Introduction

Several years ago I taught several "hands-on" courses on user interface design. In one exercise, students were given a specification, and used a prototyping tool to create a simple system. After the design solutions were completed, each individual in the class used everyone else's proposed systems to complete a task. Having experienced everyone else's ideas, the students then made changes to their original prototypes. The revised interfaces were always better than the original.

The three most interesting observations from these classes were:

• how many unique ideas (creative design solutions) individual students had initially,
• no matter how good were their original interfaces, every one could be improved, and
• how quickly students found and perpetuated good design ideas in their own products.

A few years later, Ovaska and Raiha (1995) published an article suggesting that having designers make initial design decisions independently, and then combining their results, resulted in far better user interfaces. They called this approach "parallel design."Five years later, Macbeth, Moroney and Biers (2000) found that having the original decisions made by several individuals was good, but that the original group then should evaluate all independent submissions and determine the best design solutions.

More recently, John McGrew (2001) from Decision Process Consulting published an article where he confirmed the validity of parallel design. He applied a parallel design process to develop a user interface for an invoice reconciliation program. To do this, McGrew scheduled a one-day session with several participants. He included the project manager, one person from the software and hardware design teams, two subject matter experts, a technical writer that was scheduled to do the training, three users and himself (a human factors engineer).

They began by having each person in the group independently sketch a proposed user interface on a large sheet of paper using colored felt-tip markers. The sketches then were posted on the wall for all to see and evaluate.

After viewing the design solutions proposed by others, each participant sketched two new designs. McGrew required that their new design include at least one idea from another person's design, and include an idea that no one else yet had proposed. Again, all the design solutions were reviewed by all participants. Participants began to agree on an optimal design fairly early in the process, and were able to reach a consensus on the final user interface design before the end of the day.

What is most striking, however, is that most linear processes would only have considered a few iterations of a single design. Using a parallel design approach like they did here, the design team considered 40 design alternatives before beginning the iterative process, i.e., before doing any usability testing. Consistent with my observations a few years ago, McGrew also found that participants responded immediately to good ideas. This was true even when good ideas were contained in otherwise poor design solutions.

Good user interface design requires designers first to "saturate the design space." This means that user interface designers should consider as many alternative design ideas as possible before selecting the best with which to begin the iterative process.