Feedback

This is an attractive theme that fits well with the heapsort modelling exercise. It would be good to complement the EM model of a tree introduced in that context with a tree model that expresses the different style of observation and animation associated with search. I think you could consider additional issues such as how far complete knowledge / comprehensive observation of the tree is required to carry out the search algorithms - presumably local knowledge at each node will suffice. On this basis, the possibility of a tree being dynamically generated as it is searched might be of interest. A possible extension of interest might involve linking the abstract searching activity to some specific context of use (such as surveying the nodes of a graph).

Good to try as far as possible to look critically at the claims for constructionism and learning. Am intending to make the recent paper on this theme (Computer Support for Constructionism in Context) available. If possible, it would be good to analyse your model-building as it proceeds, and see how far it bears out the idea that modelling and learning are closely linked. (Some of the learning will be about semicolons I fear.) It would be interesting to constrast this with what you think you might need to do to build similar models using other techniques.