[quoting Hannafin , 20 May 99] ...make no mistake about it: scaffolding of necessity becomes progressively less explicit and domain referenced as the enabling context becomes increasingly open and learner-defined. Whereas in externally imposed/induced, we can support cognitive processes and tactics linked to known domains/problems under study, in internally defined enabling contexts we can only support cognitive cognitive processes and tactics. The dilemma, of course, is that it is precisely this type of learning (informal education, informal learning) that most needs improved tools and scaffolds to support individually defined, often spontaneous learning needs. This is where the risks are greatest, but so too are the payoffs.
It seems to me that this is where Dave Merrill's ideas about knowledge objects will help. If our knowledge objects consist (roughly) of flat pictures, applets, and text, then the scaffolding has to be supplied by an external process that knows about the objects. If the objects know about themselves (that is, they are objects with descriptions within ontologies), then the user-interface can (theoretically) generate scaffolding based on the nature of the objects.
For example, suppose I want to learn about maps. If a kind of map object, a Mercator Projection, "knows" that it is a kind of map, that it can be calculated from a formulas (and formulas are blah, blah), and that rhumblines are straight lines on a Mercator map, and that rhumblines are concepts and that rhumblines are defined as X and that rhumblines can be instanced by Y, and that longitudes and latitudes are parts of maps, and that latitudes are concepts, and that etc., etc., then the scaffolding may, in principle, be calculated on the fly. As Dave has pointed out, we know something about teaching concepts and procedures and definitions. The interface takes the knowledge objects and according to the actions of the user, navigates a trajectory through a linked ontology.
Really no distinction between external or internal contexts need be made, since that is in the head of the user. If a teacher orders me to learn about Mercator Projections or I do it because I'm personally interested, the trajectories may be the same.
The downside of this is that specialized systems (where the knowledge is in the design, not the objects) will probably be "cooler."