Research has repeatedly demonstrated the critical importance of social connectedness in older age. Social isolation and loneliness, the main constructs of social disconnectedness, are prevalent in Americans aged 65 years and older, with prevalence rates approaching nearly 25% for the former [1] and ranging from 17% to 57% for the latter [2-4]. Social isolation is objectively defined as the lack of social contact, communications, and relationships [1,5], while loneliness refers to a subjective feeling resulting from a perceived deficit in the quantity or quality of meaningful social relationships [6]. Older adults who are socially isolated or feel lonely exhibit a wide range of negative health outcomes, including depression, cardiovascular diseases, poor cognitive function, reduced quality of life, and increased mortality [2,5,7,8].

Social isolation and loneliness are pressing concerns for older adults living in affordable housing (hereafter referred to as “residents”) because they have limited social networks, transportation options, and access to information and communication technology (ICT) [9]. Chronic health conditions and disabilities are also common in this population [10,11], which affect their ability to establish and maintain meaningful social interactions. Also, affordable housing, particularly in high-rise structures, may lack communal areas [12] or discourage residents from inviting others over due to suboptimal building conditions [13]. The complex interplay of health, housing conditions, and social and financial factors may contribute to elevated levels of social isolation and loneliness among residents [14]. Innovative approaches are necessary to increase meaningful social connections and reduce loneliness in this population. It is recognized that there is no “one-size-fits-all” approach to address this issue and there is a need to design interventions customized to meet specific groups or the varying levels of social isolation and loneliness they may experience [15].

A notable gap exists in technology-based interventions tailored to enhance social connectedness among residents. While face-to-face programs and services for improving social connections are available for older adults (eg, community senior centers and senior cafés), these may be inaccessible to older adults residing alone in affordable housing for multiple reasons mentioned previously. The growing interest in voice-operated smart speakers (eg, Amazon Echo and Google Home) has increased an exploration into how this technology can enhance social connectedness in community-dwelling older adults [16-18]. The hands-free voice control may improve accessibility and usability for people with vision impairment, mobility limitation, or poor dexterity. Artificial intelligence (AI)–based voice agents (eg, Amazon Alexa) on smart speakers have the potential to help older adults in limited-resource settings feel less lonely and engage more in social activities, for example, by providing easy access to communicate with others. Studies provide preliminary evidence that older adults enjoy social health benefits from interactions with an AI voice agent, such as a source of companionship reducing loneliness and improving mental health [18,19].

Despite the potential benefits of smart speakers, unique challenges to their adoption and use may arise, particularly when the needs of target end users are overlooked or inadequately addressed. Moreover, historically marginalized communities, including low-income older adults and racial or ethnic minority groups, still face inequity in ICT access as well as challenges in its adoption and use [20-22]. Low ICT adoption and utilization are attributed to various barriers, including technology costs, internet accessibility, functional impairments, low literacy, lack of self-efficacy, privacy concerns, and mistrust of the technology [23-28]. Recognizing the unique needs of users and the factors influencing acceptance and usability within the user’s living context is critical for implementing technology-based interventions. This can be accomplished through user-centered design (UCD), which involves exploring user needs, preferences, and technology requirements; engaging users throughout the design process; and iteratively developing technology prototypes [29].

The purpose of the study was to engage older adult residents living alone in affordable housing in an iterative UCD process to develop prototype ideas of innovative solutions leveraging smart speaker technology for social connectedness. Through a 3-stage co-design process, we sought to (1) understand the social connection needs and technology preferences of residents, (2) collaboratively develop and refine prototype scenarios that address these needs, and (3) identify design requirements and implementation considerations for future deployment. Our work focused on early prototyping to develop technology solutions and build a shared understanding of the social connection needs in this population. Our long-term goal is to develop smart speaker–based applications called Voice2Connect for deployment in affordable housing to enhance social connectedness among their residents.

We conducted a UCD study to engage residents in developing smart speaker–based applications for social connectedness. Our process consisted of three major stages conducted between April 2021 and April 2022: (1) needs assessment and idea generation, (2) iterative prototype development, and (3) validation and implementation planning (Table 1).

We recruited participants from 3 housing buildings in Richmond, Virginia, United States (referred to as housing A, B, and C) that serve older adults with limited income and individuals with disabilities. Our research team had established relationships with these communities through prior collaborative research and wellness programs. Eligible participants were older than 55 years, lived alone, could speak and understand English, and were able to participate in all study activities. We excluded individuals with severe sensory impairments preventing engagement with audiovisual materials, terminal illness, or a score ≤8 on a short version of the Montreal Cognitive Assessment indicating moderate to severe cognitive impairment. Participants were divided into 2 groups based on smart speaker experience: users (n=13) who had been using Amazon Echo Dot devices deployed by housing A since January 2020, and nonusers (n=16) from all 3 housing sites.

We recruited participants through multiple channels including housing management referrals, community wellness program participant lists, on-site information sessions, and flyers. Due to pandemic-related visitor restrictions, eligibility screening was conducted by phone, during which we also collected demographic and technology experience information. Detailed descriptions of recruitment procedures can be found elsewhere [30].

The Institutional Review Board of Virginia Commonwealth University determined that this project posed minimal risk and waived written informed consent (HM20020008).

We conducted focus groups with older adults in phases 1, 3, and 5 to perform an iterative co-design process and gather data. The groups were stratified by housing building and current smart speaker use (Table 2).

We conducted 4 focus groups (n=17) in April 2022 to validate refined scenarios. Team members again used role-play to present scenarios, explicitly noting incorporation of previous feedback. We solicited perceptions, desired additional features, and implementation recommendations. Sessions were transcribed verbatim and analyzed as described in the Data Analysis section. Throughout the study, participants received US $20 gift cards at each session. Focus groups were limited to 6 participants to encourage initiated conversations while minimizing groupthink [34].

Before phase 1, we conducted a phone-based survey to collect data on demographics, technology experience, social well-being (social isolation, loneliness, and satisfaction with current social interaction levels; Multimedia Appendix 2), and other aspects of health such as functional status, depression, sleep, and physical activity. This manuscript reports only on demographics, technology experience, and social well-being.

We used a combination of inductive and deductive qualitative approaches across the 3 stages of the UCD process, selected to match the analytic goals of each phase (Table 3).

Twenty-nine older adults participated in 1 or more focus group sessions (Table 4). The sample was predominantly African American (n=23, 79%), with 55% (n=16) male. Out of 29 participants, 20 (69%) participants had completed high school education or less. While 21 (72%) participants owned smartphones, only 15 (52%) participants had computer or laptop access, and 12 (41%) participants had never used the internet. Thirteen (45%) participants were smart speaker users at study inception, 11 (85%) of whom reported prior internet use and 2 (15.4%) did not. Among the 11 participants with prior internet use, the majority (n=8, 73%) were daily internet users.

At baseline, participants had a mean Lubben Social Network Scale-6 score of 15 (SD 6.7), with 8 (28%) out of 29 participants meeting the threshold for social isolation (Lubben Social Network Scale-6 score of <12). Mean University of California, Los Angeles (UCLA) Loneliness Scale-Short Form score was 4.6 (SD 1.9), with 9 (31%) out of 29 participants meeting the threshold for loneliness (UCLA Loneliness Scale-3 ≥6). While just over half of the participants (16/29, 55%) reported being satisfied with their current social relationships, 34% (10/29) reported being very dissatisfied and 10% (3/29) somewhat dissatisfied, and fewer than half (12/29, 41%) reported participating in any organizations, religious groups, or committees.

Phase 3 evaluation revealed both enthusiasm and concerns. Participants appreciated applications that could foster connection and help them stay informed about housing events. They strongly preferred technology facilitating human connection over purely automated interactions. The social companion features were seen as useful for those experiencing loneliness, although some worried about replacing rather than supplementing human contact.

Privacy and obtrusiveness emerged as primary concerns. Participants worried about other residents or management checking-in at any time, unintentional sharing of location and activity information, and technology exerting excessive control (described as “too Orwellian” by one group). Responses to the wellness check scenario indicated that participants were receptive to the use of wearable-based data for monitoring physical inactivity and detecting potential falls. However, they expressed a clear preference for systems that prompt rather than diagnose, flagging activity changes and facilitating human connection, rather than inferring emotional states from sensor data.

Participants provided concrete improvement suggestions for each scenario (Multimedia Appendix 7). In summary, these include (1) user control—add “do-not-disturb” features and allow selection of check-in frequencies, (2) privacy protection—create close contact lists rather than building-wide access and ensure that health information is not shared during check-ins, (3) natural interaction—make voice interfaces sound conversational rather than mechanical, (4) minimal management oversight—limit housing management role to essential check-ins only, and (5) personalization options—allow users to set speaking times, choose conversation perspectives, and toggle features on/off.

The depth and quality of participant contributions did not differ by prior smart speaker experience. Analysis of focus group interviews across all 7 design scenarios revealed comparable levels of engagement between smart speaker users and nonusers. Per focus group, both groups generated similar numbers of positive themes (users: 11; nonusers: 11) and design recommendations (users: 14; nonusers: 11.3). The major design themes, including privacy and user control, perceived usefulness for socially isolated residents, trusted contact network preferences, and voice-based security, emerged consistently across both groups. While users occasionally drew on their lived device experience, nonusers contributed equally substantive and actionable recommendations, including novel suggestions such as video interfaces for hearing-impaired residents, information integration, and peer activity coordination via Alexa.

Based on this feedback, we revised all 7 scenarios (Textbox 1). Key changes included adding opt-in or opt-out capabilities, do-not-disturb functions, contact list controls, and explicit privacy protections. For example, in the management check-in scenario, we specified that Alexa would confirm resident status without sharing additional details the resident mentioned (eg, having a cold). Figure 1 shows a graphic representation of a scenario in the checking-in category.

This study used an iterative UCD process to engage older adult residents in affordable housing in designing prototype scenarios for smart speaker–based applications to promote social connectedness. Throughout 3 stages of co-design, we developed 7 scenarios across four functional categories: (1) checking-in for safety verification, (2) conversational AI as social companions, (3) community involvement, and (4) wellness monitoring. These scenarios provide a foundation for developing a technology platform tailored to address the specific social isolation and loneliness experienced by underserved older adult populations, who often face distinct challenges in locating and accessing the social and financial resources necessary for aging in place. Baseline social well-being data underscore this need. While more than half reported satisfaction with their current social relationships, a meaningful subgroup was experiencing clinically significant loneliness and social isolation, lending ecological validity to their strong interest in features supporting social connectedness in their living communities.

The findings of this study can be interpreted through the lens of Self-Determination Theory [41], which posits that human motivation and well-being are grounded in 3 psychological needs: autonomy, competence, and relatedness. Self-Determination Theory maps meaningfully onto the design preferences and concerns expressed by participants across all phases. The demand for customization, opt-in controls, do-not-disturb features, and close contact lists reflects a fundamental need for autonomy, the desire to maintain personal agency over one’s social environment and information sharing. Participants’ requests for interfaces that are simple, conversational, and responsive to their voice reflect the need for competence, the ability to engage effectively with technology without feeling overwhelmed or excluded. Finally, participants’ desire for building a livable community for themselves and peers and for preserving human relationships in the loop reflects the need for relatedness, belongingness, and meaningful connection with others. Designing smart speaker–based applications that satisfy these needs may be essential not only for usability but for sustaining intrinsic motivation for continued technology engagement and well-being outcomes in this population.

Throughout the design process, it became clear that the study’s focus on social connectedness strongly resonated with participants due in part to the isolation they experienced during the COVID-19 pandemic, a period during which residents faced significant restrictions on in-person social interaction. The strong participant interest in peer check-ins and community engagement features may partly reflect this context, and it is plausible that the relative prioritization of these features could differ in a postpandemic sample. Importantly, however, a critical insight emerged: for affordable housing residents living alone, social connection needs cannot be separated from safety and survival concerns. Participants emphasized that any technology system must address both relationship-building and emergency support, reflecting the reality that all residents live alone and some face immediate safety vulnerabilities, including declining health and fall risk, that may arise at times when human assistance is not immediately available.

This dual emphasis shaped our design approach fundamentally. Rather than creating purely social applications, we prioritized checking-in functionality that serves both social and safety purposes. This finding aligns with research suggesting that older adults’ technology adoption is strongly influenced by whether innovations address their most pressing daily concerns [42]. This study shows that for affordable housing residents with limited support networks, safety vulnerabilities are not abstract future concerns but immediate realities shaping their daily experiences of living alone [43].

Our findings indicate that residents were genuinely interested in making a livable community within their housing by building social infrastructure [44], not just to address social needs but also to promote safe, independent living. Participants suggested various ideas, including virtually connecting with individuals in need, conducting remote safety check-ins, advertising community events, and facilitating coordinated trips to essential destinations (eg, grocery stores). Participants were interested in expanding and enhancing social networks within and beyond their housing community. Research suggests that digital tools connecting older adults to existing social networks may not significantly increase social activity for individuals lacking networks and at risk of social isolation [45]. However, tools designed to facilitate new social connections or broaden access to community resources can play a critical role in addressing these gaps [46]. Our scenarios address this through community-building features (eg, virtual bulletin boards, coordinated meetups, and shared activity planning) that help residents discover and connect with neighbors who share similar needs and interests.

The emphasis on community involvement also reflects participants’ recognition that their housing buildings could become more than just places to live in isolation. They wanted technology that would help them build “something lively,” a community with regular interaction, mutual support, and shared activities. This vision extends beyond typical conceptualizations of smart speakers as individual assistive devices and positions them as one component of a broader, community support ecosystem that can facilitate communication about opportunities and access to them, while the core work of building and sustaining community remains a fundamentally human endeavor. Importantly, technology alone is insufficient to establish such infrastructure. The foundation must be built through human-centered programming, dedicated staffing, and sustained relational investment. Smart speaker–based applications, therefore, should be understood not as substitutes for this work but as complementary tools that facilitate awareness of and access to existing and emerging community resources, functioning as bridges to human connection.

To our knowledge, this is one of few studies using UCD with older adults of lower socioeconomic status. Such populations are often excluded from design and testing processes for new technologies conceptualized to improve quality of life and health [47], probably due to concerns about their ability to engage meaningfully and provide valuable input. Limited access to ICT and low digital literacy also may have contributed to the underrepresentation of these communities in technology research and design initiatives. The limited participation of marginalized communities perpetuates the exclusion of their perspectives [48] and contributes to disparities in digital health access and outcomes. Given that widespread digital technology access has created new opportunities for self-management, quality of life, and social well-being in older adults [49-51], recognizing the importance of engaging community members of underrepresented populations in a technology development and testing process is essential.

Our study demonstrates that older adults in underserved communities, including those with limited prior technology experience, can actively participate in and meaningfully contribute to technology design processes. Our participants across age groups, technology experience levels, and housing communities demonstrated enthusiasm and active engagement and expressed a sense of ownership in the design sessions, contributing valuable insights that shaped prototype development. The 153 ideas generated during brainstorming activities, the nuanced feedback on privacy and control features, and the concrete implementation recommendations all demonstrate that concerns about this population’s ability to engage in design work are unfounded.

Our research has methodological implications for UCD research with and for underserved communities. The 3-stage iterative approach we used successfully balanced rigorous design principles with the unique contexts and abilities of participants. Our inclusive methods included rapid brainstorming sessions, facilitated mini-group discussions to ensure that everyone had a voice and that discussions remained focused and productive, audiovisual materials to enhance nonusers’ understanding of the target technology and stimulate idea generation, scenario creation and role-play demonstrations to reduce uncertainty about the technology and facilitate discussions around privacy and security, and iterative design revision informed by continuous feedback.

These methods were thoughtfully chosen based on the literature [31-33] and the team’s extensive experience working with affordable housing communities. Beyond specific UCD techniques, the team prioritized creating a comfortable, nonjudgmental environment where participants, especially those not familiar with smart speakers, could freely propose ideas, suggest changes to scenarios, and identify concerning features. We found that using pictorial sheets allowed participants with varying literacy levels to express ideas, while mini-group formats prevented more vocal participants from dominating discussions.

This study also demonstrates the need for flexibility in balancing rigorous design principles with the unique contexts and abilities of participants, as previous research demonstrated [32]. For example, during phase 1, we opted not to conduct mini-group sessions with focus group 5 (nonusers) after observing their difficulty engaging in technology-related discussions. Instead, we adapted our approach to facilitate a large group discussion focused on preferred functionalities for social connectedness, emphasizing their role as co-designers. This adaptive method adds important considerations specific to a UCD process with underserved populations, suggesting that methodological fidelity should sometimes yield to participant comfort and authentic engagement.

The established trust relationships our research team had developed through prior community-engaged work [52] proved essential to recruitment and retention. This suggests that successful UCD with marginalized communities requires sustained institutional commitment beyond individual research projects, and researchers cannot parachute in for design sessions without prior relationship-building and expect meaningful engagement.

Participants’ privacy concerns emerged consistently across all 3 stages and shaped significant design revisions. Their worries about unintentional information sharing, management surveillance, and loss of autonomy led to specific requirements: opt-in or opt-out controls, do-not-disturb features, close contact lists rather than building-wide access, and explicit privacy protections ensuring that sensitive information would not be shared with management or other residents. These concerns reflect broader issues of power, trust, and control that are particularly salient for minority, low-income older adults who may have experienced systemic marginalization and surveillance [53]. The description of mood detection features as “too Orwellian” signals deep discomfort with technologies that could exert control over their lives without consent. Our design iterations directly addressed these concerns by centering user control and transparency.

Importantly, participants distinguished between features they found helpful (checking-in on neighbors and being reminded to connect with others) and those they found intrusive (automatic mood detection and unrestricted management access). This nuanced feedback demonstrates why participatory design is essential because researchers and technologists might assume that more automated, comprehensive monitoring would be beneficial, while users clearly articulated the need for boundaries and personal agency.

Participants expressed interest in voice agent companionship while simultaneously voicing concerns about replacing human interaction. This tension reflects broader societal questions about the role of AI in addressing loneliness and social isolation. Participants recognized that for those experiencing severe isolation, having “somebody to talk to” through Alexa might be “better than nothing,” while also insisting on the primacy of human relationships.

Our scenarios attempted to navigate this tension by positioning voice agents as facilitators of human connection rather than replacements for it. For example, the social companion scenarios included Alexa suggesting the user connect with friends for walks rather than only offering its own companionship. The wellness check scenario prompted connection to friends or health care providers rather than only providing automated support. This design approach aligns with emerging ethical frameworks for AI companionship that emphasize augmentation rather than substitution [18]. However, participants’ concerns suggest that implementation will require careful monitoring to prevent vulnerable, isolated individuals from becoming overly reliant on AI interaction at the expense of human relationships. Future development should include safeguards that encourage and facilitate human connection rather than allowing AI to become the primary or sole social contact.

The varying levels of enthusiasm across scenarios suggest that a modular, customizable system would be most appropriate. Rather than implementing all 7 scenarios uniformly, housing communities should be able to select and configure features based on resident preferences, building infrastructure, and management capacity. This aligns with participants’ emphasis on personalization and control. Moreover, given participants’ limited experience with ICT and struggles with voice user interface interactions shared in phase 1 focus groups [11], resident training sessions and ongoing technical support are required for successful implementation. These supports are inherently service-oriented, requiring dedicated staff, sustained resourcing, and the cultivation of trust with residents, and should be considered alongside, rather than assumed to be fully addressed by the technology itself.

In the future, research should examine how smart speaker–based social connection tools interact with other community resources and interventions. Our participants identified transportation barriers, lack of information about community activities, and limited social networks as interconnected challenges. Technology solutions are most likely to succeed when integrated into broader community-based approaches that address structural barriers to social connection.

Learning emerged as a prominent brainstorming category in phase 1, comparable in volume to Socializing With Others. Participants—all nonusers—generated ideas spanning news and current events, language acquisition, practical skills, and digital literacy. Although this domain falls outside the social connectedness focus of this study, these observations point to an interest in intellectual engagement within this population, consistent with a growing literature on lifelong learning in older adults [54,55]. Future work might more explicitly examine learning-oriented smart speaker–based applications as a complementary pathway for supporting cognitive engagement and quality of life.

There are several limitations of this study. First, due to the COVID-19 pandemic, we were unfortunately unable to follow up with some participants throughout the process, resulting in their loss from the study. The pandemic also necessitated off-site meetings with transportation provision, which may have excluded some potential participants with mobility limitations or discomfort with van travel. Second, established trust relationships that facilitated recruitment may have introduced response bias, as participants with prior positive experiences with the research team may have been more inclined to provide favorable feedback. Given that real-world implementation will inevitably encounter residents who are skeptical, mistrustful of institutional technology, or uninterested in smart speaker–based applications, purposive recruitment of such individuals would elicit insights that a self-selected, engaged sample cannot provide. Third, incorporating diverse and sometimes contrasting ideas into the design was challenging. While we used multiple data synthesis methods, a more systematic approach to identifying and prioritizing user demands (eg, conjoint analysis or formal prioritization exercises) would have been beneficial for making design trade-off decisions explicit. Fourth, hearing-related concerns, such as hearing loss that hinders meaningful engagement with smart speakers, did not emerge spontaneously in focus group discussions. This absence may reflect sample characteristics or participants’ limited awareness of the relationship between hearing ability and voice interface usability. Although smart speaker designs incorporate features such as adjustable volume and visual display companions that may support more equitable access, their adequacy in practice remains unclear. Future work should more explicitly examine how hearing loss shapes engagement with smart speakers. Fifth, while we adapted facilitation approaches to encourage broad participation, differential comfort with technology discussions likely shaped whose voices were most prominent in certain design decisions. Furthermore, differences in age, education, technology access and usage, and prior smart speaker experience among participants may have influenced the outcomes at each phase of the study. Finally, this study focused on early-stage prototyping and scenario development. The scenarios represent participant aspirations and preferences but have not been validated through implementation.

Our study identified the needs for meaningful connections and safe, independent living among older adults living alone in affordable housing and showed the potential of smart speakers and their voice agents to address social isolation and reduce loneliness. Through a 3-stage iterative UCD process, we successfully engaged residents, including those with limited technology experience, in co-designing 7 prototype scenarios across 4 functional domains. The UCD methods used in this study effectively incorporated end user feedback to develop a technology platform for reducing social isolation and loneliness in this population [42]. Findings emphasize the importance of accessible, thoughtfully designed technology tools for social connectedness and more broadly for health and well-being tailored to older adults with limited resources and digital literacy. Further research is needed to fully develop these prototypes into functional applications, pilot-test them in real-world settings, and evaluate their effectiveness in addressing the unique social health needs of underserved older adults. Future evaluation should also extend beyond self-selected participants to deliberately engage skeptical residents, individuals with institutional mistrust, and those for whom this technology is not an intuitive fit. Rather than assuming universal applicability, research should iteratively test and refine these prototypes across diverse residential contexts, trust environments, and user dispositions to generate a more grounded and transparent evidence base to implementation in affordable housing communities.