A little more on the afterthought
I’ve mentioned before yesterday that I thought prior use context was among the most important types of information metadata could capture about a particular piece of learning material, and also that capturing prior use context is among the most difficult tricks to accomplish in metadata. That’s because you are trying to capture networked relationships, which works fine when those relationships are between people, each of whom is willing to enter his or her metadata. Unfortunately, instructional materials aren’t so socially motivated.
As I said in the post from last August, I had hacked together a metadata capture database before coming to MIT that used FileMaker’s lookup function to partially automate the creation of metadata records and capture relationships between materials in the process. The level of effort was still not trivial, but you could at least fairly painlessly develop metadata records at several levels of granularity for a group of instructional materials. Borrowing from a David Wiley paper, I used four levels of granularity, which I called Course, Unit, Resource and Object. By creating the top-level granularity record, and then creating child records down the chain using the lookup feature, you could have the smaller granularities “inherit” much of the necessary metadata, and only have to make a few changes.
The best part, though, was that by concatenating the ID’s for parent-child strings, you could develop a “DNA” for the material in question that expressed its relationship to other materials. For instance, the system could spit out a string like 25253-67465-47577-47773, which essentially told you that object 47773 (say, a .jpeg) was a part of resource 47577 (an .html page), which in turn was part of a series of related pages 67465 (a unit on narrative arc) that was one unit with a course 25253 (say, Introduction to Creative Writing: Fiction). There was no requirement within the system that the chain start at the course level–if the unit was the top of the chain, the DNA would look like: 00000-67465-47577-47773. A stand-alone object would look like 00000-00000-00000-47773. You could express a whole unit as 25253-67465-00000-00000.
Once the records for the materials’ initial use were created, it was possible to adopt previously created materials into new higher-granularity materials. For instance, Advanced Fiction Writing (88326) might adopt the whole unit on narrative arc, expressed as 88326-67465-00000-00000; or just a page in that unit for a new unit on narrative structure (93326), expressed as 88326-93326-47577-00000; or simply adopt the image on a new page on narrative structure (67688), expressed as 88326-93326-67688-47773.
Through all of this it was possible to query a series of DNA strings related to a given piece of material, such as:
This query can be done on records of any granularity, providing access to a record of all the uses to which the material has been put over time.
I drag you through all of this because, as I’ve also said before, the workflow tracking engine we use at MIT OCW is based on this system, uses a very similar parent-child granularity tracking and generates similar chains. In developing the MIT version, we didn’t at the outset implement the “adopt” feature, but have just recently done so, largely to track intellectual property metadata for objects we’ve previously obtained permission to publish and are using for a new course.
There are now several hundred such cases, many of which are associated with updates of the same course (i.e. the Spring 2005 version uses many of the same images we cleared for the Fall 2002 version), but other instances are old materials being reused in totally different courses. The primary interest here, of course, is keeping the IP straight and eliminating the effort of clearing the same object twice, but eventually there will be quite a bit of valuable data on the reuse of educational materials that can be drawn pretty easily out of the system. I’m looking forward to being able to look at this down the road.