The primary question in the design of Goldberg is "What is the appropriate metaphor for the interaction between the composer and Goldberg?" Because we started with a vision of how Goldberg should work, this fundamental question did not occur to us until we started analyzing survey results and discussions. It became clear that the original metaphor (an intelligent music assistant that makes suggestions) would turn off a large portion of our potential audience (the traditional composers who disdain computer composition).
We then recognized that there is a spectrum of possible interaction styles ranging from an in-control human and obedient computer to the other extreme of an intelligent computer and passive human who just chooses among options.
We decided the appropriate way to address this question was to create mockups of a range of interfaces along this spectrum and test them on composers of different styles.
An equally important, though less fundamental, question is "How should Goldberg present the operations it can perform and the variations it has created?"
The basic Goldberg idea was inspired by a system for interactively creating images. In that system, it was easy to present the user with a large number of choices for her to choose among, because the human visual system can scan many things quickly and light on a few interesting ones. With music, however, we might have to listen to the choices one at a time to be able to rate them. This is potentially time-consuming and limits the number of choices that can be presented.
These issues raise a number of questions that can be addressed in user studies:
The interfaces we tested on people attempt to explore these questions by employing different solutions.
