Studies of smart speakers highlight issues people face with understanding why unexpected behaviour occurs and with recovering from mistakes due to uninformative responses. Yet, our understanding of such intelligibility issues in smart speakers -difculties in understanding the device's behaviour -remains limited, in particular, for long-term and frequent smart speaker users who may encounter more complex situations than frst-time users. We conducted an online survey and interviews with smart speaker enthusiasts to investigate how they form an understanding of the device's behaviour and what strategies they use to recover from breakdowns. We identifed seven diferent breakdown recovery strategies and found that enthusiasts particularly struggled with breakdowns in their IoT infrastructure. Informed by our results, we propose three research directions: infrastructural breakdowns as learning opportunities for understanding the smart speaker's behaviour; leveraging aspects of non-verbal communication as opportunities for design; and considering passive users' intelligibility and control needs. CCS CONCEPTS• Human-centered computing → Human computer interaction (HCI); Empirical studies in HCI; Human computer interaction (HCI); Interaction paradigms; Natural language interfaces.

Findings: 6 Themes 1 | Mirroring, mimicking motions Dynamic Smart Speaker with Expressive Physical Motion User study Study participant Wizard-of-Oz lab study: 12 participants interacted with QUBI in 4 scripted scenarios + Questionnaire/interview Examples of expressive physical motion Nodding Wiggling Pointing QUBI 2 | Body language with voice 3 | Anthropomorphism 4 | Audio trumps motion 5 | Reaffirmations 6 | Emotional reactions Study environment QUBI, Wizard-of-Oz controlled 12 participants in total for study Interpretation of motionFigure 1: Visual abstract summarizing research of QUBI -a smart speaker design with multiple degrees of freedom for expressive physical motion, and key findings of our user study about people's interaction with QUBI.

In this paper we activate theoretical principles for HCI and CSCW in analysis and design of collaborative technologies. We critically evaluate theoretical principles about how technologies mediate collaborative activities through two empirical cases regarding smart speakers and collaborative writing, respectively. Common to them is that both common objects and mediating artifacts are set in a mixed virtual and physical setting. Principles from activity theory are examined, specifically with a focus on how people use artifacts and objects as mediators to collaborate in joint activity. As part of questioning the cases, we also position collaborative affordances in relation to the principles. We engage with the cases by using the theoretical principles both analytically and constructively. Our analysis specifically targets traces as a possible construct that has potential to constructively aid collaboration across the two cases. We explore what seems to be a lack of support, and in particular theoretical framing, of traces of past activities, connected objects, other users, etc. With an initial hypothesis that traces could be a mechanism that would support these two kinds of collaborative activities better in a future redesign, we set out to further explore the cases and their future possibilities, and how they could be conceptually grounded and supported. We present a set of analytically and constructively oriented questions to help researchers and designers of artifacts identify mediators for and instances of collaboration.

Voice-controlled smart speakers with intelligent personal assistants (IPAs) are increasingly becoming ubiquitous in homes. They play a key role in home automation as a hub that interfaces with various appliances. Previous work has suggested that ubiquitous computing systems should be intelligible and controllable, by informing users about the system's underlying behavior and enabling users to intervene during breakdowns. In my PhD, I investigate physical intelligibility for smart speakers: the use of physical motion and interaction to provide intelligibility and control, informed by observations of users approaching and physically interacting with smart speakers.