Exploring the context of use for voice user interfaces: Toward context-dependent user experience quality testing

Voice user interface (VUI) systems, such as Alexa, Siri, and Google Assistant, are popular and widely available. Still, challenges such as privacy and the ability to have a dialog remain. In the latter example, the user expects a human-like conversation, that is, that the VUI understands the dialog and its context. However, this VUI feature of context-aware interaction is rather error prone. For this reason, we intend to explore the VUI context of use and its impact on interaction, that is, relevant user experience (UX). We see a demand for context-dependent UX measurement because analyzing the context of use and UX assessment are both critical human-centered design (HCD) methods. Therefore, we examine the VUI context of use by asking users about how, where, and for what they use VUIs, as well as their UX and improvement proposals. We interviewed people with disabilities who rely on VUIs and people without disabilities who use VUIs for convenience or fun. We identified VUI context-of-use categories and factors and explored their impacts on relevant UX qualities. Our result is a matrix containing these elements; thus, it provides an overview of the contextual UX of our target group's VUI interaction. We intend to develop a VUI context-of-use conceptual structure in the future based on this matrix, which is needed to create an automated context-dependent UX measurement recommendation tool for VUIs. This conceptual structure could also be useful for automated UX testing in the context of VUI. Voice user interface (VUI) systems such as Alexa are popular, but challenges remain, for example, privacy and fully understanding the dialog and its context. Therefore, we explored the VUI context of use and its impact on user experience (UX) by interviewing people with and without disabilities. We examined their interaction (e.g., dictation) to identify categories (e.g., environment) and factors (e.g., accent) and their impacts on UX (e.g., understanding). Our results help to meet the demand for automated context-dependent UX measurement recommendations. image