Virtual assistants in smart homecare: An interactional qualitative analysis

With the aim of tackling the national care ‘crisis’, assistive technology, particularly virtual assistants (VAs), are often framed as accessible ‘solutions’ that enhance service users’ independence, provide access to medical support, and alleviate loneliness. Marketing claims emphasise the use of ‘natural’ interaction to control these devices, presenting their interactions as human-like and seamless. This discourse portrays VAs as conversational partners which conversational analytic studies question, reconceptualising interactions with technology as structured and machine-oriented which users must adapt to. Accordingly, within this thesis, I do not focus on the ostensible technological or medicalised functions of VAs, aiming to ‘fix’ service users. Instead, I examine their interactional and practical affordances in fostering collaborative care environments that can enable service users to engage in their care routines and decisions. Specifically, I focus on how such environments are shaped through interactional achievements of autonomy, interdependence, and companionship, and how VAs contribute to these dynamics, drawing on the social model of disability that views ‘dis-abling’ environments, not impairments themselves, as a fundamental barrier to participation. I also explore participants perceptions of their daily use of VAs in care settings to uncover discrepancies between what participants ‘think they do’ compared to what ‘they actually do’.

To examine these relatively under-explored areas, this empirical thesis draws on two overarching datasets: (1) interviews with older adults and disabled people who use smart technology in homecare settings, focusing on technology as underpinning interdependence (Chapter 4); and (2) video recordings of participants in their homes using VAs in their daily routines with their care assistants or alone, focusing on how (if at all) the VA takes on a role recognisable as ‘doing companionship’ (Chapter 5), and the organisation of repair in resolving breakdowns in interactions with VAs (Chapter 6). To examine these data, I employ a triangulation of three approaches. Firstly, reflexive thematic analysis is used to explore participants’ perceptions of their VA use and their constructions of the importance of collaborative care for technology use (Chapter 4). Secondly, conversation analysis as a methodological and analytic approach is used to explore and consider specific moments and routine methods in the talk, visible behaviours, and spatial arrangements of the human-VA interactions (Chapters 5 and 6). Finally, discursive psychology is paired with conversation analysis (in Chapter 5) to examine ‘psychological’ concepts (i.e., companionship) that are practically implicated in certain interactional practices that accomplish or relate to the concept.

Reflexive thematic analytic findings indicate that VAs are constructed as facilitating collaboration within the care routine. Participants alluded to the challenges they faced with their VAs, particularly when their care assistants were not available to assist. They also discussed privacy concerns as an emic dimension, rather than ‘virtual privacy’ of standardised data protection and information sharing risks, as underpinning autonomy and interdependence. Privacy was constructed as a practical concern of personal space, linked to the practicalities of care, meaning that participants could perform private tasks or spend time with others using their technology, with a care assistant available, but not physically co-present.

Conversation analytic findings, with discursive psychology, highlight interactional concepts such as burden and blame when facing interactional difficulties with VAs for supporting agency, interdependence, and collaboration in care settings. In both Chapters 5 and 6, the majority of the ‘interactional burden’ lies with the user when managing miscommunication involving VAs. In interactions where a care assistant was co-present, this burden often constituted a resource for soliciting and offering collaborative care, creating opportunities to provide assistance without diminishing a disabled person’s sense of agency. Blame attribution was vivid in both Chapters 5 and 6, where findings show that when alone with a VA and repair sequences are longer and often unresolved, actions such as complaints about and blamings directed towards the device are achieved through shifts of ‘footing’ which, as I argue, constitute a ‘companioned practice’ for alleviating anxieties and loneliness (Chapter 5). However, in multi-party interactions including a VA, blaming (through prosody, intonation, turn design, and embodied movements) provided a resource for agency, where care assistants can shift blame for failed attempts to use the VA away from the service user and towards the device.

Overall, this thesis contributes deeper knowledge to the affordances of VAs and smart technologies in the care sector. I conclude that these devices can provide resources for building collaborative care infrastructures and users’ autonomy without the need for high-tech, cutting-edge technology that claims to emulate human emotions, action, and replace human care.