Towards Adaptive Generation of Faded Examples

Faded examples have been investigated in pedagogical psychology. The experiments suggest that a learner can benefit from faded examples. For realistic applications, it makes sense to generate several variants of an exercise by fading a worked example and to do it automatically. For the automatic generation, a suitable knowledge representation of examples and exercises is required which we describe in the paper. The structures and metadata in the knowledge representation of the examples are the basis for such an adaptation. In particular, it allows to fade a variety of parts of the example rather than only solution steps.