PI: Prof. James Landay, PhD

Abstract

Presentation tools are used by millions of people to communicate ideas. Much of this effort is design work, and yet most of these users have no knowledge of design processes. A powerful technique for producing good designs is parallel prototyping. Novices, unfortunately, do not know the benefit of this technique and often do not overcome the effort barrier to creating design alternatives. However, if a system automatically creates design alternatives and presents them to users, the users’ task shifts from creating alternatives to merely recognizing good alternatives. Previously, automatically generating well-designed slides was hard. Today, we can mine the web for large datasets of highly-rated slides and use machine learning to extract patterns to create alternatives. We will use these technologies to create a PowerPoint plugin that automatically generates design alternatives. Through quantitative experiments we will measure novice design recognition ability. Next, we will test the correlation between design recognition ability and the quality of novices’ designs. After introducing our tool, we will test whether automatically generating alternatives improves design quality. We will also determine the key times to present alternatives to maximize quality. This new tool introduces novices to parallel prototyping and enables them to improve their designs.