the creation, storage and distribution of education-based learning content over the next five years.

Specifically, we are going to witness a transition from LMS platforms as repositories of content to LMS platforms that serve as distribution shells for intelligent learning object repositories (LORs). In order for that evolution to occur, however, we still have to create LORs that have built-in information structures that are mapped deeply to education disciplines and their business needs. Such LORs will not only store our content more accurately but they will also be able to anticipate what we want to do with that. This, in turn will lead to automated course creation, multiple LMS and e-book distribution options, and to the semi-automatic and rapid development of complex learning objects.

Current education LORs have created a solid foundation to build on even though we’re still a ways off from the final goal. Representative LORs such as MERLOT, Careo, and MLX provide different levels discipline structure in their metadata (from basic subject area to sub-discipline support), object descriptions, author identification, object type identification, export possibilities, and user reviews. Each supports some level of standards-based metadata schema and all possess important baseline search functionality.

The next evolution in LOR functionality and information management — one that boasts standard taxonomies and ontologies to facilitate higher-order search and production — will require a greater level of complexity. These LORS will have to ingest, autotag, and store content easily as well as anticipate distribution and new collection possibilities on the fly.

In order to get there from here we will have to add new layers of complexity to our information systems and, in so doing, elevate the level of complexity in these systems. At this point, it would probably help to review the basic tenants of complexity and complex adaptive systems.

A complex adaptive system (CAS) is a system that contains agents whose job is to learn what they can about the system (by looking at their own instructions as well as by interacting with other agents) and to use that knowledge to “improve” their behavior. “Improve” here is defined in terms of the rules of the system. In the United States, for example, building a better automobile (improving the model) can mean many things depending on the system setting the rules. The government system defines an improved automobile as getting better gas mileage with lower emission output while an automobile dealership will define improvement by a car that will provide increased sales. The system sets the rules and the many agents that comprise the system all have to keep interpreting the rules and getting better.

The idea is that in such systems, given the right circumstances and information, the agents become “emergent” (they evolve), the system becomes more complex, and the system itself actually undergoes significant evolution.

When we talk about information systems and learning object repositories in particular, we can see that our current models are not sophisticated enough to have emergent complexity. To begin with, these systems are not constructed so that the various agents have enough interaction with each other to combine or improve their behavior schema. This is true for information structures such as taxonomies as well as for user roles and interactions. The systems are stuck without such enhancements.

Another limitation has to do with feedback. CAS depend on external and internal feedback to help agents adapt in ways that will make them better (help them improve). Feedback can come in the form of market popularity, voting, or electrical currents (depending on the system and its rules). With regards to LORs, feedback means, among other things, input regarding the coherence and applicability of ontologies, metrics regarding learning content housed in the LOR, e-commerce, and communal discussion boards.

In the current world of LORs, feedback is perhaps the most important element for better (or emergent) intelligence in our information structures. In other words, if we want our libraries to be smart, we have to feed them the correct information within the best set of rules so that they can learn.

This all seems straightforward enough. Next-generation LORs will require:

• Standard taxonomies and taxonomy structures (for education worldwide)
• Ontologies that provide the semantic logic and growth structure for the system
• Lots of users
• Varied feedback from multiple sources to ensure the proper evolution

With these elements in place, educational content will move into a whole new era.

Finally, one of the big issues facing us today is finding the balance and structure regarding input. When it comes to tagging content, for example, is it best to work from the bottom up (del.icio.us or furl) or from the top down (the system provides pre-set boundaries for tagging). The answer, most likely, will be found in LORs that combine both options. A certain amount of top-down tagging is necessary to maintain structure in a library environment. On the other hand, allowing users broad freedom to tag their own content provides much of the feedback necessary for the complexity we are discussing.