Working Memory, Search, And Signal Detection: Implications For Interactive Voice Response System Menu Design

Many researchers and speech user interface practitioners assert that interactive voice response (IVR) menus must be relatively short due to constraints of the human memory system. These individuals commonly cite Miller\u27s (1956) paper to support their claims. The current paper argues that these authors commonly misuse the information provided in Miller\u27s paper and that hypotheses drawn from modern theories of working memory (e.g., Baddeley and Hitch, 1974) would lead to the opposite conclusion that reducing menu length by creating a greater number of menus and a deeper structure will actually be more demanding on users\u27 working memories and will lead to poorer performance and poorer user satisfaction. The primary purpose of this series of experiments was to gain a greater understanding of the role of working memory in speech-enabled IVR use. The experiments also sought to determine whether theories of visual search and signal detection theory (SDT) could be used to predict auditory search behavior. Results of this experiment indicate that creating a deeper structure with shorter menus is detrimental to performance and satisfaction and more demanding of working memory resource. Further the experiment provides support for arguments developed from Macgregor, Lee, and Lam\u27s dual criterion decision model and is a first step toward applying SDT to the IVR domain