What if voice interfaces were designed using conceptual metaphors?

Description

Current Voice User Interfaces (VUIs) are designed with the idea of role reproduction, exemplified by the use of system personas. However, these VUIs do not take into account the complexities of “talking with a computer.” Users’ mental models for human-VUI conversations are based on human-human conversations, and this leads to heightened expectations and eventual conversational errors. Take, for example, the “desktop” metaphor used for GUIs. This model comprises secondary and auxiliary metaphors, which combine to give a unified desktop metaphor. For VUIs, the primary humanness metaphor does not allow for such complexities. A VUI is like a butler, an assistant, a librarian, and so forth. These metaphors can be amusing during the first interaction. Still, the novelty effect wears off during subsequent interactions as the facade of humanness cracks and the hostility of the machine becomes more apparent. These metaphors are especially ineffectual during conversational errors. How do you tell a VUI—designed to be a butler—that you meant to say “celery” and not “salary” without disassociating from the absurdness of the interaction? This project aims to address the problem of VUIs that are not dependent on human metaphors and do not match users' mental models for talking with VUIs. The proposed approach involves designing VUIs with a fluid metaphor approach that considers the context and appropriate metaphor for each interaction.