From support to empowerment: AI-enabled transformative service for positive aging

The world is undergoing a dramatic demographic shift, marked by a rapidly growing aging population (Farag et al., 2024): the population of aged 65 and above is estimated to double over the next decades reaching 1.6 billion by 2050 and accounting for more than 16% of the global population (United Nations, 2023). Among older adults aged 55 and older, a recent multinational survey conducted by McKinsey Health Institute across 21 countries finds that the widely shared priorities in late life are having purpose, enjoying meaningful connections, managing stress and maintaining independence (Ahlawat et al., 2023). These late life priorities align with the concept of positive aging, which emphasizes the continued pursuit of physical, psychological and social well-being as individuals grow older (Bar-Tur, 2021).

As the aging trend accelerates globally, researchers and practitioners have increasingly highlighted the important role of lifelong learning in supporting older adults' well-being and promoting positive aging. Lifelong learning that reflects older adults' interest in intellectual engagement is an important avenue leading to positive aging (Boulton-Lewis, 2010; Sloane-Seale and Kops, 2008). Drawing on lifespan development theory (Baltes et al., 1999), researchers suggest that cognitive stimulation through education, creative activities and technology-assisted learning can foster successful aging by enhancing mental agility and emotional well-being (Kim and Park, 2017). For example, participation in creative hobbies and continuous education (e.g. languages, art, music) can promote a sense of purpose and connection, both of which are key components of psychological well-being (Ryff, 2013). Initiatives like the UK-based University of the Third Age (www.u3a.org.uk) reflect this concept by offering continuous education opportunities for learning and social interaction among older adults. Research further confirms that older adults who engage in such self-directed learning often experience improved quality of life (Morrison and McCutheon, 2019).

Against this backdrop of demographic aging and increasing emphasis on lifelong learning, the current research argues that the emergence of interactive generative AI (GenAI) shows great potential in providing transforming service to support older adults in achieving positive aging. GenAI can facilitate personalized, adaptive and emotionally resonant learning experiences, as well as companionship through interactive conversation (Grewal et al., 2024; Mende et al., 2024). Early applications such as Elliq, an AI platform tailored to older adults, are already adopted by the service industry to address the aging demographic's unique emotional and cognitive needs (Mende et al., 2024). Well-designed GenAI tools can help in extending older adults' engagement in intellectual activities and social interaction into later stages of life. However, the current academic research still lags in providing robust empirical evidence to support the potential transformative power of GenAI, particularly its impacts on the aging older adults. Despite growing conceptual interest in GenAI's service potential (e.g. Grewal et al., 2024; Huang and Rust, 2024; Wirtz and Stock-Homburg, 2025), empirical studies fail to keep pace with the rapid technological innovation, which leaves critical gaps in understanding how to optimize GenAI service for the well-being of the aging population.

To address this research gap, the current study draws on the concepts of positive aging (Bar-Tur, 2021) and AI-enabled transformative service (Huang and Rust, 2024; Mehmood et al., 2024) to explore how GenAI can be optimally designed to support older adult’ psychological well-being. Specifically, three progressive studies are conducted to investigate whether (Study 1), what (Study 2 and 3) and how (Study 1, 2 and 3) AI-enabled service design would be welcomed by older adults because of enhanced purpose of life and satisfaction with life. This research focuses on two core dimensions of psychological well-being: personal growth, which fosters a sense of purpose in life through AI-enabled self-learning service, and social connectedness, which contributes to life satisfaction via interactive GenAI engagement. By investigating these constructs, this research aims to advance theoretical insights in transformative service and AI-enabled service design, while offering practical guidelines for developing GenAI tools that address the cognitive, emotional and social needs of older adults for positive aging. The research also responds to the call by Mende et al. (2024) for a reframing of AI-enabled service design to empower aging populations, both effective for service customers and sustainable for service providers.

The integration of Artificial Intelligence (AI) into service has been significantly reshaping service design and delivery and transformed customer care journey by enhancing service speed, precision, personalization and accessibility (Huang and Rust, 2018, 2024). The progression from mechanical, analytical, intuitive, to empathetic intelligence highlights how AI's growing sophistication enables service innovation and changes the strategic design of service systems, which is captured by the service strategy framework proposed by Huang and Rust (2021) based on the three major AI intelligence levels: “mechanical AI” for standardization, “thinking AI” for personalization and “feeling AI” for relationalization. These categories align AI capabilities with specific service outcomes, indicating the transformative potential of AI-enabled service. Particularly, the development of “feeling AI” marks a critical advancement in transformative services and the emergence of GenAI emulates empathetic interactions traditionally reserved for human (Huang and Rust, 2024).

The concept of transformative service – defined as service that enhances the well-being of individuals and communities (Anderson et al., 2013) – has gained increasing attention. As AI technologies rapidly evolve and integrate into service systems, researchers have begun exploring how AI can be a catalyst for transformative service outcomes across various domains, such as healthcare, education and aging (Grewal et al., 2024; Mende et al., 2024). AI-enabled transformative service refers to the integration of artificial intelligence technologies (e.g. machine learning, natural language processing, recommendation algorithms, robotics) into service design and delivery in ways that enhance customer well-being, empower users and reduce systemic inequities (Huang and Rust, 2024; Mehmood et al., 2024). This stream of research builds upon the transformative service research (TSR), which prioritizes well-being as a core service outcome (Anderson and Ostrom, 2015), and further extends by incorporating digital and autonomous systems into the service experience.

AI may foster transformative service through the mechanisms of personalization, accessibility and co-creation (Kumar et al., 2019; Tully et al., 2025; Wang and Touré-Tillery, 2024). For example, personalized AI recommendations can match services to individual preferences or vulnerabilities, thereby enhancing relevance and satisfaction (Hermann et al., 2024). AI can reduce access frictions for marginalized populations, such as language barriers through real-time translation or cognitive challenges via intuitive interfaces (Tully et al., 2025). Co-creation of value is also enhanced through AI's ability to learn from user behaviors, enabling adaptive and responsive service ecosystems (Huang and Rust, 2018). Several domains have demonstrated how AI-enabled service may contribute to transformative outcomes. In healthcare, AI-enabled services such as diagnostic tools and personalized health monitoring promote patient empowerment, preventive care and more equitable access (Ho, 2020; Longoni et al., 2019). In education, AI tutors and adaptive learning systems provide individualized support, helping underserved learners overcome cognitive and social barriers (Kabudi et al., 2021). In aging service, AI companions and ambient technologies support social connectedness and independent living, contributing to older adults' psychological well-being and dignity (Bedaf et al., 2015).

Although conceptual frameworks and early applications suggest that AI-enabled transformative service has great potential to significantly enhance individual and societal well-being, empirical research is still limited. Particularly, research focusing on such marginalized customer groups as aging population is much underdeveloped. To bridge this research gap, the current study aims to explore whether, what and how AI-enabled service may help old adults for transformative outcomes manifested as positive aging.

Positive aging refers to the concept shifting from a deficit-focused view of aging (i.e. decline and dependency) toward a holistic and empowering aging perception that emphasizes growth, social participation and well-being in late life (Bar-Tur, 2021; Hill, 2005). Rooted in theories of successful aging (Rowe and Kahn, 1997), positive psychology (Seligman and Csikszentmihalyi, 2000) and gerotranscendence (Tornstam, 2005), positive aging is defined as the process of maintaining a high quality of life, psychological well-being and active engagement in meaningful activities as individuals grow older (Bar-Tur, 2021). Literature identifies several core dimensions of positive aging, including physical health and functioning, psychological well-being, social connectedness, purposeful engagement and adaptability (Medina, 2017). While maintaining physical health is foundational, researchers emphasize that subjective psychological well-being is equally important (Martinson and Berridge, 2015). Furthermore, studies have shown that older adults who report a sense of purpose (Ryff, 1989) tend to exhibit better health outcomes and greater life satisfaction. Research reveals that positive aging can be enhanced through effective strategies that address older adults' psychological well-being needs (Bar-Tur, 2021), such as providing positive psychology interventions and growth opportunities, and creating senses of purpose and connectedness.

Research on positive aging underscores the critical role of psychological well-being. Ryff's (1989) multidimensional model – autonomy, personal growth, self-acceptance, life purpose, environmental mastery and positive relationships – has been widely used in gerontological studies. Older adults often demonstrate what is known as the paradox of aging, indicating subjective well-being remains stable or even improves with aging despite physical decline (Carstensen et al., 2011). Hence, positive aging is not merely about living longer but about living better, with purpose and connection.

Nowadays, digital tools and AI-enabled platforms are increasingly used to support older adults' independence, cognitive functioning and social interaction (Bedaf et al., 2015; Charness and Boot, 2009). Emerging research starts exploring how digital technologies can facilitate positive aging by integrating AI-enabled services with person-centered service design. Following this stream of studies, the current research integrates AI-enabled transformative service literature and positive aging research to explore how AI-enabled service can be better designed to address older adults' psychological well-being needs. Particularly, this research utilizes AI-enabled self-learning service platform as study context and focuses on older adults' customer behaviors (i.e. service usage and referral intentions) and psychological well-being mechanisms (i.e. personal growth, purpose of life, social connectedness and satisfaction with life) to explore the optimal AI-enabled transformative service design for the aging population.

To meet their lifelong learning and personal growth needs, older adults increasingly turn to self-learning platforms as a flexible and empowering approach (Merriam and Kee, 2014; Morrison and McCutheon, 2019). Unlike the traditional classroom-based instruction, which often imposes rigid schedules, physical demands and age-related social barriers, self-learning accommodates the unique challenges many older adults face (Dunlosky and Hertzog, 1998; Jin et al., 2019). Aging is often accompanied by physical limitations such as reduced mobility, chronic health conditions or transportation challenges that can make attending in-person classes difficult or even unfeasible. Cognitive shifts, such as slower information processing or hearing and vision impairments, may also reduce the effectiveness of conventional instruction pace. Moreover, older adults may experience social deterrent, marginalization or anxiety about participating in the general learning spaces, which hinders their confidence or sense of belonging in classroom environments. These challenges are compounded by the desire for autonomy, individualized pacing and relevance in content, all of which self-learning service can offer (Boulton-Lewis et al., 2006; Roberson and Merriam, 2005). Moreover, in some cases, caregiving responsibilities or social isolation make self-learning the only viable option to continue intellectual stimulation and personal fulfillment (Boulton-Lewis et al., 2006). Therefore, older adults' turn to self-learning is not only a response to practical constraints but also a proactive strategy to remain independent and psychologically well-being in late life.

Technological advancements and improved digital literacy among aging populations further facilitate the rise of various self-learning services that align with older adults' learning preferences (Jin et al., 2019; Morrison and McCutheon, 2019), such as video-based self-learning courses and online or mobile self-learning platforms. Educational institutions and private providers offer a wide range of online courses through online self-learning service platforms for a low cost or even free (e.g. Senior Planet, Coursera, GetSetUp), which are particularly attractive to older adults as such services provide a convenient, low-pressure way to participate in lifelong learning without the barriers of traditional classroom environments. The emerging AI technologies further enhance older adults' interest in self-learning and their learning experience. Through natural language processing and machine learning, AI systems (e.g. “thinking AI”) can analyze customer profiles, usage preferences and patterns and behavioral data to deliver tailored service and learning support (Kumar et al., 2019). Personalized feeling AI enhances customer satisfaction, loyalty and engagement by providing relevant and empathetic service interactions (Huang and Rust, 2024; Kumar et al., 2019).

By meeting both practical and psychological needs, AI-enabled self-learning service may offer older adults a richer, more supportive and more personalized learning environment than non-AI service. In particular, such AI-enabled services can foster an enhanced sense of personal growth generated by the better learning and service experience – an important aspect of psychological well-being for aging population (Kim and Park, 2017). Older learners who perceive greater personal growth are likely to have deeper engagement with the service and more positive evaluations of the service. As a result, older adult learners not only would like to use such a personal growth enhancement service (i.e. AI-enabled self-learning) but also may be willing to recommend to the service to their peers. Therefore, the following hypothesis is proposed.

Additionally, AI-enabled self-learning service not only benefits service providers in terms of customer engagement and loyalty, but its transformative power can also provide customers a sense of purpose in life through fostered personal growth, which is critical for older adults' psychological well-being and positive aging (Bedaf et al., 2015). As individuals age, they often experience significant life transitions such as retirement, reduced social roles, loss of loved ones and health-related limitations. These life stage changes can lead to feelings of isolation, identity loss and diminished self-worth (Diggs, 2008). Well-designed AI-enabled self-learning service can help counteract these effects by offering a proactive and personalized way to stay intellectually engaged. Engaging in continuous learning helps older adults maintain mental sharpness and curiosity and keep personal growth. This sense of personal growth and progress contributes to a future-oriented mindset and self-worth, allowing them to look forward to new discoveries and achievements even in late life (Ryff and Singer, 1998). Hence, the following hypothesis is posited.

As AI capabilities evolve beyond mechanical and analytical tasks, increasing attention has turned to the emotional and relational dimensions of AI-enabled service, often referred to as “feeling AI” (Huang and Rust, 2024). AI systems are designed to detect sentiment and emotional cues from customer inputs, enabling more contextually aware and emotionally intelligent responses (Huang and Rust, 2024). This development supports the shift toward more empathetic service experience and customer care, aligning with the goals of transformative service research that prioritize customers' emotional and psychological well-being (Anderson and Ostrom, 2015). The emergence of interactive GenAI has made feeling AI a reality – now AI has the capability to engage in reciprocal communication and interaction with customers based on emotional detection from the emotion data analysis (Huang and Rust, 2018; Wirtz and Stock-Homburg, 2025). The communicative and interactive capabilities enable GenAI to generate humanlike responses demonstrating empathy, understanding and support based on the interaction context, and provides information, suggestions or consolation that are helpful to customers in addressing their unique needs or challenges (Huang and Rust, 2018).

Conceptual papers discussing AI-enabled service and customer care suggest that AI-generated communications with social presence such as humanlike features can increase customer evaluation, usage intention and lifetime value as well as improving customer well-being (Grewal et al., 2022; Huang and Rust, 2024). This proposition aligns with AI-enabled transformative service objectives – using affectively resonant AI or feeling AI to foster customer well-being and behavior change, whether through the social presence of virtual assistant or companion (Mehmood et al., 2024; Wang and Touré-Tillery, 2024). Thereby, applying the suggested advantages of social interaction function to AI-enabled self-learning service, the present research proposes that the humanlike interactive communication embedded in self-learning service can lead to enhanced transformative outcomes in terms of both customer engagement (i.e. service usage and referral intentions) and customer well-being (i.e. purpose of life) (Huang and Rust, 2024) among older adults as the following hypotheses.

Social isolation and loneliness can result in increased risks of depression, cognitive decline and mortality (Holt-Lunstad et al., 2015), particularly for older adults as they face increased physical, psychological, emotional and social limitations and challenges. To reduce social isolation and loneliness, researchers find that social interaction forms perception of social connectedness, which plays a vital role in positive aging (Bar-Tur, 2021). Thereby, this research argues that, together with the positive effect of self-learning on older adults' psychological well-being (Morrison and McCutheon, 2019), the social components may further improve older adults' service engagement and life satisfaction because of the enhanced social connectedness perception facilitated by the AI-enabled humanlike interactive communication embedded in the learning service experience. In addition, research finds that anthropomorphized AI, such as expressive language or humanlike avatars, can enhance user motivation by creating a sense of companionship and positive social relation during goal pursuit (Wang and Touré-Tillery, 2024), and such social connectedness consequently prompts customer engagement (e.g. service usage and referral) and improves life quality (e.g. satisfaction with life).

In sum, anthropomorphized AI agents not only foster emotional resonance but also enhance motivation and persistence in learning tasks through enhanced perception of social connectedness. As a result, the integration of interactive, humanlike communication into AI-enabled self-learning service may serve as both a cognitive support mechanism and a psychological buffer, contributing to older adults' well-being beyond traditional learning outcomes. Accordingly, the following hypotheses are proposed.

This research further explores the social strategy of inter-generational interaction in the current AI-enabled transformative service context by examining whether the inter-generational interaction facilitated by anthropomorphized AI persona can promote older adults' service engagement and positive aging through AI-enabled learning service. Inter-generational interaction – the purposeful exchange between people of different age groups, especially between older and younger generations – has emerged as an important strategy and mechanism supporting positive aging (Zhong et al., 2020). Underpinned by activity theory (Diggs, 2008) and socioemotional selectivity theory (Carstensen et al., 2011), empirical studies find that inter-generational interventions (e.g. co-learning activity) positively affect older adults' cognitive functioning, social connectedness, active aging and life satisfaction (Gualano et al., 2018). Inter-generational digital literacy programs, for example, not only empower older adults with new skills but also build reciprocal relationships with tech-savvy young generations, fostering mutual understanding and reducing age-related stereotypes (Farag et al., 2024).

Digital platforms and virtual inter-generational programs have started thriving, which is further accelerated by the recent pandemic (Farag et al., 2024). Virtual interaction and collaborative online programs show great potentials in maintaining social relevance, providing emotional support and combating isolation for older adults (Seifert and Schlomann, 2021). The integration of inter-generational strategies into AI-enabled service platform marks a new frontier in transformative service design. Through anthropomorphism effect, humanlike AI now enables service providers to simulate inter-generational exchanges by designing virtual companions that represent different age/generation personas. For instance, an AI agent may adopt the communication style or visual appearance of either a similar-age (older adult) or younger companion.

Theoretical insights from stereotype embodiment theory (Levy, 2009) suggest that older adults may respond more positively to youthful AI companion than to old-age AI companion. Younger personas may be perceived as more dynamic, energizing and socially stimulating (Hummert, 1990), thereby enhancing user motivation and emotional engagement. Moreover, simulated inter-generational exchanges may serve as a subtle but effective means of challenging internalized age stereotypes and promoting psychological vitality (Levy, 2009). As the positive role that inter-generation interactions may play in fostering positive aging (Farag et al., 2024; Zhong et al., 2020), the current research proposes that AI-enabled service may also be able to apply the inter-generational interaction advantage as the following hypothesis postulates.

Three progressive studies were conducted to examine the proposed hypotheses. Study 1 tested the basic effect – whether AI-enabled service increased older adults' service usage and referral intentions due to their perception of enhanced personal growth from an AI-enabled self-learning service experience (H1). Study 2 further explored how such an AI-enabled service and additional humanlike interaction embedded in the self-learning service experience might enhance older adults' preference for AI-enabled service and improve their purpose of life, which were mediated by their perceived personal growth gained from the experience (H2, H3 and H4). Lastly, Study 3 expanded the AI-enabled social component to inter-generational humanlike interaction to investigate whether AI could help to improve older adults' social connectedness perception, which in turn, enhanced their AI-enabled service engagement and satisfaction with life (H5, H6 and H7). Figure 1 presents the proposed conceptual model incorporating the three studies.

Study Design and Scenarios – A scenario-based experiment was conducted. Participants were randomly assigned to one of the two self-learning scenarios: either by following a self-learning video on pad without AI assistance or using an AI-enabled self-learning platform where in the learning process, the AI would adjust the teaching approaches and pace to help learners for a better self-learning experience. Following their personal preferences, participants chose from two self-learning options: new language or drawing.

Participants – The academic data collection platform of Prolific was employed to recruit participants for this research (prolific.com). A total of 109 qualified participants were recruited following the criteria of (1) U.S. participants, (2) English as primary language and (3) age of 55 years or above following the classification of older adults in consumer research (Gunter, 2012). The final sample included 52% female and 48% male. Participants' average age was 61 years old. Most participants (95%) were self-identified as Caucasian or White, 3% as African American and 2% as other ethnicity groups. Approximately 54% of the participants reported holding a bachelor's degree or higher. About 67% reported having an annual household income higher than $35 K.

Measures – The 7-poiont Likert scale was used to measure the two dependent variables and the one mediator: 1 = strongly disagree to 7 = strongly agree. Service usage intention was measured with a three-item scale adapted from Ma et al. (2024) by asking participants to rate to what extent they would like/consider/want to use such a self-learning service as depicted in the scenario they were exposed to respectively (Cronbach's alpha = 0.78). Service referral intention was measured by three items (adapted from Hwang et al., 2019), including questions such as “I may recommend this self-learning service to others” (Cronbach's alpha = 0.86). Personal growth was measured with a five-item scale adapted from previous literature (Springer et al., 2011), including questions like “I would feel more confident about trying new learning experiences than before” (Cronbach's alpha = 0.88).

The realism of the scenarios was assessed via the question: “How do you rate the realism of this scenario?” on a bipolar scale from 1 = very unrealistic to 7 = very realistic. Participants rated the scenarios highly realistic (M = 5.68). To check the manipulation of different self-learning approaches, participants were asked to recall whether the self-learning they were exposed to was AI-enabled or not. Participants' attentiveness was assessed by asking questions such as: “For this item, please select three.” Inattentive participants who failed manipulation or attention checks were automatically filtered out.

MANOVA Analysis – A one-way between-subjects multivariate analysis of variance was conducted on service usage and referral intentions. Results revealed that participants exposed to the self-learning scenario with AI assistance (vs. no AI) rated significantly higher on both dependent variables (Pillai's trace = 0.11, F_{(2, 106)} = 6.78, p = 0.002, partial η² = 0.11; service usage intention F_{(1, 107)} = 13.45, p < 0.001, partial η² = 0.11, M_AI = 5.75, M_no-AI = 5.071; service referral intention F_{(1, 107)} = 5.54, p = 0.020, partial η² = 0.05, M_AI = 5.60, M_no-AI = 5.14). The effect size indicator of partial η² indicated medium to large effects following the general thresholds of η² = 0.01 for small, η² = 0.06 for medium and η² = 0.14 for large effects (Cohen, 1988).

PROCESS Analysis – A mediation analysis was conducted following PROCESS Model 4 (Hayes, 2022). The results showed that personal growth fully mediated the independent variable of self-learning service on both dependent variables of service usage intention (direct effect – b = 0.16, 95% CI: −0.09–0.41; indirect effect – b = 0.52, 95% CI: 0.22–0.84) and service referral intention (direct effect – b = −0.04, 95% CI: −0.34–0.25; indirect effect – b = 0.50, 95% CI: 0.24–0.78) as the direct effect of IV turned to insignificant when the mediator was added to the model. Together with above MANOVA results, H1 was supported.

Study 1 confirmed the proposed basic effect regarding AI advantages and the effect sizes indicated substantial differences between AI-enabled and non-AI services, particularly in terms of service usage intention, suggesting that older adults significantly preferred the AI-enabled self-learning service because of their perception of enhanced personal growth gained from the AI-enabled experience (H1 supported). To further explore the transformative effect of AI-enabled service, Study 2 examined whether the additional social interaction embedded in the AI-enabled service experience may further enhance the basic effects discovered in Study 1.

Study Design and Scenarios – A scenario-based experiment study was conducted. Participants were randomly assigned to one of the three self-learning scenarios. Two were the same as Study 1: following a video on pad without any AI assistance or using an AI-enabled platform and AI would adjust the teaching approaches and pace to help learners for a better self-learning experience. The third scenario described a similar AI-enabled self-learning process with extra social interaction function: in the learning process, the AI communicated with the learner like an actual instructor – adjusting the teaching approaches and pace to help learner for a better learning experience, and from time to time, making some small talk conversations to keep the learner engaged.

Participants – Following the same recruitment criteria as Study 1, a new group of 155 older adult participants was recruited from Prolific. Participants' average age was 61 years old. Gender split was 47% female and 53% male. Most participants (92%) identified themselves as Caucasian or White, 6% as African American and 2% as other ethnicity groups. Around 51% of the participants held a bachelor's degree or higher. About 73% reported having an annual household income higher than $35 K.

Measures – Participants were asked about their self-learning service usage intention following the bipolar scale from 1 = unlikely/improbable/impossible to 7 = likely/probable/possible (Cronbach's alpha = 0.82) adopted from previous research (Park et al., 2025). Following the 7-point sale (1 = strongly disagree to 7 = strongly agree), purpose of life was measured via three items adapted from Springer et al. (2011) including questions such as “I feel I will have a better sense of what I'm trying to accomplish in my life” (Cronbach's alpha = 0.92). The measurements for service referral intention (Cronbach's alpha = 0.90) and personal growth (Cronbach's alpha = 0.92) followed the same scales as in Study 1.

Participants rated the scenarios highly realistic (M = 5.19). The same recall manipulation check (i.e. whether the service was AI-enabled or not) and attention check questions were asked and inattentive participants who failed manipulation or attention checks were filtered out. Additionally, the interaction manipulation was checked via the question “in the scenario's self-learning activity, to what extent do you feel interactions with others” on a 7-point scale from 1 = no interaction at all to 7 = high level of interaction. Results confirmed the successful manipulation of varying interaction levels among scenarios (F_{(2, 152)} = 77.73, p < 0.001). Specifically, the pairwise comparisons indicated that participants perceived the interaction with interaction AI (M_{interaction-AI} = 5.74) was significantly higher than non-interaction AI (M_{non-interaction-AI} = 2.45, p < 0.001) and no-AI (M_no-AI = 2.84, p < 0.001), but no significant difference between non-interaction AI and no-AI (p = 0.185) – these results confirmed the effectiveness of scenario manipulation.

MANOVA Analysis – A one-factor, three-level between-subject multivariate analysis of variance was conducted on service usage and referral intentions, and purpose of life. Results confirmed that participants showed varying responses to different scenarios across the three dependent variables (Pillai's trace = 0.22, F_{(6, 302)} = 6.14, p < 0.001, partial η² = 0.11; service usage intention F_{(2, 152)} = 18.72, p < 0.001, partial η² = 0.20; service referral intention F_{(2, 152)} = 10.56, p < 0.001, partial η² = 0.12; purpose of life F_{(2, 152)} = 9.33, p < 0.001, partial η² = 0.11). The effect size indicator of partial η² indicated medium to large effects (Cohen, 1988).

Pairwise comparisons further revealed that, compared with the two AI-enabled service, no-AI service received significantly lower responses, but varying results between the two AI-enabled service formats. For service usage intention, no-AI (M_no-AI = 4.25) was significantly lower than both non-interaction AI (M_{non-interaction-AI} = 5.32, p < 0.001) and interaction AI (M_{interaction-AI} = 5.79, p < 0.001), and there was marginal difference between the two AI-enabled services (p = 0.051). In terms of service referral intention, no-AI (M_no-AI = 4.46) was significantly lower than non-interaction AI (M_{non-interaction-AI} = 5.11, p = 0.010) and interaction AI (M_{interaction-AI} = 5.62, p < 0.001), significant difference was also observed between non-interaction AI and interaction-AI services (p = 0.034). As for purpose of life, no-AI (M_no-AI = 4.50) was rated significantly lower than non-interaction AI (M_{non-interaction-AI} = 5.14, p = 0.007) and interaction AI (M_{interaction-AI} = 5.52, p < 0.001) supporting H2 proposition, but no difference between the two AI-enabled services (p = 0.093). Therefore, H3 was supported but H4 rejected – significantly different impacts were revealed between the two types of AI-enabled service (non-interaction AI vs. interaction AI) on older adults' service usage and referral intentions (H3) but not on their purpose of life (H4). Figure 2 demonstrates the results.

PROCESS Analysis – PROCESS Model 4 was conducted to test the mediating role of personal growth on the three dependent variables. Results confirmed the proposed mediation effect on service usage intention (direct effect – b = 0.52, 95% CI: 0.30–0.74; indirect effect – b = 0.24, 95% CI: 0.11–0.39), service referral intention (direct effect – b = 0.23, 95% CI: 0.08–0.38; indirect effect – b = 0.35, 95% CI: 0.16–0.53) and purpose of life (direct effect – b = 0.19, 95% CI: 0.04–0.35; indirect effect – b = 0.32, 95% CI: 0.14–0.51). PROCESS indicator coding system (Hayes, 2022) was further used to contrast effect conditions with a baseline control condition (0 = interaction AI vs. 1 = non-interaction AI vs. 2 = no-AI). Across all the three dependent variables, results consistently showed that personal growth mediated the effect of contrasting interaction AI vs. no-AI on service usage intention (b = −0.47, 95% CI: −0.78∼−0.21), service referral intention (b = −0.70, 95% CI: −1.07∼−0.31), purpose of life (b = −0.63, 95% CI: −1.01∼−0.28); yet barely between the two types of AI (service usage intention – b = −0.23, 95% CI: −0.50∼0.03; service referral intention – b = −0.34, 95% CI: −0.72∼0.04; purpose of life – b = −0.31, 95% CI: −0.68∼0.04). These results further validated H1, and together with above MANOVA results, H2 was supported.

Findings from Study 2 replicated the basic effect found from Study 1 regarding the AI advantages – AI enhanced older adult consumers' service usage and referral intentions of the AI-enabled self-learning service (supporting H1), and further revealed AI advantages in improving older adults' psychological well-being manifested as purpose of life (supporting H2). The large effect sizes indicated practically strong effects induced by AI-enabled service, particularly impacting on service usage intention, followed by service referral intention and purpose of life. These AI advantages were mediated by enhanced personal growth that older adults experienced from the AI-enabled self-learning service. Furthermore, compared with no social interaction during the AI-enabled service, Study 2 discovered the enhancement effect of AI-enabled interaction function on older adults' service usage and referral intentions (H3 supported) but no impact on their purpose of life (H4 rejected). Additionally, compared with the full mediation effect revealed in Study 1, personal growth explained only part of the effects as it turned to partial mediation in this study. Therefore, to further examine the effect of AI-enabled social interaction in service and other potential mediator, Study 3 aimed to explore how AI-enabled social interaction function might affect older adults in other aspect of psychological well-being (satisfaction with life) and whether the inter-generational interaction effect might appear in the current research context.

Study Design and Scenarios – Participants were randomly assigned to one of the three AI-enabled self-learning scenarios. One was the same as the non-interaction AI scenario in Study 2 – participants used an AI-enabled self-learning platform and AI would adjust the teaching approaches and pace to help learners for a better self-learning experience but no AI-initiated interaction with learners. The two interaction AI-enabled service scenarios described a similar AI-enabled self-learning process with extra social interaction initiated by the AI: in the learning process, the AI communicated with the learner either like a similar-age older adult companion or a young-age companion sharing the same learning interest – the AI learning companions would adjust the teaching approaches and pace to help the learner for a better learning experience, and from time to time, making some small talk conversations to keep the learner engaged.

Participants – Following the same recruitment criteria as the prior two studies, a total of 153 older adult participants were recruited from Prolific. Participants' average age was 63 years old with 58% identified as female, 41% as male and 1% as others. Among participants, 85% identified themselves as Caucasian or White, 10% as African American, 2% as Latino, 1% as Asian and 2% as other ethnicity groups. Around 61% of the participants held a bachelor's degree or higher. About 80% reported having an annual household income higher than $35 K.

Measures – The same measurement scales for service usage intention (Cronbach's alpha = 0.95) and service referral intention (Cronbach's alpha = 0.94) were used as those in Study 2. Adapted from previous positive aging and life satisfaction literature (Diener et al., 1985; Ryff and Keyes, 1995; Springer et al., 2011), the measurement for perceived social connectedness was conducted following an eight-item scale including questions such as “I feel less lonely” (Cronbach's alpha = 0.94), and satisfaction with life via four items (Cronbach's alpha = 0.91) including questions like “I find myself in good spirits more”. Participants rated the scenarios highly realistic (M = 5.04). Manipulation check (“During the learning activity in the scenario, does the AI interact with you like a friend? If yes, what age-range does this AI friend look like – similar age as you or much younger?”) and attention check questions were asked, and inattentive participants who failed manipulation or attention checks were filtered out.

MANOVA Analysis – A one-factor, three-level between-subject multivariate analysis of variance was conducted on service usage and referral intentions, and satisfaction with life. Results confirmed that participants showed varying responses to different scenarios across the three dependent variables (Pillai's trace = 0.10, F_{(6, 298)} = 2.65, p = 0.016, partial η² = 0.05): service usage intention F_{(2, 150)} = 7.27, p < 0.001, partial η² = 0.09; service referral intention F_{(2, 150)} = 7.10, p = 0.001, partial η² = 0.09; satisfaction with life F_{(2, 150)} = 3.23, p = 0.042, partial η² = 0.04. The effect size indicator of partial η² indicated medium effects (Cohen, 1988).

Pairwise comparisons further showed that, compared with the two interaction AI service scenarios (old-adult vs. young-age AI personas), non-interaction AI service received significantly lower responses, but no significant differences between the two interaction AI-enabled service formats. For service usage intention, non-interaction AI (M_AI = 4.98) was significantly lower than both old-adult AI (M_old-adult-AI = 6.03, p < 0.001) and young-age AI (M_young-age-AI = 5.73, p = 0.009), but no difference between the two interaction AIs (p = 0.295). For service referral intention, non-interaction AI (M_AI = 4.78) was significantly lower than both old-adult AI (M_old-adult-AI = 5.59, p < 0.001) and young-age AI (M_young-age-AI = 5.46, p = 0.003), no difference between the two interaction AIs (p = 0.601). As for satisfaction with life, non-interaction AI (M_AI = 4.54) was lower than both old-adult AI (M_old-adult-AI = 5.11, p = 0.018) and young-age AI (M_young-age-AI = 5.00, p = 0.055), no difference between the two interaction AIs (p = 0.648). The results further validated H3 and supported H5; however, H7 was rejected regarding the proposed AI-enabled inter-generational interaction differences. Figure 3 depicts the results.

PROCESS Analysis – PROCESS Model 4 was conducted to test the mediating role of perceived social connectedness on the three dependent variables. Results confirmed the full mediation effect on service usage intention (direct effect – b = 0.21, 95% CI: −0.05∼0.48; indirect effect – b = 0.17, 95% CI: 0.04–0.32), service referral intention (direct effect – b = 0.17, 95% CI: −0.03∼0.37; indirect effect – b = 0.18, 95% CI: 0.05–0.32) and satisfaction with life (direct effect – b = 0.01, 95% CI: −0.17∼0.18; indirect effect – b = 0.23, 95% CI: 0.06–0.41). Then PROCESS indicator coding system was used to contrast effect conditions with the baseline control condition (0 = non-interaction AI vs. 1 = old-adult AI vs. 2 = young-age AI). For all the three dependent variables, results indicated the same pattern that perceived social connectedness mediated the effect of interaction AI: contrasting non-interaction AI vs. old-adult AI on service usage intention (b = 0.38, 95% CI: 0.14–0.68), service referral intention (b = 0.42, 95% CI: 0.16–0.70) and purpose of life (b = 0.56, 95% CI: 0.23–0.92); contrasting non-interaction AI vs. young-age AI on service usage intention (b = 0.31, 95% CI: 0.08–0.59), service referral intention (b = 0.34, 95% CI: 0.09–0.61) and purpose of life (b = 0.46, 95% CI: 0.12–0.82). Hence, H6 was supported.

Study 3 further confirmed the enhancing impact induced by the social interaction function in AI-enabled service on older adults' service usage and referral intentions (H3 supported) and their satisfaction with life (H5 supported). Although statistically medium, the effect sizes still suggested meaningful enhancement generated by the additional social function compared with AI-enabled service without such a social component. In addition, the study revealed the underlying mechanism of such interaction function advantages – enhanced social connectedness perception facilitated by the AI-enabled interactional service (supporting H6), fully explaining how AI-enabled service with social component could significantly improve older adults' responses and life quality. Interestingly, the inter-generational interaction advantages found by previous aging research among human-to-human interactions were not replicated in the current human-to-AI interaction context (rejecting H7).

This research provides empirical evidence supporting the transformative potential of AI-enabled self-learning services for older adults, particularly in enhancing both service engagement and psychological well-being. Through three progressive studies, the findings confirm that AI-enabled services significantly increase older adults' service usage and referral intentions, with these effects mediated by enhanced perceptions of personal growth (Study 1 and 2 – H1). The research further reveals that AI-enabled services can also improve older adults' sense of purpose in life, again through personal growth, thereby demonstrating AI's capability to foster meaningful psychological outcomes (Study 2 – H2). Additionally, incorporating social interaction features – such as empathetic, humanlike communication – into AI service although may not enhance sense of purpose in life (Study 2 – H4) but still increases customer engagement (Study 2 and 3 – H3) and overall life satisfaction (Study 3 – H5). And these significant positive effects are explained by elevated perceptions of social connectedness, highlighting the importance of relational components in AI service design (Study 3 – H6). However, the hypothesized inter-generational interaction effect, namely, that older adults might benefit more from AI companions anthropomorphized as young personas, was not supported (Study 3 – H7). This finding suggests a nuanced boundary condition in applying human-to-human aging literature to AI-mediated service experiences.

Together, these results contribute to both AI-enabled transformative service and positive aging literature by demonstrating how AI can serve as both a technological and psychosocial enabler. They underscore the importance of designing AI services that are not only functionally effective but also psychologically resonant, particularly for aging populations facing unique social and cognitive challenges.

This research offers several theoretical contributions that collectively advance understanding in the fields of AI customer care, transformative service and positive aging. By integrating GenAI with self-learning service platforms, the research illuminates how empathetic, personalized AI-enabled service can foster meaningful engagement and well-being among older adults, thereby extending existing literature in multiple ways.

The current research expands the extant AI-enabled service research by shifting the focus from efficiency and utility (Kim et al., 2025; Du et al., 2025) to psychological and emotional enrichment. Prior literature on AI in service has largely emphasized speed, accuracy and task completion through “mechanical AI” or “thinking AI” (Huang and Rust, 2018, 2021; Wirtz et al., 2018). While this research integrates “feeling AI” into the AI-enabled service (Huang and Rust, 2024), demonstrating that AI is now capable of delivering affective value and empathetic rapport through relational communication. The study contributes to the theoretical development of AI customer care by showing that interactive GenAI embedded with emotional intelligence – through sentiment responses and anthropomorphized features – can enhance not only service engagement but also transformative outcomes such as social connectedness and the consequent satisfaction with life. These findings suggest a transition from transactional to transformational customer care, highlighting that AI can support psychological empowerment and improve life quality for aging customers. The research therefore advances a new paradigm in service where AI does not merely assist but cares, suggesting its position as a co-agent in well-being co-creation.

The concept of transformative service emphasizes enhancing individual and collective well-being (Anderson et al., 2013; Anderson and Ostrom, 2015). The current research extends this notion by empirically validating AI-enabled transformative service as a meaningful driver of psychological benefits for aging populations. Building on the conceptual work of Huang and Rust (2024), this study provides concrete empirical support for how GenAI can manifest transformative service through personalization and humanlike social interaction. The study demonstrates that the emotional and cognitive features of AI-enabled services are not merely facilitators but mechanisms of transformation. Specifically, the three progressive studies together show how AI-enabled self-learning service increases older adults' service engagement through perceived personal growth and social connectedness, enhances purpose of life via growth and promotes life satisfaction via connectedness. These mechanism linkages clarify how GenAI serves as a “transformative agent” within service ecosystems, especially for marginalized or vulnerable populations like older adults (Fisk et al., 2018; Mende et al., 2024). In this way, this research reframes AI from a service delivery tool to a psychosocial catalyst, bridging the gap between transformative service theory and digital service design with empirical evidence support. The research enriches the AI-enabled transformative service literature by showing how socio-emotional capabilities of the new generation of GenAI can trigger well-being outcomes traditionally associated with human-to-human services.

This research also contributes to the evolving concept of “positive aging” by demonstrating how AI-enabled services align with the positive aging principles. Positive aging theory has evolved from focusing on physical health and independence to incorporating psychological growth, social engagement and purpose (Bar-Tur, 2021; Medina, 2017). The current research verifies that the right technological design, which is grounded in emotional intelligence and interactive learning, can function as an effective positive aging intervention. The research findings indicate that GenAI can be utilized as an enabler to provide older adults with a sense of personal accomplishment and social connection. This research further challenges the ageist narrative that older adults are technology-averse by showing that they actively engage with AI services when these are meaningfully designed to address their unique psychological and social needs. This confirms that positive aging can be technologically mediated and digitally sustained (Farag et al., 2024; Tully et al., 2025), enriching the positive aging literature by integrating technology as both medium and message for aging well (Medina, 2017).

Furthermore, the research's investigation into inter-generational AI interaction, although not statistically supported, raises new theoretical questions about the boundaries of stereotype embodiment and digital companionship in aging research. This opens future theoretical exploration into how identity, age representation and relational cues may shape older adults' response to AI. Previous research suggests that the positive impact of inter-generational interaction may come from both “receive” and “give” (Barbosa et al., 2021). In the current study context, older adult participants only experience receiving/learning but not much giving/mentoring. This paves the way for future investigation in how to promote effective inter-generational interactions through learning for giving – partner with community centers or senior service organizations to design hybrid AI + human inter-generational programs, blending online tools with in-person events or offering courses helping older adults to tutor and mentor young generations.

Another theoretical contribution lies in the two psychological causal paths revealed in this research and their respective facilitating factors. Drawing on Ryff's (1989, 2013) model of psychological well-being, this research confirms that the perception of personal growth achieved via AI-enabled learning can serve as a mechanism that fosters renewed life purpose among older adults. This adds theoretical clarity to how digital self-learning platforms may contribute to late-life psychological well-being. Unlike younger learners, older adults face unique social and physical constraints; yet this research shows that well-designed AI application can counter these barriers and provide a venue for older adults' personal development and sustaining a purpose of life when many of their social roles are fading way (e.g. retirement). This insight enriches both developmental psychology and transformative service literature, suggesting a new perspective of AI service as a purpose-eliciting system: AI-enabled self-learning platforms do not merely disseminate information but evoke existential fulfillment (Medina, 2017). And such a purpose-eliciting mechanism even does not request much social enabler.

Although the social interaction components do not directly enhance life purpose but significantly improve older adults' overall life satisfaction. The integration of humanlike interaction into AI-enabled services is shown to enhance older adults' perceived social connection, which in turn predicts higher life satisfaction. This supports longstanding theories that link social integration to well-being in aging (Carstensen et al., 2011; Holt-Lunstad et al., 2015), but it innovates by showing how the AI-enabled sociality can achieve similar outcomes as human-to-human social interaction. The current finding extends theoretical debates in human-technology interaction, arguing that the perception of social presence, not just the source, drives relational satisfaction and behavioral changes (Grewal et al., 2022; Wang and Touré-Tillery, 2024). AI-enabled interaction creates a psychologically valid form of companionship, serving as both a coping resource and motivational driver. This finding also supports and nuances socioemotional selectivity theory (Carstensen et al., 2011), demonstrating that emotionally meaningful exchanges, with whether human or synthetic, remain central to life satisfaction in older adulthood.

This research provides evidence-based insights that can guide practitioners in service industries, technology development and aging services in designing AI-enabled systems to promote aging well and help older adults to stay vital, happy and sharp in late life. With aging populations growing worldwide and digital transformation reshaping service industries, the current findings offer timely and practical guidance on how to build supporting, empowering and socially connected AI-enabled services for older adults.

First, this research informs how to design emotionally intelligent and adaptive “feeling AI” services for aging populations. Rather than focusing solely on analytical tasks such as recommendation algorithms, AI-enabled services should be equipped with empathetic, context-aware communication capabilities that foster cognitive, emotional and relational engagement (Huang and Rust, 2024). Humanlike communication that highlights warmth, encouragement and personalization can significantly enhance older adults' service engagement and psychological well-being. For example, AI-enabled service design needs to recognize and accommodate diverse customer profiles by customizing features based on varying cognitive capabilities, learning styles and emotional needs, thereby reducing technology-induced anxiety and enhancing usability through adaptable interfaces and interactive navigation. In addition, service providers may embed features such as sentiment detection, memory of prior conversations and simulated companionship in the AI-enabled service to create meaningful interactions that counteract loneliness and enhance life satisfaction.

The research also provides practical implications about how to address the psychological challenges of aging through service design. A major contribution of this research is identifying the linkage from AI-enabled self-learning to personal growth and subsequently to an enhanced sense of purpose in life (Ryff, 1989). Older adults would appreciate such service designs when their learning is self-paced, interest-driven and autonomy-supporting. Therefore, AI-enabled services may offer customizable learning progress, goal-tracking and opportunities to explore diverse topics. These features not only enable cognitive stimulation but also help users rebuild their sense of identity and direction, especially during challenging life changes such as retirement, bereavement or reduced mobility. Moreover, transformative service design needs to position older adults not as passive recipients but as empowered co-creators of their own service experiences, whether through feedback mechanisms, co-design roles or community-based mentorship programs that integrate AI tools.

In addition, this research emphasizes the role of AI in reducing social isolation and promoting life satisfaction through simulated social interactions. Service providers can leverage this advantage to foster AI-enabled digital companionship or develop hybrid service models that blend virtual tools with in-person engagements through interest-based matching and shared learning platforms (e.g. technology themed community center by Senior Planet from American Association of Retired Persons/AARP). Furthermore, service design needs to accommodate older adults' physical and cognitive changes by incorporating accessible features, such as large icons, simplified workflows and modular content that supports short, satisfying user sessions.

Finally, resonating with Fisk et al.’s (2018) “Design for service inclusion: creating inclusive service systems by 2050”, the present research calls for systemic support through public policy and institutional action. Governments and policy makers need to consider funding pilot programs and public–private partnerships that deliver affordable, inclusive AI services for older adults, particularly those in rural or underserved communities. By embedding AI-enabled learning and engagement tools in national aging strategies, society as a whole can move toward more inclusive models of positive aging.

As an early effort to explore how emerging GenAI and emotionally intelligent feeling AI can empower older adults for positive aging, this research has certain limitations that direct potential avenues for future inquiry. First, the current study only utilizes one service context (older adults use self-learning service platform to learn languages or drawing) to investigate the transformative potentials of AI-enabled service. Future research may continue integrating other relevant socio-environmental contexts, transformative service formats, technological advancements and psychological theories to develop inclusive models that support aging as a fulfilling and dynamic stage of life. Future theoretical investigation can also delve into the topic of technology-mediated social capital, that is, social connections and resources can be cultivated via well-designed anthropomorphized AI, especially in contexts of social isolation or marginalization like aging. This research is conducted only in one country (U.S.) through an online platform, which may result in participant bias. Future research can expand this study to a broader range and more diverse older adult sample, such as cross-cultural participants, to further validate the current findings.