Book chapter 475 views
Conversational AI: Respecifying Participation as Regulation
Stuart Reeves 
,
Martin Porcheron
,
Martin Porcheron
Handbook of Digital Society
Swansea University Author:
Martin Porcheron
Abstract
When we talk about AI-driven systems there is a tendency by researchers to treat people encountering them as ‘participants’ in human-machine interactions. This seems particularly true for so-called conversational AI, such as voice interfaces or chatbots. The pervasiveness of this position is encapsu...
| Published in: | Handbook of Digital Society |
|---|---|
| Published: |
SAGE Publications
|
| URI: | https://cronfa.swan.ac.uk/Record/cronfa60090 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Abstract: |
When we talk about AI-driven systems there is a tendency by researchers to treat people encountering them as ‘participants’ in human-machine interactions. This seems particularly true for so-called conversational AI, such as voice interfaces or chatbots. The pervasiveness of this position is encapsulated by the popular adoption of Nass et al.’s statement that “Computers are Social Actors” (CASA), which argues people are “mindlessly” applying human “social scripts” to AI systems; in other words, people act like participants as a kind of social reflex action. We think this is mistaken and find that a cursory look at actual interactions with (in our case) conversational AI systems reveals a different picture. Taking an ethnomethodological and conversation analytic perspective, we present a series of recorded fragments of people interacting with domestic voice interfaces. These show the organised ways in which conversational AI systems are embedded into everyday action. In doing this we reframe people’s use of interactive AI technologies: far from being mindless or perfunctory, interactions with conversational AI are inextricably situated and interwoven with the sociality of a setting. Crucially, we show how AI systems are regulated within that sociality, via a wide range of practical (in our case conversational) methods. Understanding mundane regulatory work, then, is more pressing from a design perspective than working out how to design AI-driven systems to be better ‘participants’. |
|---|---|
| College: |
Faculty of Science and Engineering |