The KMR-group is basing much of its development work on a kind of knowledge-pulling educational architecture that we call a Knowledge Manifold. A Knowledge Manifold consists of linked Knowledge Patches - each maintained by a custodian called a Knowledge Gardener. A KP in turn consists of a set of learning resources that are tied together with maps that represent the subject domain context model of the KG. Such context maps are preferably constructed using ULM, since the resulting context maps have clearly defined and verbally coherent visual semantics. This makes it easy to cognitively integrate the conceptual relations and achieve an overview of the context. Moreover, making the context visually explicit provides important support for the conceptual calibration activities that form an integral part of the learning process
The KM-archtecture defines (at least) seven different Knowledge Roles that can be taken dynamically by both teachers and learners. The first three of these knowledge roles involve construction, while the last four involve performance. Hence, in the KM architecture, there are three constructing and four performing knowledge roles.
There are at least three important reasons why these knowledge roles should include learners at all stages of their learning process.