Feedback

This is an imaginative idea which lends itself to empirical modelling in the standard informal sense. It will be necessary to think carefully about where observation dependency and agency fit into the picture - it will be important for assessing your paper and model to know whether the same features/results could be obtained just as effectively (perhaps even more effectively!) by using programming tools other than tkeden. As I read it, your focus is on monitoring the patterns of navigation, and trying to make inferences from the empirical evidence. You may find a model such as trafficlightMendis1997 useful here, as it invokes a similar interpretation of 'empirical analysis', and could give you ideas about how to simulate visits automatically etc. You might also want to consider the issue of whether 'counting visits to a page' is subtle enough to give the insight you need into web-design, and discuss / address appropriate extensions of your model to deal with this. We wouldn't expect experienced users to consult the Tools pages of the EM website often for instance - indeed once to download, and occasionally to browse for other tools and manuals might be quite reasonable. Perhaps EM could offer support for a more semantically interesting model than mere statistical analysis can support, and it would be good to say something about this theme.

The discussion of real / fictional websites is a little confusing. I assume that you mean the model to be developed in conjunction with some actual website to which it refers, whether or not the website is drawn from a real application or contrived by the modeller. For ths purpose of your project, it would probably be good to have a number of specific examples in mind: a very small-scale toy site for basic illustrative purposes, something a bit more ambitious that does justice to the potential of your model as far as you are able to build it within the scope of the assignment, and an actual site that you could refer to as an indication of what might be possible in future extensions (perhaps with proposed changes to tkeden itself).

Your reference to the COG model is interesting - I assume that you are thinking of the construction of graphs through direct manipulation. You can probably get away with building much simpler utilities for the same purpose (the 2-view Donald editor - 2vde in wmb/public/projects/tools/editors may be one prototype for such). Another option that might suit your application (and further justify the weighting towards model-building) would be building your own notation for describing websites abstractly.

'amount of visits' -> number of visits
'principle area' -> principal area
'little or no visits' -> few or no visits