I thought a bit about the possible usage of the learning companion and how to frame its support in the learning experience. Basically I thought that it should not provide interaction regulation, a kind of mirroring system as proposed by Jermann (2004). On the other hand it should not work with a guiding principle, because not all the task/problems have a singular/simple definition of the task.

As the approach I propose is tailored in the distributed cognition theory, what this agent should do is essentially to propose examples of good strategy to solve the problem. The pace and the modality at which this should happen should be triggered by the learner activity in the simulation.

Very rough at the moment. I hope it will condensate into something more concrete as soon as I complete some more readings …