'This is not what we wanted'

Abstract

Design is increasingly said to be about constructing conversations with end users [1]. Advances in underlying voice-related [2] technologies, coupled with the spread of voicedriven agents and dedicated devices such as the Amazon Echo, Google Home, and HomePod, lend weight to the notion of so-called conversational interfaces. In spite of the hyped anticipation of an AI-powered future, however, it is not always clear how the vision of conversation with machines measures up to lived reality, or if it is even relevant to actual design problems.

Progressivity for voice interface design

“…Clark et al [6] drill down on this notion through an elaboration of how "human-agent interaction [should be regarded] as a new genre of conversation" rather than as a simulation of human-human conversation. Further still, Reeves et al [20] argue that interaction with VUIs is not to be misunderstood as a conversation, and instead better understood as "sequentially organized moves around request and response". In the same vein, Porcheron et al [18] propose designing with the notion of responses from VUIs as resources for further user interaction with the VUI.…”

“…CA takes this basic observation and applies it primarily to talk (and secondarily, bodily action). EMCA has previously been applied to examine VUI use in talk-in-interaction [18,20]. We present fragments of data which make use of the Jeffersonian transcript notation [2].…”