Finding a Home for Voice Assistants: A Domestication Calculus Across Three Years and Thirty Households

Source: Mahla Alizadeh, Minha Lee, Dave Randall, Peter Tolmie, Dominik Pins, Gunnar Stevens; CHI 2026 · 2026-04

Abstract HCI has explored voice assistant (VA) use across various social settings, highlighting their impact on personal and familial dynamics. Yet, the progressive domestication of these devices over time and their longer-term impact on relationships remain underexplored. We present findings from a three-year study of 30 households using interviews and diaries. Our analysis introduces the concept of a domestication calculus that captures how VAs find—or fail to find—a home over time through shifting spatial arrangements, relational roles, and household routines. Domestication unfolded not as a linear sequence of stages but as a dynamic process in which devices were either embedded into routines, withdrawn from use, or repurposed in response to changing circumstances. Across these trajectories, participants attributed four recurring roles to their VAs: (1) negotiators, (2) separators, (3) mediators, and (4) amplifiers of shared life. We conclude with implications for designing VAs that support long-term domestication.

Introduction. Voice Assistants (VAs) are rapidly becoming an integral part of the daily lives of many. At scale, this shift is evident in recent estimates showing that Google Assistant and Amazon Alexa are embedded in over 8.4 billion devices worldwide, alongside rising household adoption, including projections of digital assistant ownership exceeding half of households and smart speaker ownership reaching up to 75%. As modern-day VAs evolve—offering functions from managing smart homes to entertaining with family games, jokes, and stories—their influence on family life and interpersonal dynamics within households becomes more and more pronounced.

Historically, HCI research has examined VA adoption in the home and their impact on family dynamics (e.g., [28, 36, 65, 83, 89]). For example, Garg and Sengupta analyzed interview data and activity logs from 18 families with children to compare the usage patterns of adults and children. Beirl and Rogers observed that VAs nurtured specific family dynamics, especially in directing children’s interactions. Similarly, Lee et al. showed that VAs could positively contribute to family life by increasing conversations among household members, while also becoming anthropomorphized in the process.

However, studying VAs in the household context is not without challenges. Over time, these tools undergo a process of “domestication”, seamlessly integrated into the tapestry of home life and its intricate web of relationships. As Harvey Sacks originally remarked, “the technical apparatus is, then, being made at home with the rest of our world. And that’s a thing that’s routinely being done ... what happens is that the object is made at home in the world that has whatever organization it already has” ([92], p. 548). In other words, technologies do not remain fixed in their use: how and by whom they are used shifts as they move from novelty to routine, continually adapting to the changing circumstances of household life.

A few studies have provided glimpses into how VA use changes in its early stages. Cho et al. studied eight first-time Alexa users over twelve weeks and identified stages in their experiences—preadoption, adoption, adaptation, stagnation, and acceptance—finding that use declined significantly as novelty waned. In a two-week study, Voit et al. distinguished between households that quickly abandoned VAs and those that integrated them into daily routines. Analyzing conversation logs from assisted-living communities over two years, Oewel et al. found that older adults engaged in personal and relational conversations with VAs, even disclosing sensitive information over time.

Although existing studies —including a few longer-term ones—demonstrate that VA use can shift over time, much of the domestic research has examined relatively short periods, often weeks or a few months. These perspectives capture adoption and early adjustment but rarely follow households long enough to observe how longer-term domestication unfolds across extended trajectories and varying household circumstances. Consequently, we still know little about how VAs are renegotiated once their novelty fades or how their roles evolve over longer trajectories. This gap is increasingly important as VAs gain new capabilities through large language models (LLMs). Early work already points to agents taking on roles in family coordination and intergenerational communication, and LLM-enhanced systems are expected to support increasingly complex, relational, and emotionally responsive forms of cooperation. Understanding domestication over time is therefore increasingly important for anticipating how these evolving capabilities intersect with shifting social contexts and when VAs ultimately become embedded, transformed, or abandoned in everyday life.

To address this, we ask:

• RQ1) How do people’s experience of VAs at home unfold over time as they progressively domesticate them? • RQ2) What impacts do VAs have in shaping or altering the domestic life over time?

In our research, grounded in qualitative traditions common within HCI and CSCW, we conducted a longitudinal study of thirty households over three years. We combined iterative interviews with photographs, home maps, and a three-week diary study in the project’s second year, enabling us to trace not just moments of use but also changes in placement, roles, and routines over time. This multimodal approach allowed us to capture the spatial and temporal movements of VAs, the social roles they came to play within households, and the fragility or durability of everyday routines built around them.

Related work. In this section, we situate our study within three strands of HCI and CSCW research. First, we review work on how VAs are taken up in domestic life, focusing on everyday routines, multi-user interaction, and the social and relational dynamics that shape use in shared households. Second, we revisit domestication theory and its role in HCI as a lens for examining how technologies become integrated in everyday practice. Third, we consider research on short-term and long-term studies of living with VAs, highlighting what is known about early engagement and where temporal gaps remain.

Across these strands, we identify a key limitation in existing work: while prior work captures early adoption, we know far less about how VAs evolve as domestic life changes. Our longitudinal study shows that life transitions—such as moving, relationship shifts, or changes in caregiving—are central to VA domestication, yet rarely reflected in shorter-term research.

2.1 Voice Assistants in the Home The notion of a smart home equipped with various technologies has been of research interest for decades, VAs like Alexa in domestic life, however, have further changed the notion of what it means to live with at-home technology. HCI research has explored VA use in domestic environments from multiple angles, including concerns about privacy and surveillance (e.g., [2, 19, 21, 31, 34, 56, 57, 81, 82, 104]), everyday usage patterns (e.g., [5, 7, 62, 71, 110, 119]), and their role as aids to particular user groups (e.g.,[17, 20, 38, 98, 99]).

Much of this work has examined how VAs raise questions of privacy, surveillance, and trust in the home. Studies consistently show that the “always listening” quality of VAs generates user concern (e.g., [19, 56, 94, 95, 104]). Edu et al. identified some of the main security risks associated with VA use, including the complexity of the underlying technology and the nature of the AI that such devices rely on.

Other studies have examined general usage patterns of VAs at home. Despite the high expectations that surrounded their introduction, households largely use VAs for a narrow range of tasks. Ammari et al. showed that music, search, and IoT control dominated everyday use, and other studies showed similar lightweight uses—such as setting timers and checking the weather—once initial enthusiasm faded. Luger and Sellen described VAs as “like having a really bad PA,” capturing the frustration many users feel when devices fail to live up to expectations of competence. Ma et al. further argued that adoption orientations vary across enthusiasts, pragmatists, and skeptics, suggesting that voice interfaces cannot be designed for a singular “average user”.

Method. For long-term and contextually driven results, we performed a longitudinal, three-year research project (January 2020-December 2022). Ethical approval for the study was granted by the University of Siegen. We began by recruiting households and, in line with domestication research that conceptualizes technology integration as an ongoing and evolving process rather than a one-time event, we intentionally included both new and experienced VA users. This allowed us to capture a broad range of trajectories—from early encounters and initial adjustments to long-term reconfigurations and stabilized routines. Regardless of prior experience, all households received new devices from us (e.g., Amazon Echo, Echo Dot, Echo Show; Google Home, Google Nest Mini, Google Nest Hub) based on their preferences, ensuring a comparable starting point within the study context.

Participants were recruited until June 2021 via social media, personal networks, university mailing lists, and snowball methods. This resulted in 36 German households. Six households opted out of the study over three years due to unforeseen changes such as severe illness or alterations in their living situation. Our final group consisted of 30 households and 33 participants (M = 21, F = 12, average age = 36): 23 with Alexa, 14 with Google Assistant, and 3 using both, as detailed in Table 1. Among the participants, fifteen were experienced VA users, having used a device for at least six months prior to joining our project. Participants’ living situations varied from living alone (N = 10), cohabitating with a partner (N = 10), and sharing a space with roommates (N = 7), to residing with family and children (N = 6). During the project, nine participants’ living situations underwent changes (as indicated by (*) in Table 1). P3, P4, P12, P11, and P19 moved to live alone from shared spaces or parental homes. P14 moved from a shared flat with her partner. P18 began a weekday long-distance relationship. After P31 and P32 separated, P31 shifted to a shared flat, while P32 remained in their original house.

Households were informed that they would receive compensation for participating but were not told the amount during recruitment to avoid bias; each household later received a €100 gift card. They were also permitted to retain the devices after the project concluded. All data collection was conducted in German and later translated to English.

Discussion. Although prior HCI research has examined how VAs enter the home—focusing on early appropriation, initial use, and the first stages of domestication—it has offered only a partial view of how these systems evolve within the shifting rhythms of domestic life. To address this gap, our contribution and methodological strength lies in capturing the home as a nexus of spatial, relational, and temporal change. From this empirical grounding, we conceptualize domestication not as a stage-based progression but as a calculus that unfolds across spatial, relational, and routine dimensions. In what follows, we first synthesize how these dimensions interact and reshape one another over time (Section 5.1), then introduce the domestication calculus as an analytic framework for understanding long-term technological change in the home (Section 5.2), and finally discuss implications for designing VAs that can better support households across life transitions and long-term use (Section 5.3).

5.1 Interdependence of Domestication Dimensions Over Time: Integrating Spatial, Social, and Routine Dynamics (RQ2) While our findings were presented across three analytically separate dimensions—spatial, social, and routine integration—domestic life is not experienced in discrete parts. To understand how VAs become consequential over time, these dimensions must be examined in relation to one another. In this section, we bring together our longitudinal findings with shorter-term VA studies, illustrating how extended timescales reveal forms of domestication that shorter deployments have not captured.

5.1.1 Spatial arrangement. Consistent with prior research, many participants initially preferred to place the VAs in communal areas such as kitchens and living rooms, reflecting expectations of shared use. Shorter-term studies often describe these placement decisions as relatively stable across the adoption period. Across three years, however, we observed that spatial arrangements were far more fluid. When households moved, separated, or when children grew older, VA placement shifted—sometimes substantially. For instance, as discussed in Section 4.3.3, a participant who moved out of a shared flat with her partner stopped placing the device in communal areas due to privacy concerns, despite having previously shared it freely with flatmates. This kind of relationally contingent spatial reconfiguration is largely absent from short-term studies, which rarely capture major life transitions. Beyond movement and reconfiguration, we also observed age-related spatial trajectories. Older participants tended to expand the presence of VAs across multiple rooms, reflecting increasing utility—a pattern consistent with the long-term findings in and (see Section 4.1.1). In contrast, some participants—particularly millennials —gradually marginalized the device, placing it near the TV, on shelves, or in visually peripheral locations, treating the VA more as a decorative or dormant object than an actively used technology (see Section 4.1.3). This contrast also resonates with prior research showing older adults’ positive engagement with social technologies, such as companion robots, and may reflect both increased time spent at home and different everyday responsibilities compared to other cohorts.

5.1.2 Interactions and roles. Prior work has shown that VAs can take on a range of social roles—mediating family cohesion, offering playful companionship, or serving as conversational partners for children.

Conclusion. The use of VAs is highly complex, mitigated by a range of factors across space and time. This complexity is unlikely to reduce as the technology evolves. Voice assistants have clearly enormous potential to become intricately woven into the fabric of family life. With the advent of LLMs and sophisticated natural language processing, their capacity to become social actors can only be enhanced. However, a significant gap remains in research that thoroughly examines the roles of VAs in real-life home settings and their progressive domestication into daily life.

By tracing spatial arrangements, relational roles, and routine entanglements across thirty households over three years, our study shows that VAs can foster intimacy and playfulness, mediate communication, support coordination, and at times unsettle domestic routines. These dynamics reveal that domestication is not a linear progression but an ongoing negotiation shaped by shifting living arrangements, relationships, and household rhythms.

We propose the domestication calculus as a framework for analyzing and designing for these long-term processes. The calculus captures how spatial, relational, and routine dimensions evolve together, offering future researchers a structured way to examine how domestic technologies transform alongside household transitions. We see this work as a foundation for further studies of agentic systems in more diverse cultural and socio-economic contexts and for design approaches that support technologies capable of adapting, repurposing, and renegotiating their roles over time. With their domestication in the lives of the young, in particular, VAs bring a new potential to the way we conceive ‘family’ life.

Limitations. Here, we address certain limitations of our research. Firstly, while our sample covered various generations, it was not evenly distributed across age groups. As such, we refrain from making broad claims about specific age groups. Instead, we offer trends discerned from this sample. A more diverse sample across age, culture, and social class might yield a deeper understanding of home practices.

Secondly, despite our efforts to include all household members, our study mainly involved primary device users, with only three cases featuring non-primary users. This skew might underscore a dynamic where primary users typically dominate interactions, and suggest that the viewpoints of non-primary users, who were scarcely represented, might be different.

Thirdly, several participants already owned VAs at the beginning of the study, which may have shaped their initial expectations and practices. However, the breadth of our sample and the diversity of household configurations allowed us to observe a wide range of trajectories, and we reached conceptual saturation in identifying domestication patterns across homes. Relatedly, not all participants took part in every stage of the study (first interview, diary, and final interview).