Abstract
Keywords
Introduction
In the World Social Report 2023, the United Nations (2023) stated that the global population aged 65 and over was 761 million in 2021, expected to double to 1.6 billion by 2050. 1 The number of older adults living alone in China increased to 118 million in 2021, expected to exceed 200 million by 2030. 2 Since China has adopted the home-based pension model as its primary form, 90% of older adults over 60 live at home for the aged. However, the primary problem many older adults face living alone is how to fulfill their needs for social participation. Social participation is a fundamental factor influencing whether older people can live independently and actively. 3 According to the continuity theory, 4 older people's social needs do not diminish with the decline of physiological function or retirement. However, due to the decline in physical function and mobility of older adults, retirement from work, shrinking social circles, the impact of the postepidemic era, and the need to cope with the sadness caused by widowhood and the loss of loved ones and friends, 5 all of these factors can lead to a lack of social interaction between older people and the rest of society, resulting in their loneliness. Loneliness is associated with depression and has a synergistic effect, which, in turn, reduces the happiness of older people. Loneliness also increases the risk of death in older adults. 6 A large body of evidence shows that social isolation and loneliness in later life are risk factors for adverse physical and mental health outcomes, underscoring the need for scalable, home-based social connection supports among older adults. 7 Recent reviews indicate that older adults’ use of digital/social technologies—including social media and video-based information and communication technology (ICT)—is generally associated with reduced social isolation and modest gains in wellbeing, although effects vary by intervention and measurement.8–10 Studies have shown that social interaction and participation in social activities can not only help alleviate loneliness among older adults, 11 but also help reduce the risk of Alzheimer's disease. 12 Appropriate social activities are crucial for successful aging. 13 Therefore, enhancing the social participation of older adults living alone has emerged as a significant practical issue.
At the same time, technological advances have accelerated the advent of the digitalization and intelligence era, accelerating the intelligent innovation of the family unit system. The smart-home screen series is a category of products that has only begun to emerge in recent years. It is different from current household appliances in that it can control smart-home products and is equipped with a touchable large screen, allowing users to interact with it and perform operations such as video calls and online chats. In recent years, Amazon, Alibaba, Baidu, and other companies have released smart-home screen products. This kind of product is becoming increasingly popular and essential for smart aging. Intelligence and digitization are critical in helping older adults improve their lives. 14 Smart-home social media primarily refers to users utilizing smart-home screen products and leveraging mobile communication network technology to facilitate social interaction. Social applications are presented in the form of information systems, with diverse social content and formats encompassing voice, text, graphics, and video. Research have shown that older adults use social media for social engagement, which is conducive to enhancing their positive emotions and life satisfaction and further having a positive impact on health.15–18 The use of WeChat by older adults living alone in China can increase their intergenerational support and social activities, thereby improving their subjective wellbeing. 19 The inclusivity, recognition, and mobility satisfaction brought about by social media communication can provide social compensation for older adults. 20 Thus, it can be said that older adults’ use of social media may help them achieve a certain degree of social compensation, thereby enhancing their subjective wellbeing. However, whether smart-home social media can help older adults gain more social compensation remains unclear. A primary predictor of social compensation is computer-mediated communication (CMC), which possesses attributes that enable users to mitigate their psychosocial vulnerabilities. 21 Research indicates that psychological barriers and health obstacles are the most common reasons why older adults discontinue using the Internet, 22 making it difficult to enjoy the conveniences of new technologies and products. Therefore, a research question arises: as a type of information system, does smart-home social media possess attributes that can compensate for the psychosocial vulnerabilities of older adults?
Despite increasing research on older adults’ social media use, there is no concise, psychometrically validated scale that specifically assesses design-controllable features of smart-home social media delivering social compensation for older adults. This measurement gap limits cumulative evidence and design translation, motivating the development of a brief, design-oriented scale. Accordingly, this study aims to develop and validate a scale for social compensation design among urban older users in the context of smart-home social media from the perspective of information systems design. The theoretical motivation for approaching this from a design perspective is underscored by Gregor, 23 who emphasized the importance and necessity of design science research as a critical theoretical branch in the field of information systems. Supporting this viewpoint, Hooker 24 further asserted that, given the highly practical nature of design, its research could leverage theories from other disciplines, intersecting with design studies, to construct and enrich the research paradigm in design science. For instance, social psychology theories can be employed to aid design practice, some of which hold the potential to develop design science theories and exhibit certain patterns. The practical motivation for constructing the scale arises from the increasing focus of smart-home social media applications on establishing the acceptance behavior of older adults, aspiring to differentiate marketing strategies through social compensation, and designing competitive social media. However, all these efforts still need an implementable theory and a measurable tool. The purpose of developing the scale is to understand the multidimensional nature of social compensation design, which is crucial for establishing solid relationships between smart-home social media and older users. The scale measures the extent to which the design elements of smart-home social media can compensate for the vulnerable psychosocial aspects of older users during use, thereby enhancing their subjective wellbeing. This study examines the multidimensionality of social compensation design and discovers that, as a relatively new construct, it serves as an effective core variable for older users’ acceptance behavior towards smart-home social media. Social compensation design aids in understanding older users’ psychology and decision-making processes, prompting the design of smart-home social media to focus more on fulfilling their physiological and emotional needs. Therefore, understanding and deconstructing the multidimensional characteristics of social compensation design can help smart-home social media track and measure the degree of social compensation among older users. This understanding is critical to grasping the emotional connection between older users and smart-home social media, comprehending their needs, expectations, preferences, willingness, and behavioral patterns. Consequently, this leads to the design of social applications that align more closely with the psychological traits of older users, the adoption of appropriate marketing strategies, and the provision of strong physiological and emotional fulfillment for older adults, thus stimulating and maintaining their willingness and behavior to use smart-home social media.
Literature review
Social compensation
Tracing the origins of social compensation, Davis and Kraus 25 proposed the compensation hypothesis in the 1980s when studying the relationship between social behavior, loneliness, and mass media use, in which the medium was used to compensate for the lack of social connection. At that time, the coverage of telephones and mobile phones was not high, and the popularity of the Internet was even lower. Therefore, the research focused on the more traditional medium, such as the telephone, television, movies, books, and newspapers. The use of mass medium in this study compensated for the psychological loneliness caused by the lack of real-life social interaction. With the further development and popularization of the Internet, the medium has shifted from the traditional mass medium to the emerging medium, mainly smartphones, computers, and smart homes. In this type of medium, social media has emerged. Social media refers to the tools and platforms people use to share information, experiences, and opinions based on the Internet. They rely on the medium to provide people with social services and facilitate their online social activities. McKenna and Barge 26 were among the first to suggest that the Internet may be more beneficial to some people than others. They argued that people who experience high levels of anxiety in face-to-face social situations would find it easier to socialize via the Internet. They also suggested that lonely individuals lacking offline social connections may be more likely to turn to the Internet for re-establishing relationships. Since then, these early propositions have generated broader predictions that constitute the social compensation hypothesis.
In summary, it can be seen that medium compensation includes relatively more diverse types of medium, such as television, movies, music, phone calls, text messages, and even tapes that do not have access to the Internet, which can play a compensatory role. 27 Furthermore, social compensation is more to distinguish between face-to-face communication and online-mediated communication, which mainly refers to the emerging medium equipped with social media being used to compensate for the loneliness caused by the lack of social relationships among different groups.21,28 Social compensation means that people who have difficulties with offline face-to-face communication will compensate for offline deficits through online means.29,30 Social compensation is sometimes called the “poor-get-richer” hypothesis, 31 because online networks can compensate for inadequate offline networks. Based on the above research, Ma et al. 32 proposed that social compensation studies the process of substitution and psychological compensation by the medium carrying social media in a system for different groups of people who do not have enough social or interpersonal relationships. This eventually leads to equilibrium, strengthening social relationships, thereby enhancing their subjective wellbeing. Social compensation focuses on how CMC can provide relational benefits to individuals who have difficulty interacting with others face-to-face due to a lack of social skills or a low sense of wellbeing. 33 The main prediction of social compensation is that CMC has specific attributes that enable users to compensate for their psychosocial vulnerabilities. 21 For instance, owing to the characteristics of computers that enhance the ability to control messages or enable anonymity, online communication may be more comfortable than face-to-face interactions; it also facilitates users in easily finding like-minded individuals to establish connections. For these reasons, individuals with social psychological vulnerabilities are considered to have a greater inclination toward the online environment.
Two key elements of social compensation
Social compensation builds on the characteristics of online social interaction compared to offline face-to-face communication. As a result, individuals who cannot communicate face-to-face or have difficulty socializing offline are more likely to communicate through CMC. Thus, one of the core aspects of social compensation lies in its focus on the characteristics of the medium used. Wang and Shi 28 suggested that the attributes of the medium, as one of the elements of social compensation, can be analyzed from dimensions such as interactivity, temporality, accessibility, replicability, storage capacity, content permanence, retrievability, portability, social cues, and information capacity. In addition, another core element of social compensation is the different groups. Due to the differences in groups and the respective functions of the medium, the mechanisms of social compensation are also different. These mechanisms vary according to the specific physiological and psychological characteristics, needs, usage behaviors, and sociocultural environments of the different groups at their developmental stages, as well as according to the technical characteristics, functions, and affordances of the medium themselves, which release their compensatory potentials in their ways. Therefore, the mechanism of social compensation is an issue that requires comprehensive consideration of multiple factors.
Social compensation for different groups
The current research stage on social compensation for different groups focuses on adolescents, college students, adults, and older adults. Firstly, social compensation is supported in friendship formation in adolescents. For example, Selfhout et al. 34 collected questionnaire data from 307 Dutch secondary school students (mean age 15 years). The results of the study indicated that for adolescents who perceived friendships to be of lower quality, using the Internet for communicative purposes predicted less depression, whereas using the Internet for noncommunicative purposes predicted more depression and more social anxiety. Secondly, research on college students also supports social compensation. For example, Liu and Yang 35 investigated the potential relationship between social interaction needs and gaming experiences of 118 college students. The quantitative study results indicated that players’ level of social compensation was positively correlated with immersion. Thirdly, social compensation is also supported in online dating for adults. For example, Yang et al. 36 surveyed 459 users of online dating apps in China. The quantitative findings support social compensation, that is, a positive correlation exists between rejection sensitivity and the use of online dating apps. Finally, studies on older adult groups also support social compensation. For example, Ma et al. 32 examined the factors influencing social compensation design in 24 older adults who used smart-home social media to socialize online from the perspective of information system design. The results of the qualitative study indicated that the influencing factors of social compensation design included graphic features (GF), information architecture (IA), human–computer interaction (HCI), human interactivity (HI), intelligence (INT), socialization (SOC), shareability (SHA), user-generated content (UGC), social security (SE), and empathy (EMP). Kong and Lee 20 randomly sampled 392 older adults. They found that social media can help older adults gain social compensation.
In summary, although several scholars have researched social compensation for different groups, it is noteworthy that there has been a significant lack of attention to social compensation research for older adults. Strengthening compensatory interventions for older adults is essential for active aging. 37 Given the rapid growth of the world's aging population, social compensation research targeting older adults needs to be brought to the attention of the academic community. Therefore, this study develops and validates the social compensation design scale (SCDS) from the perspective of information system design to support future empirical research on social compensation based on older adults.
Social compensation for different medium
Although the medium's unique properties enable users to overcome the limitations of time and space and make up for claims that are difficult to realize in offline activities, it is still important to clarify that different mediums have different characteristics and functions. Therefore, the occurrence of social compensation to a certain extent depends on the choice of medium; that is, the use of different mediums will result in different social compensation, which is determined by the characteristics of the medium itself. The current research on older users of social media has mainly focused on the medium of smartphones. Online social activities through mediums such as mobile phones enable older adults to obtain higher social support and social connections. 38 Older adults can achieve higher levels of wellbeing by staying in touch with friends and family through smartphones.39,40 By expanding older adults’ interpersonal circle and following the trend, they can gain more social recognition, which makes them emotionally satisfied and compensated. 20 The use of WeChat by Chinese older users living alone has been shown to increase their intergenerational support and social activities, thereby enhancing their subjective wellbeing. 19 Furthermore, the inclusivity, recognition, and mobility satisfaction derived from social media communication can provide social compensation to older adults. 20 Evidently, social media is facilitating the establishment and maintenance of interpersonal relationships with unprecedented convenience, creating a form of social compensation for many individuals. 41
However, existing research on older social media users has largely overlooked other smart devices, such as smart-home screen products. These products have emerged as a result of the digitalization and intelligent development of home systems. According to the Chinese National Standard (GB/T 39190-2020), a smart-home application service platform is defined as one that supports internet protocols and browser functionality. It receives information from the Internet and IoT (Internet of Things) smart-home terminals, automates the management of home devices through these IoT terminals, and allows remote access to the smart-home network, community network, and Internet applications via the Internet. 42 Therefore, in this study, smart-home social media is defined as the use of smart-home screen terminals by users, leveraging mobile communication network technology, to facilitate social interaction from home. The social applications are presented in the form of information systems, with a variety of social content and forms, including voice, text, graphics, video, and services. Compared to traditional online social networks, smart-home social media features enhanced interactivity, real-time communication, and intelligence. Smart-home social media is particularly beneficial for older adults living alone, as it allows them to create and share content and engage in social interaction and participation without the constraints of time and space.
Currently, smart-home social media can be categorized into five types based on whether they are purely social:
Instant Messaging Social Media: Examples include WeChat, QQ, and DingTalk. These platforms primarily focus on instant messaging and leverage their large user base to become more tool-oriented, extending into other areas. Information Social Media: This category can be further divided into graphic/text-based and short video social media. Representative apps are Xiaohongshu (Little Red Book), Toutiao, Douyin (TikTok), and Kuaishou. While social interaction remains a primary feature, instant messaging is less emphasized. E-commerce Social Media: Examples are Taobao, JD.com, and Pinduoduo. These platforms primarily focus on e-commerce, with social interaction mainly occurring through sharing, purchasing, and group buying among acquaintances. This type of social interaction is considered “light,” and users do not typically use these platforms as their main social media. Service-Oriented Social Media: Such as Alipay and JD Finance. These platforms are primarily focused on payments and various services, also utilizing acquaintance-based social interaction to enhance user engagement. Entertainment and Gaming Social Media: Examples include card games like “Fight the Landlord” and chess games. These platforms are primarily gaming-oriented, using social features to increase online rates and sharing, and are also considered “light” social interaction platforms.
In light of the above, it becomes apparent that further research into the social compensation of older users using smart-home social media is merited. Currently, no researchers have embarked on describing and measuring social compensation from a process-oriented perspective. Therefore, this study focuses on the context of smart-home social media, and from the perspective of information systems design, aims to develop and validate a scale for social compensation design targeted at older users, thereby underpinning subsequent empirical research in this field.
Methodology
This study was divided into three phases (Figure 1), namely, development of SCDS (Phase 1), exploration of SCDS (Phase 2), and validation of SCDS (Phase 3). Phase 1 was mainly to define and form the initial scale from the analysis of related literature based on the influences of social compensation design. Then, the experts further evaluated the content validity of the initial scale and deleted the poor items. Phase 2 collected effective questionnaires from 340 urban older users and conducted reliability and validity tests using SPSS 25.0 software. Then, exploratory factor analysis (EFA) was conducted to explore the scale's construct validity through principal component analysis (PCA). Four principal components were extracted, 12 poor items were deleted, and 29 remained. Phase 3 collected a total of 357 valid questionnaires, and confirmatory factor analysis (CFA) was carried out by Amos 28.0 software to verify the rationality of the scale structure. The reliability, convergence validity, discriminant validity, and model fit were calculated. The factor loadings, composite reliability (CR), and average variance extracted (AVE) of each measurement model and the chi-square values/degrees of freedom (CMIN/DF), root mean square of the approximation error (RMSEA), standardized residual root mean square (SRMR), comparative fit index (CFI), adjusted goodness-of-fit index (AGFI), and Tucker–Lewis Index (TLI) of the combined model were reported. By deleting 13 poor measurement items, all the above indicators were statistically excellent. Finally, this study proposed a measurable and validated multidimensional SCDS containing 4 dimensions and 16 measurement items. The exact wording and response anchors of the validated 16-item SCDS are provided in Supplemental File S1. Reporting followed the COSMIN guidance for studies on measurement properties where applicable.
Three phases of the study.
Two types of participants were involved in this study. One group consisted of four experts from the HCI design, participating in the scale development phase. The other group comprised urban older users in China who participated online in two rounds of data collection—the exploratory phase and the validation phase.
During the expert interview phase, the four experts invited were from different universities in China and Italy, comprising one professor, one associate professor, one assistant professor, and one lecturer, specializing in design. Two of them were from the School of Design at Hunan University and Qingdao University of Technology in China, while the other two were from the Design Department at Politecnico di Milano in Italy. The primary areas of expertise for these four experts were visual and information design, as well as psychological and user behavior studies.
According to the regulations of the State Council of China (1978) 43 on retirement age, workers in state-owned enterprises, institutions, and mass organizations who have not engaged in heavy physical or health-damaging work can apply for retirement when men reach 60 years, and women reach 50 years of age, provided they have at least ten consecutive years of work experience. Additionally, considering definitions of older adults in other countries, the American Association of Retired Persons (2024) 44 also describes older adults as individuals aged 50 and above. Moreover, Anderson and Langmeyer 45 noted that age can be used for market segmentation, identifying motivational, planning, and cost differences between age groups over and under 50 years.
Based on this, the target population defined in this article is urban older individuals in China who are 50 years and older, living alone, and are currently using smart-home social media for social communication. This age range definition is also consistent with other recent studies.46–48 The selected target group for this study is based on a key pre-requisite: participants must be active internet users, as this is a fundamental requirement for using smart-home social media. Younger older individuals, compared to other age groups of older adults, have higher digital literacy and are more willing to use trendy smart devices. This study anticipates that as more older people start using smart-home social media, the research findings can be extended to other age groups of the older population.
Alkis et al. 49 pointed out that collecting data through methodology of paper and online questionnaires might impact the results in an as-yet-undiscovered manner. They advised using either of these methods as the sole medium for data collection to ensure the validity and reliability of the scale. Online surveys are a rapid and cost-effective means to obtain responses from online users. 50 Given that this study examines the behaviors of urban older users of smart-home social media in China, and considering factors such as pandemic-related restrictions, internet-based online surveys are deemed an appropriate method for data collection. Two rounds of data collection were conducted using online questionnaires created on the Tencent Questionnaire payment platform, with the questionnaires distributed through Tencent Questionnaire's Aging Group. Efforts were made to ensure the authenticity and validity of the collected data during the data collection phase. Tencent Questionnaire has accumulated a group of over 500,000 older users, forming an aging group that covers 300 cities across China. This aging group can be utilized to invite members to participate in any specific study based on the criteria of each research project.
The SCDS was developed and psychometrically validated within this study (item generation and expert review, followed by EFA and CFA across two independent samples). Therefore, it is not a previously validated external scale but one established herein. The questionnaire used in both rounds of data collection was divided into three parts: (1) Introduction, explaining the purpose, social value, scope of information collection, potential privacy risks, and countermeasures, as well as some of the terminology involved in the questionnaire; (2) the multidimensional SCDS, consisting of measurement items; (3) basic user information. Studies have shown that large-scale scales are significantly better than small-scale scales regarding reliability and validity. 51 Therefore, all items were measured using Likert's seven-point equidistant scale. 52 Among them, “1” means strongly disagree, “2” means disagree, “3” means relatively disagree, “4” indicates uncertainty, “5” indicates relatively agree, “6” indicates agree, and “7” indicates strongly agree. The last item in the questionnaire design is a simple question (Please note that this study is important. Please check “I don't know”), which checks whether participants are focused and giving meaningful responses, as choosing the answers “I know” and “I don't care” reflects the respondent's carelessness.
Participants were urban older adults living alone (inclusion: age ≥ [50/60], living alone, having experience with smart-home social media). Recruitment used an older-adult online user panel covering multiple cities in China. Two independent waves were fielded with no repeat participation. Quality control included attention-check items, minimum completion-time thresholds, and pattern-response screening; inconsistent or low-quality responses were removed prior to analysis. To obtain high-quality data, the two rounds of data collection used rigorous procedures and systems to screen responses. If a respondent answered a survey question for too short (less than 80 seconds) or too long (more than 1000 seconds), if the respondent's answers followed a pattern (e.g., all 1 or all 7), or if the respondent answered the last attention-testing question incorrectly (selecting “I know” and “I don't care”), the respondent's data were deleted because the pattern indicated that the responses were not credible. 53 Respondents were not allowed to participate in the survey more than once, and each respondent was tracked by their WeChat ID. As a bonus, each survey respondent received a cash prize.
Phase 1: Development of SCDS
Wang and Shi 28 suggested that the attributes of the medium, as one of the elements of social compensation, can be analyzed from dimensions such as interactivity, temporality, accessibility, replicability, storage capacity, content permanence, retrievability, portability, social cues, and information capacity. Building on this, Ma et al. 32 examined the factors influencing social compensation design in 24 older adults who used smart-home social media to socialize online from the perspective of information system design. The results of the qualitative study indicated that the influencing factors of social compensation design included GF, IA, HCI, HI, INT, SOC, SHA, UGC, SE, and EMP.
This study builds on the research of Ma et al. 32 and, referencing relevant literature and combining the characteristics of smart-home social media with the authors’ personal experiences in using such media, defines the factors influencing social compensation design from the perspective of information systems design, as shown in Table 1. Based on this, the study further developed an initial scale for social compensation design among older users in the context of smart-home social media. This scale comprises 10 dimensions and 50 measurement items, 54 as presented in Table 2.
Definitions of social compensation design.
Initial scale.
Phase 2: Exploration of SCDS
To further test the scale, reliability and validity analyses were conducted to screen the scale items and form the final scale. EFA was utilized to test the scale's validity. In the development of scale, EFA can be used to continuously attempt to achieve the optimal state of factor structure and establish structural validity by deleting measurement items. 84 Therefore, through EFA, various latent variables and measurement items of the scale can be determined.
At this phase, the questionnaire was designed based on the revised scale with the participation of experts. The SCDS, consisting of 41 measurement items, was mainly used to measure 10 dimensions. Four hundred questionnaires were collected in this survey, excluding 60 invalid questionnaires and 340 valid ones. In addition, according to the recommendations of Zeng et al. 85 and MacCallum et al., 86 it is best to have a sample size of 5 to 10 times the measurement items in factor and regression analysis. Therefore, the SCDS for this phase measured 41 items, with 340 valid questionnaires. The sample size is about eight times that of the measurement items, which meets the requirements of factor analysis. The demographic statistics are shown in Table 3. All 340 participants in the study had at least 6 months of experience using smart-home social media. Therefore, they were more familiar with the interfaces of smart-home social media compared to other older individuals.
Demographic statistics.
Phase 3: Validation of SCDS
After EFA, CFA was used to ensure the structure validity made sense. It means the relationship between latent variables and measurement items and the fit between the scale and the actual data was checked.
A questionnaire was designed based on the 29 selected measurement items at this phase. Four hundred twenty-one questionnaires were collected, excluding 64 invalid questionnaires and 357 valid ones. In addition, regarding the sample size, structural equation modeling (SEM), as a large-sample sampling analysis method, Hair 87 suggested that the sample size should generally be 10 to 15 times the number of measurement items. After counting 72 papers on SEM, Loehlin and Beaujean 88 found that the median sample size was 198. Barrett 89 argued that since SEM generally uses the built-in maximum likelihood method, the chi-square value will expand seriously when the sample size is greater than 500, resulting in a poor fit of the model. Therefore, the scale measured 29 items, and the valid questionnaires were 357. The sample size is about 12 times more than the measurement items, and the number of valid questionnaires is less than 500. The number of samples needed for factor analysis is satisfied. The demographic statistics are shown in Table 4. All 357 participants in the study had at least six months of experience using smart-home social media. Therefore, they were more familiar with the interfaces of smart-home social media compared to other elderly individuals.
Demographic statistics.
Results
Phase 1
The study continued to use the Delphi method to ensure that the content of the initial scale reaches a certain level of validity. Lynn 90 suggested that at least three experts should be invited at this phase, but no more than 10 experts at most. Therefore, four experts in HCI design were invited to assess the content validity of the initial scale in this phase. The purpose was to check that the initial scale was suitable for measuring the social compensation design influences of older smart-home social media users and the clarity and accuracy of the measurement items. In the specific implementation process, their consent was first obtained, and then during the Tencent meeting, the background, research objectives, and research methodology of this study were introduced to them. Then, the “Request for Revision of the Initial Scale of Social Compensation Design for Older Adults Based on Smart-home Social Media” was distributed to them. They were invited to provide their opinions on the clarity of each specific measurement item, the correlation between measurement items under each dimension, and the degree of interpretation of the measurement items for the higher dimensions.
Subsequently, feedback from the four experts was collected, and a preliminary summary of each expert's comments was organized. After multiple rounds of careful consideration and comparison, items that were deemed repetitive or controversial by the experts were removed. Additionally, measurement items that were semantically ambiguous or difficult to understand were modified and rephrased. Following this round of evaluation by the experts, a total of 9 measurement items were deleted from the initial scale, leaving 41 items. After expert review, the initial scale was refined, yielding the “Expert-Modified Initial Scale” (Table 5). To facilitate the following exploratory factor analysis, measurement items for the dimensions of GF and IA were represented by UIQ + numbers. The dimensions of HCI, HI, and INT were represented by IQ + numbers. The dimensions of SOC, SHA, and UGC were represented by CQ + numbers. Finally, the dimensions of SE and EMP were represented by SQ + numbers.
Expert's revised initial scale.
CQ: content quality; IQ: interaction quality; SQ: service quality; UIQ: user interface quality.
Phase 2
Reliability analysis
Testing the data quality of the measurement results is an essential pre-requisite to ensuring the subsequent correlation analysis. The internal consistency of each dimension is generally examined through Cronbach's alpha coefficient for reliability testing. The value of Cronbach's alpha coefficient ranges from 0 to 1, and the higher the result value, the higher the reliability. Generally, a Cronbach's alpha coefficient below 0.6 is considered to have failed the reliability test, and the questionnaire needs to be redesigned, or the data needs to be recollected and analyzed again. In exploratory research, Cronbach's alpha coefficient between 0.6 and 0.7 indicates that there is reliability, and the reliability is acceptable. A Cronbach's alpha coefficient between 0.7 and 0.8 is relatively reliable, between 0.8 and 0.9 is reliable, and between 0.9 and 1 is very reliable. 91
After calculation (Table 6), Cronbach's alpha coefficients of all latent variables are greater than 0.7, some are greater than 0.8, a few are greater than 0.9, and the overall Cronbach's alpha coefficient is 0.958. The results indicate good internal consistency and reliability among the measurement items in this survey questionnaire. The sample measurement data has high reliability, which meets the requirements for reliability between measurement items proposed by Hair et al. and Fornell et al.
Reliability analysis.
Validity analysis
Due to the rigorous evaluation of the validity of the scale content before the empirical research, all measurement items were checked. Therefore, the scale has good content validity. Then, before conducting EFA, the KMO (Kaiser–Meyer–Olkin) test and Bartlett's sphericity test were conducted to verify the suitability of factor analysis among the measurement items. 92 This study uses the KMO test to see if the sample data is good enough for further factor analysis. It works by checking to see if some information overlaps between the measurement items in each latent variable in the sample data. 93 The KMO should be greater than 0.7 94 and at least greater than 0.6. The p-value for the Bartlett's sphericity test should be less than 0.05. According to the analysis of the data using SPSS 25.0 software, it was found that the KMO value of the scale was 0.946, and the Bartlett Test was significant at the level of 0.000 (Table 7). Therefore, the results indicate that there are common factors between the measurement items of this questionnaire, which are very suitable for factor analysis.
KMO and Bartlett's sphericity test.
KMO: Kaiser–Meyer–Olkin.
EFA
This study used PCA to conduct EFA and explore the structural validity of the scale. The maximum-variance orthogonal rotation method was used to rotate factors. Factors with rotated eigenvalues greater than 1 were chosen to retain common factors. We then tested whether the factor structure reached its optimal state and whether each item could be attributed to an explicit latent variable, thus forming a reasonable factor structure. 95 If the factor structure was unreasonable, items were deleted to achieve the ideal situation. Tabachnick et al. 96 noted that when the factor loading was greater than 0.55, it explained about 30% of the variance of the measured items (a good situation); when the factor loading was greater than 0.71, it explained about 50% of the variance (ideal). 97
In this study, SPSS 25.0 was used to perform EFA on the measurement items, resulting in the extraction of four principal components. Items were retained when the primary loading ≥ 0.55 and the factor pattern was conceptually consistent; items were eliminated based on a priori rules (e.g., low primary loading < 0.55 or salient loading on an unintended factor/substantial cross-loading). Accordingly, UIQ5, IQ10, IQ12, IQ13, IQ14, IQ15, IQ18, IQ19, SQ38, SQ39, SQ40, and SQ41 were removed. Moreover, since the first principal component included all measurement items from the dimensions of SOC, SHA, and UGC, this indicated a close semantic relationship among these items. Therefore, it was necessary to further assess the independence of residuals during CFA and eliminate items with closely related meanings.
Through EFA, 12 items were deleted, leaving 29 items. The retained items included essential parts of relevant literature and interviews, and all items had factor loadings greater than 0.55. The 29 items did not have multiple loadings, and the total variance explained by the four factors was 61.788%, as shown in Table 8. Item-level EFA loadings and retain/drop decisions are provided in Supplemental File S2 (sheet S2-EFA-41), together with the a priori rules. Meanwhile, considering that the questionnaire survey method could lead to multicollinearity issues, Harman's single-factor test 98 showed that the total variance extracted by a single factor was 29%, far below the expected threshold of 50%. Therefore, it indicated that the designed scale had appropriate factor classification and good validity. Finally, the first principal component was named content quality (CQ), the second was named user interface quality (UIQ), the third was named service quality (SQ), and the fourth was named interaction quality (IQ).
The results of EFA.
Note: Extraction method: PCA; Rotation method: Kaiser normalized maximum variance method.
CQ: content quality; IQ: interaction quality; PCA: principal component analysis; SQ: service quality; UIQ: user interface quality.
Phase 3
The four principal components and their measurement items were used as the measurement models. CFA further validated the validity of measurement items in Amos 28.0. Segars 99 argued that poorer measurement models might lead to erroneous findings and conclusions; Anderson and Gerbing 100 argued that a good measurement model is a prerequisite for further causal analysis; Jackson et al. 101 argued that among the general SEM research process, the measurement model is usually assessed first and the structural model is assessed after meeting the conditions; Brown 102 argued that in many cases, problems in SEM are due to the measurement model and that these problems can usually be identified and resolved through CFA; Byrne 103 argued that a better model fit metric is a pre-requisite for further analysis such as regression in SEM. A better model fit metric indicates that the matrix of the SEM theoretical model is closer to the sample matrix.
Therefore, this article is based on the criteria for convergent validity proposed by Hair, 87 Fornell and Larcker 91 : (1) Standardized factor loadings for each measurement item should be greater than 0.5; (2) CR for each dimension should be greater than 0.6; and (3) AVE for each dimension should be greater than 0.5.
Model fit was analyzed using the recommendations of Jackson et al. 101 for the CMIN, DF, CMIN/DF, RMSEA, SRMR, CFI, goodness-of-fit index (GFI), AGFI, and TLI values. 104 Among them, the smaller the CMIN, the better; the larger the DF, the better; the CMIN/DF should be greater than one and less than 3; the RMSEA should be less than 0.08; the SRMR should be less than 0.08; and the CFI, GFI, AGFI, and TLI should all be greater than 0.9, which means that the model is well fitted. 105
Validation of the UIQ measurement model
The standardized factor loadings of the UIQ measurement model ranged from 0.578 to 0.722, with a CR of 0.813 and an AVE of 0.422. Since the AVE was less than 0.5, the measurement items UIQ3, UIQ4, and UIQ6 were deleted according to the recommendations. After deleting them, the AVE was 0.518. The results showed that the measurement model was reliable, with good reliability and convergent validity.
Validation of the IQ measurement model
The standardized factor loadings of the IQ measurement model ranged from 0.475 to 0.802, with a CR of 0.811 and an AVE of 0.469. Among them, the factor loading of IQ9 was 0.475, which was lower than 0.5 and should be deleted. After deletion, the CR was 0.818, and the AVE was 0.533. The results showed that the measurement model was reliable and had good reliability and convergent validity.
Validation of the CQ measurement model
The standardized factor loadings of the CQ measurement model ranged from 0.665 to 0.809, with a CR of 0.946 and an AVE of 0.558. The results indicated that the measurement model was reliable, with good reliability and convergent validity.
Validation of the SQ measurement model
The standardized factor loadings of the SQ measurement model ranged from 0.723 to 0.874, with a CR of 0.875 and an AVE of 0.637. The results showed that the measurement model was reliable and had good reliability and convergent validity.
Validation of the combined model
In the model fit, the CMIN/DF was 3.296, which is acceptable. The RMSEA was 0.08, which is unacceptable. The SRMR was 0.06, which is excellent. The CFI was 0.885, which is unacceptable. Therefore, the model needs to be adjusted. Item-level standardized CFA loadings for the initial 29-item model and the corresponding retain/drop decisions and rules are provided in Supplemental File S2 (sheet S2-CFA-29). Based on the Modification Indices (MI) value, the MI value between measurement items IQ10 and IQ11 was the highest at 53.630. It means there is residual nonindependence here, which goes against the principle of residual independence. IQ11 was removed because its standardized factor loading was higher than that of IQ10. The MI between CQ15 and CQ16 was 52.171, which indicates that there is residual nonindependence here, violating the principle of residual independence, and because CQ15 and CQ16 have the same standardized factor loadings, CQ15 and CQ16 were deleted. The same analytical principle continued deleting measurement items CQ13, CQ14, CQ18, CQ22, CQ24, and CQ25.
The results of the modified model fit are shown in Table 9, with all nine indicators reaching the level of excellence. It indicates that the model constructed in this study has a relatively good fit. The reliability and convergent validity of the modified model are shown in Table 10, and all indicators meet Hair's suggested criteria, indicating that the measurement items are reliable. It can be seen from Table 11 that the diagonal bold is the arithmetic square root of AVE, and the lower triangles are Pearson's correlation coefficients between each construct. This set of values is consistent with the recommended standard of Bagozzi and Yi, 106 that is, the arithmetic square root of AVE should be greater than Pearson's correlation coefficients of each construct. It suggests that the model has good discriminant validity among constructs.
Indicators of model fit.
AGFI: adjusted goodness-of-fit index; CFI: comparative fit index; CMIN/DF: chi-square values/degrees of freedom; GFI: goodness-of-fit index; RMSEA: root mean square of the approximation error; SRMR: standardized residual root mean square; TLI: Tucker–Lewis Index.
Reliability and convergent validity of the combined model.
Note: * p < 0.050, ** p < 0.010, *** p < 0.001.
AVE: average variance extracted; CR: composite reliability; CQ: content quality; IQ: interaction quality; SMC: squared multiple correlations; SQ: service quality; UIQ: user interface quality.
Discriminant validity of the combined model.
Note: * p < 0.050, ** p < 0.010, *** p < 0.001; diagonal bold is the arithmetic square root of AVE.
AVE: average variance extracted; CR: composite reliability; CQ: content quality; IQ: interaction quality; MSV: maximum shared variance; SQ: service quality; UIQ: user interface quality.
The standardized path structure of the combined model is shown in Figure 2. Based on the above analysis, this article finally obtained the SCDS, as shown in Table 12. The scale consists of 4 dimensions and 16 measurement items. Among them, UIQ is measured by three items; IQ is measured by three items; CQ is measured by six items; and SQ is measured by four items.
Standardized path structure.
Social Compensation Design Scale (SCDS).
CQ: content quality; IQ: interaction quality; SQ: service quality; UIQ: user interface quality.
Discussion and conclusion
The development and validation of the SCDS were conducted strictly with standardized procedures and principles. After the validity assessment of the initial scale content and two further rounds of empirical research, a formal measurement scale with four dimensions was formed and obtained good reliability and validity. The results of EFA showed that the theoretical structure of the scale had a good match with the actual data. In the context of smart-home social media, the social compensation design of older users is mainly manifested in UIQ, IQ, CQ, and SQ. The results of EFA also showed that the four dimensions of the scale differed significantly in content and structure, reflecting that social compensation design is a structure composed of four dimensions. The results of the CFA indicated that the modified model had excellent indicators and a reasonable degree of fitting. Moreover, the CR of all four dimensions was greater than 0.7, and the AVE was greater than 0.5. This indicates that the scale has good internal consistency and its structure has good stability and reliability,87,91 which can be used for quantitative measurement. Therefore, the scale development process is scientific and rigorous, and the measurement has good content validity. In summary, the SCDS developed in this study has good reliability and validity, providing a basis for subsequent quantitative research on social compensation.
Existing research on the acceptance behavior of older users of social media focuses on discussions about relatively limited antecedent variables such as demographic characteristics, social factors, technological factors, and particular theoretical perspectives. It lacks a social compensation design perspective to understand and validate older adults’ adoption of social media. Ma et al. 32 explored the factors influencing social compensation in older adults from the smart-home social media perspective. The study pointed out that social compensation is the process of psychological substitution and compensation of smart-home social media to older adults with insufficient social relationships, which ultimately leads to equilibrium, that is, a complement to the social relationships, which leads to older adults’ perception and formation of positive emotions, and ultimately the acceptance of the use of the process. It is also supported by the study of Wang and Shi. 28 On this basis, this article develops and validates the SCDS for older users in the context of smart-home social media. The cumulative explainable variance of the four dimensions of the SCDS is 61.788%, of which the CQ explains 29.524% of the variance, the UIQ explains 11.542% of the variance, the SQ explains 10.382% of the variance, and the IQ explains 10.34% of the variance. It indicates that the designed scale factors were categorized appropriately.
Theoretical contributions
Based on the previous study, 32 this study collected data through expert interviews and two rounds of questionnaires and developed the SCDS using standardized procedures. As the SCDS was developed based on the prior research and literature review, there was a clear pathway for forming each measurement item, dimension, and relationship, thus ensuring the completeness and clarity of the SCDS as a structured scale. There has been no quantitative research on social compensation design before, which has led to confusion about the underlying issues of social compensation design. Therefore, this study explored the concept and structure of social compensation design through a standardized procedure tested for validity and reliability. This study not only develops the SCDS for the smart-home social media context based on the perspective of Chinese urban older users but also bridges the gap in the previous literature on the role of mechanisms in the use behavior of older social media users.
In previous studies on smart-home social media contexts, scholars have not effectively measured the structure of social compensation design. This article extends social compensation to smart-home social media contexts and develops the SCDS by combining expert interviews and collecting data from Chinese older smart-home social media users to better describe the predictive ability of older users’ perceptions and behaviors in smart-home social media contexts. It not only provides a specific extension and application of social compensation theory from general media to the specific virtual environment of smart-home social media but also provides a new perspective and valuable reference for the study of the compensation relationship between older users and smart-home social media and provides a foundation for future empirical studies related to social compensation.
Practical implications
The increasing focus of smart-home social media applications is on establishing the acceptance behavior of older adults, aspiring to differentiate marketing strategies through social compensation, and designing competitive social media. However, all these efforts are still in need of an implementable theory and a measurable tool. The social compensation design measurement tools developed in this study are divided into four dimensions. Among them, CQ explains the most significant difference, which suggests that social media developers should pay special attention to strengthening social content quality when establishing connections between older users and smart-home social media.
The purpose of developing the scale is to understand the multidimensional nature of social compensation design, which is crucial for establishing strong relationships between smart-home social media and older users. The scale measures the extent to which the design elements of smart-home social media can compensate for the vulnerable psychosocial aspects of older users during use, thereby enhancing their subjective wellbeing. This study examines the multidimensionality of social compensation design and discovers that, as a relatively new construct, it serves as an effective core variable for older users’ acceptance behavior towards smart-home social media. Social compensation design aids in understanding the psychology and decision-making processes of older users, prompting the design of smart-home social media to focus more on fulfilling their physiological and emotional needs. Therefore, understanding and deconstructing the multidimensional characteristics of social compensation design can help smart-home social media track and measure the degree of social compensation among older users. This understanding is key to grasping the emotional connection between older users and smart-home social media, comprehending their needs, expectations, preferences, willingness, and behavioral patterns. Consequently, this leads to the design of social applications that align more closely with the psychological traits of older users, the adoption of appropriate marketing strategies, and the provision of strong physiological and emotional fulfillment for older adults, thus stimulating and maintaining their willingness and behavior to use smart-home social media.
Research limitations
Like most studies, this research has limitations. First, the four SCDS dimensions were conceptualized and validated from an information-systems design perspective; we did not examine other potential determinants of social compensation (e.g., cultural heterogeneity among older adults, differences in digital literacy). Future work should incorporate older adults’ perspectives to identify additional determinants from complementary viewpoints. Second, participants were recruited online from an older-adult panel in China (Tencent Questionnaire); the absence of offline and rural samples may limit generalizability. Given that the SCDS is newly developed, future validation across offline contexts, rural populations, and diverse cultural settings is needed to strengthen generalizability.
Supplemental Material
sj-xlsx-1-dhj-10.1177_20552076251411270 - Supplemental material for Development and validation of social compensation design scale for urban older users in the context of smart-home social media
Supplemental material, sj-xlsx-1-dhj-10.1177_20552076251411270 for Development and validation of social compensation design scale for urban older users in the context of smart-home social media by Ke Ma, Ying Zhao, Francesco Ermanno Guida, Meng Gao, Renke He and Jinjun Xia in DIGITAL HEALTH
Supplemental Material
sj-xlsx-2-dhj-10.1177_20552076251411270 - Supplemental material for Development and validation of social compensation design scale for urban older users in the context of smart-home social media
Supplemental material, sj-xlsx-2-dhj-10.1177_20552076251411270 for Development and validation of social compensation design scale for urban older users in the context of smart-home social media by Ke Ma, Ying Zhao, Francesco Ermanno Guida, Meng Gao, Renke He and Jinjun Xia in DIGITAL HEALTH
Footnotes
Acknowledgments
ORCID iDs
Ethical considerations
Informed consent
Contributorship
Funding
Declaration of conflicting interests
Data availability statement
Guarantor
Supplemental material
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
