Talking with machines? — Voice UI and conversation design

There is much excitement about conversation as a new material for design, driven in part by the increased accessibility of voice user interfaces and commoditisation of AI techniques. As part of increased adoption, devices like the Amazon Echo, Google Home and Siri are providing platforms for designers to interact with users in new ways. In spite of this (often hyped) anticipation of an AI-powered future, it is not always clear how the vision measures up to lived reality of ‘having a conversation’ with machines.

I will offer a window into work being done at the University of Nottingham that is empirically examining how voice UIs like the Amazon Echo actually come to be used in social settings. Myself and colleagues (Martin Porcheron and Joel Fischer) are capturing naturalistic recordings of Echo use in participants’ homes, which enables us to start building a very rich picture of how users actually ‘get stuff done’ with voice UI, as opposed to ‘have a conversation with it’. In transcribing and analysing our recordings in detail, we get a close and perhaps surprising look at how requests are formulated by users in particular ways that are sensitive to the device and the setting. Conversely, we examine responses produced by voice UI and how, in a range of practical ways, users make them ‘fit into’ the unfolding flow of talk and the broader activities that form the life of the home.

Our findings lead us to several conceptual challenges for designers tackling voice UI to develop ‘conversations’ with users.