Répondre aux questions d’un robot

Abstract

Current trends in artificial intelligence and in the digitization of everyday life bring forward the idea that technological artefacts -such as humanoid robots, embodied conversational agents or voice user interfaces -might serve as autonomous assistants or companions in the home, in public and in work settings. From early on, it has been pointed out that their "success […] hinges not only on their utility but also on their ability to be responsive to and interact in a natural and intuitive manner" (Breazeal 2003: 167). As a basis, research in informatics/engineering endows robotic systems with advanced capabilities of monitoring their environment and for acting upon the users' communicational conduct. However, it remains a major challenge for technical systems to deal with the situated, contingent and in principle unpredictable nature of human communicational and interactional conduct (Suchman 1987, Button 1990, Schegloff 1996. In this article, I suggest that interdisciplinary cooperation between informatics/engineering and research on multimodal communication based on Ethnomethodological Conversation Analysis might help researchers to respond to this challenge and to create a novel research instrument for studying situated (inter-)action and human social practices (Pitsch et al. , 2014.To achieve this goal, a team of researchers are exploring how small groups of visitors -comprised of children and adults -engage in question-answer sequences 2 with an autonomous humanoid robot. As part of a larger research 1 The author acknowledges the financial support of the Volkswagen Foundation for the Dilthey Fellowship 'Interaction and Space' and from the Cluster of Excellence Cognitive Interaction Technology (EXC 277, project IP-18 conducted in cooperation with Dr. Sebastian Wrede). The author is indebted to her colleagues Sebastian Wrede, Raphaela Gehle and Timo Dankert for collaborative work on the project and the study investigated in this paper, to Lukas Hindemith and Hannah Pelikan for making the log files accessible, and to Vivien Ebben for her support in anonymizing the video stills.2 Stivers & Robinson (2006) distinguish between 'answers', i.e. specific types of responses which prototypically deal with the conditional relevance established through a question, and 'responses', i.e. anything that occurs (or not) subsequent to the previous turn. Here, I use the concept of 'answer' rather than the broader Postprint Karola Pitsch (2020): Answering a robot's questions. Participation dynamics of adult-child-groups in encounters with a museum guide robot. In: Réseaux, 220-221 (2-3), 113-150, https://doi.org/10.3917/res.220.0113. Postprint Version'response' as the human participants' try to answer within the conditions posed by the robot's questions -even though this is sometimes difficult.

“Pepper, what do you mean?” Miscommunication and repair in robot-led survey interaction

“…Second, coordinating who speaks when appears to be troublesome in interaction with robots, evidenced by overlapping turns and lenghty silences [4]. Third, in case of trouble, co-present others may "interfere" and explain a required action, for example to speak louder or to answer verbally [5]. Overall, users were found to adapt to robots over time, orienting to their capacities and adjusting their turns, as a result of interactional trouble and successes [4,20].…”

“…On the other hand, communication with robots is still difficult and rigid, arguably asocial. The human has to adapt to the robot [e.g., 4,5]. This article uses a multimodal conversation analytic approach [cf.…”

Staging Paro: The care of making robot(s) care

“…In another context, the mediating role of engineers (and at the same time investigators) has proven to be a decisive element of the 'situational dynamics' of humanmachine couplings between preverbal infants and the RUBI robot (Alac et al, 2011); although 'the robot [was] not under the absolute control of the roboticist …, when enacting [its] social character, the roboticists respond [ed] to the complex interactional trajectories often initiated by the toddlers' (p. 919). Based on a similar approach, another thought-provoking work was Pitsch's (2020) account of question-and-answer sequences in a museum, between visitors and Nao, an autonomous humanoid robot. She brought to light the way adults fulfilled a role of 'participation facilitators ' (p. 127), in charge of analyzing the structural prescriptions provided by Nao and supporting children in responding properly and at the right time.…”