Configuration of user environment is probable the closest thing to profiling the regular user comes. I.e. the user chooses background image and other desktop specifics, in applications the user chooses which tools to be visible and similar stuff. This is just personalization of the tool, intended to let the user feel more comfortable and hopefully also interact more efficient.
Taking a step further to allow profiling to simplify the task performed in the tool is another matter. For this to work in general some understanding of the task from the profiling software side is required. This is in general to hard and is closely related to the field of Artificial Intelligence. Intelligen Tutoring Systems (ITS) is a subset of the AI-field that deals mainly with user modeling and how to design clever (heuristicall or statisticall) algorithms that extrapolates and guesses what the next step in the learning process should be.
Today there is various attempts to attac thess problems, P3P and VCard ar just to examples of standards that are relevant for expressing personal preferences as well as vital user information.
This project focuses on investigating how to do profiling in an information search, retrieval and presentation aspect. Together with emerging semantic web techniques this restricts the general problem to a manageble size. The idea is not to do the final solution but rather investigate existing standards and techniques that could be used and provide a basic pluggable architecture where different profiles and strategies could be tested.
Conzilla have some features that could be used, actually the filtering fetures could be used as a starting point for simple profiling. The meta-data support should definitely be used and applied on the user as well, maybe use VCard for a start.
A closely related approach would be to use the query capabilities, i.e. adapt or extend queries depending on the user profile. Some thought has to be put into the differences between querying, filtering and adaption. There is strong connections to the context-search project, i.e. searches being adapted in a way that depends on the context.