Konuşmacılar
Açıklama
In digital discourse studies, “blended data” approaches that combine an analysis of digital content with an ethnographically grounded context are increasingly popular (Androutsopoulos, 2021; Varis & Hou, 2020). The main advantage of this methodology is that it provides background information on users, their positionings and thus paints a more comprehensive picture of the digital data. There are several issues that arise when attempting to apply such an approach to study conversational AI data where technologies, humans, and languages are entangled in heterogeneous ways.
In this talk, I discuss the methodological challenges and ethical issues in the study of voice-based conversational AI and propose ways of approaching and analysing discourse data as co-curated by humans and algorithms. Drawing from an ethnographically informed study with multilingual voice assistant users (e.g, Siri) as a case study, I examine the practices of Turkish-speaking migrants in Germany with their machines, how they reflect on their practices, and how their utterances are translated into text by algorithms. The data collected for this study comprises interviews, user archive data from the Alexa app as well as online and offline participant observations.
Privacy is a significant concern in this context, with users unaware that their voices are stored and transcribed each time their devices are activated, at times without their intention. These transcriptions only partially reflect the situation, as users’ utterances may have not been recognized correctly and many conversational cues important for contextual information are not transcribed (Habscheid et al., 2021). It is thus not plausible to treat these data as representing user practices as in blended data approaches. Rather, it is necessary to critically examine the types of texts algorithms produce and store, the social indexicalities of these texts, and how they shape user practices or remain concealed from them (Jones, 2016).
Keywords: Conversational AI, voice assistants, digital discourse analysis, sociolinguistics, methodology
Institution / Affiliation / Kurum
European University Viadrina, The Faculty of Social and Cultural Sciences, Linguistics Department
| Presentation language / Sunum Dili | EN (English) |
|---|---|
| Disciplines / Disiplinler | Linguistics / Dilbilim |
| E-mail / E-posta | leblebici@europa-uni.de |
| ORCID ID | 0000-0002-7104-2289 |
| Country / Ülke | Germany |
