Guiding The Behavior Design Of Virtual Assistants

In recent years there has been an explosive proliferation of virtual assistants such as Amazon's Alexa or Google's Home Assistant. These virtual assistants are becoming increasingly sophisticated and able to manage many aspects of our lives. Through bundled sales campaigns, consumers are encouraged to install multiple devices throughout the house and to share their use across the household. Unfortunately, competing devices do not behave consistently with each other. Furthermore, the increased popularity of third party behavior extensions (e.g. Alexa Skills [1] or Google Home Actions [5]) has the potential to add to this behavioral discontinuity. Ambiguities in multi-user behavior combined with disparate agent and extension behavior can lead to consumer confusion when the agent does not behave as expected. This report documents ongoing research intended to identify users' inherent assumptions about a smart agent's behavior through user studies. The results can be applied to inform design guidelines that lead to a more consistent user experience, with agents clarifying their behavior in situations identified as ambiguous.