Ben’s rebuttal in our debate on context can be found here. I’ll be following this up offline with Ben, because there are definitely fundamental divergences in our views on this matter. In fact I’ve written a small thesis in response, but I’ll spare the general public from it. The gist of my reply is this:
- I am aware of Paul Dourish’s work, and I absolutely reject the idea that his so called phenomenological view of context is any different to what most computer scientists understand context to be. I think he’s cleverly narrowed the definitions contained in other works to exclude the concepts of relations between objects and actions, scoping (i.e. deciding what is relevant, which is something that I’ve already agreed is problem to be solved) and dynamic context definitions.
- Ben’s closing statement takes our debate from being about something very specific (i.e. context) to a more general discussion about UbiComp in general. I completely agree that we need solutions that support people (in fact the opening passages of my thesis state something very similar). But this is very high-level and abstract, and eventually concrete concepts (such a context) must be identified and dealt with if we are to actually solve any problems rather than just talk about them.
- Ben says that context is subjective because the way the world is interpreted right now is shaped by past experience and action. But that’s just another way of saying that current context is partly constituted by past experiences (i.e. historical contexts). Many people have thought about context histories and the way it affects the current context. (Note that I did not rig the results of Google Scholar: the highest ranked result just happens to be a Pervasive 2002 paper written by some people at ITEE/DSTC.)
- Just because something exhibits emergence does not mean it cannot be represented. And assuming, for the sake of argument, that something can’t be represented, then how will it ever lead to the betterment of the human-computer interface (because, eventually, if you actually want to progress a concept, it needs to be represented)? James Cole whose Ph.D topic poses the question "What is Information?" can probably shed more light than I can on the fundamental importance of representation.
- Finally, while it is always possible to find new facts, relations and context types that may be relevant to a specific context, the general definition of context does not change, regardless of whether you look at it from an engineering stand point, a HCI standpoint or any other standpoint.