The interpretation of symbols and gestures used in the design process is determined on the spot, through negotiated understanding, supported by personal experience, the social context, and the common understanding of practitioners in the field. Using computational facilities to support some of the manipulations of symbols in this process comes at a cost. In order to perform operations on symbols, computers require the interpretation of these symbols to be rigid and well defined. Yet, the role of computer tools should not be to constrict the fluidity of the design process nor the fluidity with which meaning is attached to the symbols used in this process. Instead, computer tools should make representation only rigid when necessary for computational operations, at the appropriate level of abstraction chosen by the user.

Thus the central idea of this work is to dissolve the static association between representation, its structure, and its interpretation and experiment with a dynamic user driven association between representation, structure and interpretation as the central mechanism for using computers.

Instead of choosing an application and then "filling in the blanks", this approach puts representation before structure and interpretation. Using a pen, the user draws marks on an electronic display surface, and only when needed, possibly at some later point in time, identifies a subset of the marks and applies some interpretation to it.

There are general questions which arise out of this approach and will be explored in this thesis. The three large areas in which these questions arise are: rigidity of information, mix of abstractions, and preservation of fluidity.

Rigidity of Information
Information needs to be rigid in order to be object to computational operations. Current computer applications require that all the information on which the user expects to perform operations are formalized a priori, when entering them in the system. A word in a word processor is a a sequence of characters when entered. A datum in a cell in a spreadsheet is a datum in a spreadsheet cell when entered. The approach taken here is opposite from this point of view. Any mark goes, and marks can later associated with interpretations that can either transform marks into other kinds of representations, or enable the execution of interpretation on these marks. This process can be reversed as well, disassociating interpretations with marks.

Usually computer applications exhibit very coarse grained structural requirements. In contrast this thesis in on the quest for finding fine-grained minimal structures that still allow interesting operations to be performed. By choosing such find-grained interpretations the user can apply the appropriate level of abstraction to the information at hand.

What are appropriate fine-grained abstractions? How would the user know which one to choose? Which ones are available? How does one deal with information that can not be processed by the interpretation choosen? How would one deal with deconstructing representations other than ink back into atoms?

Mix of Abstractions
In contrast to current computer system there will be a large mix between information on various levels of abstraction, or interpretation. The user will be able to perform operations to information in the context of its current interpretation. The association between information and interpretation can be nested.

How does one visuallize the interpretation associated with marks and the operations it entails? How does one visualize nested interpretations? How would the user know which gestures make sense in the current interpretation context? How would the user choose information to be associated with an interpretation?

Preserving Fluidity
The practical success of such a system depends on how well the fluidity of interaction can be preserved. The user has to be explicit about the structure (what marks to group together), and the interpretation (how to treat that group of marks). This is not as necessary on a regular whiteboard. Certain interpretations might need to be negotiated between members of a design group working on a whiteboard, but most associations between marks and interpretation are established through cultural context, or work practice. Will the advantage of having a computational component in the written media outweigh its added requirement for explicitness?