Staging Natural Conversation: A Dramaturgical Analysis of Voice Assistants

• In Event: The Social Lives of Data: Attention, Agency, Exploitation

Abstract

Voice assistants are becoming active social actors in our everyday lives. Google Assistant, Apple's SIRI is integrated with our smartphones. We are controlling our smart home devices with Amazon's Alexa. These voice assistants try to mimic natural human conversations. While developing such Artificial Intelligent Systems, the developers have to consider two things: the multitude of ways a user can interact with it and that this interaction should feel “natural”. However, in reality, conversations involving human are never predefined and predictable. This uncertainty poses challenges for the voice assistants as they are faced with situations for which they are not well-rehearsed and thus struggle to have sustained interactions with users. Through dramaturgical analysis, Erving Goffman (1959) explains social interactions as if it were a play performed on a stage for an audience. This framework has been applied in the analysis of human-human interactions on online platforms (e.g., Hogan, 2010; Bullingham & Vasconcelos, 2013) and in the analysis of human-machine interactions (e.g., Bucher, 2014; Lee, Frank, Beute, de Kort., & IJsselsteijn, 2017). We extend Goffman's performative framework to ask, what are the different types of strategies the voice assistants can employ for "impression management"? How can we analyze these strategies without having access to the "backstage"? How do the conversational agents maintain decorum of expected behavior? As the “backstage” of both human and non-human remain inaccessible, we focus on the performative nature of the social interactions where voice assistants are treated analytically equivalent (Latour, 1996) to their human counterparts.

Authors

• Dipanjan Saha, International Institute of Information Technology Bangalore

• Bidisha Chaudhuri, International Institute of Information Technology Bangalore