Forschungs- & Publikationsdatenbank |  |
         | |
  |
|
Publikation Nr. 3954 - Details Even, C. (2023). Users' adaptation processes when learning to talk with a conversational interface, ICCA 2023, University of Queensland, Brisbane. URL: https://www.conversationanalysis.org/icca-2022/ Abstract Users’ adaptation processes when learning to talk with a conversational interface Conversational interfaces (CIs), like Amazon Alexa, Apple Siri, or Google Assistant, are software that are supposed to “enable people to interact with smart devices using spoken language in a natural way—just like engaging in a conversation with a person.” (McTear et al., 2016, p.1). Talking with CIs is often considered being easy and intuitive (e.g., Siegert, 2020). Yet, in recent studies, users report of having had to learn how to talk with a CI (e.g., Pins et al., 2020). Insights into a learning process are, however, mostly drawn from retrospective self-reports in interview studies. There are only few articles that present a micro-analytical reconstruction concerning user adaptations (e.g., Pelikan & Broth, 2016; Velkovska & Zouinar, 2018). In these articles, users’ adaptation work is described and analyzed for problems in terms of turn taking and repair, leading to users’ adjustments in form of reducing the complexity and length of turn construction in subsequent sequences. This presentation is going to update previous findings by presenting complementary user practices of adaptation. Based on multi-perspective video recordings of 20 volunteers (aged 25 – 72), who are novice users of the CI Apple Siri, it is investigated how they gradually adapt – or struggle to adapt – to the CI’s interface structures when entering appointments. Analyses of the sequential and sequence organization are carried out following the multimodal conversation analytic approach (e.g., Mondada, 2008). Findings show that users initially transfer social interaction patterns to their first exchanges with Siri. Most users (except of a sub-sample of six older adults) then deviate from natural human conduct due to recurrent trouble sources and repair sequences: 1. Lexical changes in turn construction: a) Linguistic alignment (Branigan et al., 2010) in repair initiations: Users change the way of initiating repair for entering the appointment’s activity. They adopt expressions used by Siri after having encountered trouble before when speaking naturally. b) Suppression of linguistic alignment in sequences of temporal information transfer: Users avoid repeating or recycling the interlocutor’s lexico-syntactical choices when answering Siri-requests for temporal information, as such linguistic alignment leads in these cases to an unintended task shift by Siri and ensuing repair. 2. Changes in sequential ordering: Users deviate from the interactional pattern of delivering type-conforming answers to questions after having encountered trouble: Instead, they add non-requested information in order to avoid otherwise ensuing repair. Or they deliver self-initiatedly information in earlier sequential positions. Branigan, H. P., Pickering, M. J., Pearson, J., & McLean, J. F. (2010). Linguistic alignment between people and computers. Journal of Pragmatics, 42(9), 2355-2368 McTear, M., Callejas, Z., & Griol, D. (2016). The Conversational Interface: Talking to Smart Devices. Springer Mondada, L. (2008). Using Video for a Sequential and Multimodal Analysis of Social Interaction: Videotaping Institutional Telephone Calls. Forum Qualitative Social Research, 9, 1–35 Pelikan, H. R., & Broth, M. (2016). Why That Nao? How Humans Adapt to a Conventional Humanoid Robot in Taking Turns-at-Talk. In CHI'16, 4921–4932 Pins, D., Boden, A., Essing, B., & Stevens, G. (2020). "Miss understandable" - A study on how users appropriate voice assistants and deal with misunderstandings. In MuC'20, 349–359 Siegert, I. (2020). “Alexa in the wild” – Collecting Unconstrained Conversations with a Modern Voice Assistant in a Public Environment. In LREC 2020, 615–619 Velkovska, J., & Zouinar, M. (2018). The Illusion of natural conversat Attribute:
|