Two conditions of possibility for artificial ‘guides’:From detached descriptions to timely resources for action

Read full text Summarise with VIRAT
Authors: Damien Rudaz, Barbara Nino Carreras, Sara Merlino, Carl Joel Wester, Brian Lystgaard Due, Barry Alan Brown
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This presentation documents praxeological challenges currently at play during the prototyping of multimodal conversational agents designed to serve as assistants, ‘guides’, or co-participants for blind individuals. Using a conversation analytic approach, we illustrate some ongoing obstacles through the granular analysis of two representative fragments from a larger corpus: we compare the practices co-produced by an experienced human remote sighted assistant (RSA) and a blind participant—as they collaborate over the phone to locate a stain on a blanket—with those achieved when the same blind participant worked with a multimodal voice agent on the ‘same’ task, a few moments earlier. Relying on this heuristic contrast, on our own corpus, and on the existing literature, we highlight that among humans, most of the practices produced by sighted ‘guides’ rely on the ability to initiate (M. H. Goodwin & Goodwin, 2012; Keisanen, 2012) or modify (C. Goodwin, 1979; Mondada, 2016, 2024) turns-at-talk in real time, in response to emergent features of the visual environment—for example, the gesture of a blind person about to reach a desired location (Simone & Galatolo, 2020, 2021). In other words, these two fundamental mechanisms—the ability to initiate or modify turns-at-talk contingently upon the perceived environment—appear to constitute a necessary condition for a vast number of ordinary assistive practices initiated by sighted participants, and co-constructed with blind participants (Bouaouina et al., 2025; Due & Lüchow, 2023; Kreplak & Mondémé, 2014). We suggest that, so long as multimodal voice agents’ turn-taking models (Castillo-López et al., 2025; Skantze, 2021; Skantze & Irfan, 2025) do not enable them to initiate environmentally occasioned vision-based actions, they will lack a key resource relied upon by human ‘remote sighted assistants’ during object-centered inspection sequences (Tuncer et al., 2019). On this basis, we discuss several obstacles—related to temporality—to designing an artificial agent that can be mobilized by blind individuals within a non-representational approach (GahrnAndersen, 2024; Thrift, 2008): that is, a conversational agent that does not produce ‘scene descriptions’ (in the sense of formulations produced by a detached observer, removed from any practical urgency, who apprehends the environment as a set of objects with specific properties to be ‘put into words’; Bourdieu, 1977; Dreyfus, 1990), but, rather, that provides timely verbal contributions that can serve as resources for action (Linell, 1998; Mondada, 2018).

Fields of Study & Subjects

Language Discourse Communication Strategies Assistive Technology in Communication and Mobility Speech and dialogue systems

Publication details

Journal / source
Research at the University of Copenhagen (University of Copenhagen)
Publication date
1 January 2026
Language
EN