User-centric Smart City Services for People with Disabilities and the Elderly: A UN SDG Framework Approach

2.1 The Role of Sustainability in Shaping Smart Cities

The Brundtland Report of 1987, “Our Common Future,” authored by the World Commission on Environment and Development, provides a definition of sustainability, which states: sustainable development is that which satisfies the needs of the current generation without jeopardising the ability of future generations to meet their own needs. Within the context of urban development, sustainability emerges as a crucial paradigm, advocating for the balanced progression of economic vitality, social equity, and environmental stewardship. This concept is further articulated through the framework of the “3Ps” – prosperity, people, and planet – proposed by the OECD (2013) and is expanded by the inclusion of two additional “Ps”, peace and partnerships, underscoring the comprehensive nature of sustainability.

Smart city technologies are increasingly being viewed through the lens of sustainability, focusing not just on system efficiency but also on the social dimensions of urban environments. Caragliu et al. (2011) define a smart city as one that prioritises citizen well-being and development, combining traditional and modern infrastructure to foster sustainable economic growth and enhanced quality of life. This approach emphasises investment in people and wise management of natural resources, integrating public administration participation.

Focusing solely on technological usability can complicate the pursuit of sustainability, necessitating a balanced integration of technology and social needs. This integration requires informed and educated policy-making (Grace et al., 2023), community involvement in smart city development, and innovative management approaches. Pira (2021) suggests that sustainability in smart cities can be viewed either broadly across various sectors or through specific areas such as health, energy, education, and transport (Carafoli et al., 2016; Frez et al., 2019; Gimpel et al., 2021; Rahman et al., 2021). To realise sustainable smart cities, there must be continuous interaction between residents and their urban environment, with information processing (Ko et al., 2018) adding on-going value (Zygiaris, 2013, p. 2).

The most comprehensive approach to understanding smart cities incorporates the principle of sustainability (Bibri & Krogstie, 2021). The author argues that his approach utilises innovative data-driven technologies from smart cities to optimise, enhance, and maintain the performance of sustainable cities. To realise this integration, significant institutional transformations are needed, encompassing new practices and competences.

To address the gap identified in the literature, the subsequent sections will analyse smart city services from the end users’ perspective, prioritising the social aspect over technology. This approach aligns smart cities with sustainability, focusing on the impact on end users as a key driver for a sustainable future.

2.2 The Convergence of Universal Design and Assistive Technology in Smart Urban Environments

The term “consumer normalcy” represents an individual’s engagement in the marketplace across four dimensions: presence, distinction, competence and control, and equality (Baker et al., 2002). The consumer experience of PwD can be viewed through two lenses: the medical model, which concentrates on the individual’s deficiencies, and the social model, which mandates institutional responsibility in ensuring equal service provision through design (Baker et al., 2001).

This article advocates for the social model within the context of smart sustainable cities. The preference for the social model is rooted in the recognition that PwD often belongs to lower socioeconomic statuses and may lack the resources to afford assistive technologies independently.

The broadened definition of disability, for the context of this article, encompasses any condition that diminishes an individual’s quality of life and development, whether temporarily or permanently (Institute of Medicine, 2007). Furthermore, the National Academies Press discusses the impact of impairments on an individual’s ability to engage in life situations. Functional limitations, whether at the physiologic or anatomic level, are considered measures of the net impact of an impairment or impairments on an individual’s capacity. This view recognises that impairments, regardless of their duration, can significantly affect quality of life and daily functioning. The scope of inclusive design now extends to include women experiencing high-risk pregnancies and individuals with temporary injuries, reflecting the growing work on the inclusion of diverse populations.

Universal design, synonymous with inclusive design or design for all, emerged from the need to support the elderly in living independently into later life. This approach has since broadened to encompass the needs of PwD, recognising the overlapping requirements of these demographic groups (Design Council, 2008).

Universal design in the planning of spaces and facilities accessible to all plays a crucial role in enhancing the quality of life for PwD, reducing injury risk and associated costs of rehabilitation, healthcare services, and institutionalisation. Implementing universal design principles enhances public facilities and services, potentially boosting tourism in areas where such inclusive practices are evident.

This research article, in the following section, will analyse smart services through the lens of universal design, particularly focusing on adaptability for PwD and the elderly, who may experience sensory and motor challenges that technologies aim to compensate for. To help promote inclusive smart city designs, we are presenting Figure 1a set of recommendations for universal design, drawing from the guidelines established by the Government of Ireland. These recommendations are substantiated by real-world examples that demonstrate how universal design principles have been effectively implemented in various settings to enhance accessibility and inclusivity.

Figure 1

Recommendations for Universal design based on the Government of Ireland enhanced by real-life examples. (a) Intuitive spatial navigation and the accessibility of outdoor areas – Memorial to the Murdered Jews of Europe in Berlin, (b) and (c) The main entrance of facilities should be universally accessible, such as The Imperial College London, allowing all individuals to enter the building, d) Depicts a driver assisting an individual with a motor disability in boarding public transport, exemplifying the crucial role that trained personnel play in facilitating accessibility and mobility for all citizens. Berlin subway accessibility presented in (e) Information accessible in various formats and emergency exits accommodating everyone, (f) Subway accessible to citizens who use wheelchairs, (g) Demonstrates how Berlin’s subway system communicates accessibility information effectively, where wheelchair-friendly stations are indicated through visual aids; the use of large fonts and contrasting colours facilitates easy navigation. (h) There are visual cues for deaf individuals, alongside audio notifications that assist blind passengers by informing them of their current location throughout their journey (authors’ own resources).

2.3 Multidimensional Sustainability – The Core of Inclusive Smart City Ecosystem

Based on the current state of the art in literature and our own work, we are suggesting the design of a holistic, multidisciplinary smart city that necessitates a breadth of knowledge across various domains (Zhao et al., 2021). Inclusive planning requires the involvement of all stakeholders, including designers, planners, community members, municipal managers, and residents. Such inclusivity ensures equality in decision-making, fostering a shared vision and respect for collective goals. It is particularly important to integrate PwD and the elderly as active participants in the problem-solving and planning processes to ensure the services meet their specific requirements.

Smart city design emphasises accessibility for all, with effective governance ensuring progress in education, health, and transportation sectors (Carafoli et al., 2016; Frez et al., 2019). The smart city value chain revolves around the gathering, processing, and utilisation of information by both public administrations and individual users to enhance service efficiency and daily living. The smart city ecosystem, supported by diverse stakeholders and public–private–people partnerships (Gao et al., 2023), should be driven by social entrepreneurship and social marketing that originates from the needs of the end customer.

Services within sustainable smart cities should adhere to the cradle-to-cradle principle of resource utilisation. For example, residents could donate old smartphones to city administrations for redistribution to those in need, potentially receiving incentives such as public transport discounts in return.

Sustainability within private companies and social enterprises is most effective when it aligns with organisational strategy and culture, supported by leadership and a forward-thinking ethos. Training for all employees in both public and private sectors on sustainability principles ensures the organisation’s commitment to sustainable practices, enhancing the company’s reputation.

Universal design and user-friendly products add commercial value for manufacturers, increasing user satisfaction and brand loyalty. The architecture of smart devices and their adaptability to the needs of PwD is critical (Baker & Moon, 2008). Manufacturers and lawmakers play key roles in ensuring that devices are accessible, catering to the growing population of PwD and the elderly. Political directives can motivate businesses to address the needs of these demographics.

Key components of an inclusive sustainable smart city are outlined in Figure 2. Each element within the diagram – Smart people, Social marketing, Inclusive smart city services, Social entrepreneurship, Smart government, Universities, Non-governmental organisations (NGOs), corporate social responsibility (CSR), and foundations such as Sustainable development goals (SDG) goals and Triple bottom line reporting – plays a vital role in the functioning and development of a smart city. Following are the definitions for each suggested component (Table 1).

Figure 2

Sustainable inclusive smart city ecosystem.

Elaborating more on a conceptual framework for an inclusive smart city, we are highlighting how various elements influence each other and work together to create a sustainable, smart urban environment.

Smart People through their education, creativity, and digital competence drive the demand for inclusive services and influence the nature and effectiveness of social marketing efforts. They are both the beneficiaries and participants in the city’s smart governance and services. Social Marketing is a tool that can significantly impact Smart People by influencing their behaviour and encouraging their participation in sustainable practices. Inclusive Smart City Services are the result of cross-sector collaboration involving Smart Government, Universities, NGOs, and initiatives driven by Social Entrepreneurship.

Social Entrepreneurship initiatives often fill gaps in city services, innovating in areas where traditional corporation solutions may fail to address specific social needs. Smart Government can leverage partnerships with Universities and NGOs to drive innovation and inclusivity in service provision.

Universities contribute to this ecosystem by researching and developing new technologies and methodologies, training Smart people, and fostering a culture of innovation. NGOs often advocate for underserved populations and can be instrumental in ensuring smart city services are truly inclusive. CSR initiatives can support the goals of an inclusive smart city by funding innovations, providing services, or through advocacy and partnership efforts. SDG and Triple Bottom Line Reporting provide a strategic foundation and an evaluative framework for all the other elements. They ensure that the smart city’s development aligns with broader global sustainability objectives.

When addressing sustainability in smart cities, we suggest that the United Nations SDG should be used, with appropriate indicators that would be used as a measure to implement sustainability of the city. The United Nations (2022) is providing several targets and indicators under United Nations Sustainable Development Goal 11, which is focused on making cities and human settlements inclusive, safe, resilient, and sustainable. These targets address a range of issues from ensuring access to adequate housing and basic services, developing safe and sustainable transport systems, promoting inclusive urbanisation, safeguarding cultural and natural heritage, reducing disaster impacts, improving environmental conditions in urban areas, to providing access to safe and inclusive green public spaces. Each target is accompanied by specific indicators to measure progress towards achieving these objectives by the year 2030.

It becomes clear that the advantages of implementing smart inclusive services extend beyond mere accessibility and that the broader social implications are profound. As individuals, particularly those with disabilities and the elderly, become proficient in navigating one digital solution, they inherently equip themselves to overcome further digital challenges. This mastery can significantly boost self-confidence, igniting a drive towards overcoming additional barriers and achieving complete societal integration.

3.1 Research Methodology

In this research, a variety of general and specific scientific methods and techniques were employed to ensure the acquisition of dependable results. The early phase involved utilising analytical methods to review existing literature and evaluate current smart city solutions. The descriptive method provided a detailed exposition of the challenges related to the inclusion of PwD and the potential solutions offered by smart city services. Inductive reasoning underpinned the formulation and validation of hypotheses.

The results are instrumental in validating the central hypothesis that posits an increase in inclusivity and ease of daily activities for PwD using smart city services. This methodological approach was selected to ensure a comprehensive study of the research process across all stages – problem identification, research design planning, data collection and analysis, and conclusion formulation – abiding by the fundamental tenets of scientific inquiry.

3.2 Data Collection Process

In the context of research concerning the Balkan region, particularly for the nations of Serbia, Bosnia and Herzegovina, and Montenegro, it is often considered acceptable to treat them as a singular population for analytical purposes. This is due to the relatively homogenous cultural characteristics that these countries share, stemming from their historical, linguistic, and social ties. As of the latest available statistics, Serbia has a population of approximately 7 million (Statistical Office of the Republic of Serbia, 2022), Bosnia and Herzegovina around 3.5 million (Agency for Statistics of Bosnia and Herzegovina, 2022), and Montenegro close to 620,000 (Statistical Office of Montenegro, 2022). Within these populations, it is estimated that around 8–10% of people are living with disabilities, and the elderly (aged 65 and over) constitute approximately 17–20% of the population. OECD (2022) The analysis benefits from the cultural congruence across these populations, allowing for a more streamlined approach to data interpretation and policy recommendation.

Participant access was facilitated through an online platform provided by Aparteko, a company with a user base of approximately 8 million across the Balkan region, as of 2011 (Aparteko, 2011). Aparteko is renowned for its “Slagalica” quiz, a highly popular family-oriented quiz show in the region, initially broadcast on TV RTS1 in 1993 (RTS, 2019). This database was specifically selected due to its diverse demographic coverage in terms of age, gender, and other variables, thereby meeting the requirements for a varied survey sample. The questionnaire was integrated into the user experience and presented to participants prior to engaging in gameplay across multiple platforms, including Facebook, as well as Apple and Android mobile applications. The survey was conducted from 1–20 August, 2022, during which time a mixed, representative sample segment was successfully compiled.

The study employed an anonymous online questionnaire consisting of 26 questions (Table 2). These included four demographic questions, 21 items evaluating the participants’ readiness to adopt inclusive smart city services, and one open-ended question allowing respondents to voice additional concerns or comments on the given topic. A total of 302 individuals from Serbia, Bosnia and Herzegovina, and Montenegro participated in the study. Comprehensive demographic details of the participants are provided in Table 3.

1. Informed consent: Informed consent was obtained from all subjects involved in the study.

3.3 Statistical Analysis

To evaluate the proposed conceptual model, the authors opted for the SEM analysis. SEM is a statistical multivariate analysis developed to explore and evaluate causal relationships (Kline, 2005). The SEM analysis stands on the principles of two statistical analyses: Principal component analysis (PCA) and multiple linear regression. The PCA allows the grouping of measured variables, in our case items or questions, into latent variables or constructs. Latent variables are variables which are newly formed as a linear combination of the measured variables which make them. Multiple linear regression allows the evaluation of the causal relationships between latent variables or between latent and measured variables (Asparouhov & Muthén, 2009).

When it comes to the SEM analysis, authors can choose between two methods: covariance-based SEM (CB-SEM) and partial least squares SEM (PLS-SEM). The CB-SEM was the first SEM method proposed and is observed as a parametric SEM. Being a parametric SEM analysis, there are multiple assumptions that must be met to conduct it, such as Normality of the data, sample size, and number of items per construct (Jöreskog & Sörbom, 1984). On the other hand, the PLS-SEM is seen as a non-parametric approach to SEM, which has loosened preconditions. Also, there is a slight difference in their goals: CB-SEM is primarily used to confirm theories, while the PLS-SEM represents a causal–predictive approach to SEM that emphasises prediction in estimating models (Hair et al., 2021).

The application of SEM analysis, both CB-SEM and PLS-SEM is wide and ranges from application in the field of marketing to human resource management (Jovanović et al., 2022) and assessment of organisation performance (Farrell & Mavondo, 2005). In the beginning, most of the studies in which the SEM was applied were related to the CB-SEM. However, a recent bibliometric study shows that PLS-SEM use has gained momentum relative to factor-based SEM in recent years (Ciavolino et al., 2022). Due to the above-mentioned and the wide range of software available for its implementation, PLS-SEM analysis has been increasingly applied to verify conceptual models in different fields of study (Hair et al., 2014). To verify the conceptual model proposed in this study, the authors used PLS-SEM and the SmartPLS Software version 4 (Smart PLS GmbH, 2023).

To evaluate the statistical significance of the paths between the constructs, we performed bootstrapping on 5,000 samples. Bootstrapping is a standard procedure in the application of PLS-SEM analysis. The aim of this approach is to randomly create samples from the overall sample and to apply the PLS-SEM to each of the created samples (Cheung & Lau, 2008). The obtained results are used to generate an empirical sampling distribution for each model parameter and to test its statistical significance (Hair et al., 2018).

To compare the models between different groups, we used the PLS Multigroup analysis (PLS-MGA). This approach is explained in detail in the works of Sarstedt et al. (2011) as well as Hair et al. (2018). Basically, a PLS-SEM model is created for each of the observed groups and different statistical tests are used to determine whether there are statistically significant differences between the calculated path coefficients. It is important to note that these tests do not compare the overall models, whereas they compare only the values of the path coefficients.

Regarding the field of the study, several conceptual models have been proposed and verified using SEM analysis. Pal et al. (2018) wanted to observe the end user perspective on the IoT (Abadía et al., 2022) and smart homes for elderly care. The authors created a complex nine-construct model which they validated using PLS-SEM and showed that the extended UTAUT model is valid in the Asian context. Maswadi et al. (2022) created a model of factors influencing the elderly’s behavioural intention to use smart home technologies and tested it in Saudi Arabia. The results of the SEM analysis pointed out that with slight modification the model can be successfully implemented. Also, in a recent study, Dadhich et al. (2022) created a structural model to determine the purpose of using IoT gadgets in the clinical background and tested its validity using PLS-SEM. The same authors also noted that the PLS-SEM has been used with success in IoT healthcare research and provided valuable insights which can lead to further developments of these devices.

Led by these examples of good practice, as well as the fact that nonparametric SEM techniques are appropriate for multigroup analysis (Henseler et al., 2016), we chose the SEM analysis, its PLS variant, and the Smart PLS software to verify the proposed conceptual model.

4.1 Sample Characteristics

The descriptive analysis of the sample shows that 302 participants took part in the survey. Half of the sample is females (50.66%), while males make up 47.68% of the respondents. The remaining 1.66% of the respondents declared their gender as Other. The age of the respondents was measured in five groups. Most of the sample (28.28%) are respondents older than 64 followed by those who are between 45 and 64 years of age (23.84%) and those who are between 25 and 34 years of age (20.86%). The remaining two groups are almost equally represented. When it comes to educational attainment, the respondents were grouped into two classes: those with lower and middle education (61.26%) and those with higher education (38.74%). For this variable, we opted for level collapsing. The idea of level collapsing is to fuse the levels of the categorical variables presuming their effects to be equal (Gertheiss & Tutz, 2010). Herein, we collapsed the categories of primary education and secondary education. The last characteristic of the sample presented here was the type of disability the respondent has or has not any type of disability. Most of the respondents can be classified as those who do not have a disability (41.72%), followed by those with motoric disability (31.46%). A smaller percentage of the respondents has other disabilities (17.89%) and sensory disability (8.93%). The sample is presented in Table 3.

The three questions within the construct Hedonic motivation were measured as the number of information the respondent wants to get in a public place when using the internet, the number of activities the respondent uses the Internet, and priorities in the respondents’ perspective regarding the introduction of innovation for the disabled. To provide in-depth analysis and insights, in the next paragraphs, we present their answers in more detail.

When it comes to question HM1 and the information, the respondent wants to get in a public place when using the internet, 302 respondents provided 779 answers. On average, 2.57 answers per respondent. The obtained frequencies are presented in Figure 4. The most chosen information was the information regarding maps and navigation (22.5% of all answers), followed by public transport (19.9% of all answers), and weather (17.2% of all answers). The results also indicate that the respondents were the least interested in the information regarding air pollution, just 6.4% of all answers.

Figure 4

The information the respondents are interested in obtaining when in a public place, using the Internet.

Respondents provided interesting insights regarding what they use the internet for (HM2); 897 answers were collected indicating that on average, the respondents used the Internet for 2.97 activities. The obtained frequencies are presented in Figure 5. The most chosen activity was sending and receiving messages (23.4% of all answers), followed by finding navigation (22.7% of all answers). The results also indicate that the rest of the offered activities have been chosen in a significant percentage. The activity which was chosen the least number of times was fun (music, games, and video), 11.7%.

Figure 5

The activities respondents use the Internet for.

It was valuable to collect information regarding the respondents’ perspective on what the local communities and governments should focus on regarding the introduction of innovation for the disabled. In total, 1,024 answers were collected, indicating that on average the respondent had 3.39 priorities marked. The innovation which was most frequently marked was the security and data security of inclusive smart services (19.9%), followed by transport suitable for people with temporary or permanent disabilities/Elderly people and information about transport (16.3%), and access to healthcare facilities (14.6%). The frequencies are presented in Figure 6.

Figure 6

The priorities of the respondents’ perspective regarding the introduction of innovation for those with disabilities.

4.2 Construct Validity Tests

One of the prerequisites when conducting SEM analysis is to observe the internal consistency and discriminant validity of the defined constructs. As well, it is valuable to inspect the possible issue of multicollinearity in the model. In the following paragraphs, we present the results of the conducted tests.

The most used metric of scale reliability is Cronbach’s alpha (Cronbach, 1951). Average Variance Extracted (AVE), and Dijkstra-Henseler’s rho A (Dijkstra & Henseler, 2015) indices take values from 0 to 1, where the closer they are to 1, the better the internal consistency is, thus indicating that the scale is reliable. The widely accepted level of acceptance of Cronbach’s alpha is above 0.7 (Gliem & Gliem, 2003), while for AVE and rho A, it is above 0.5 (Dijkstra & Henseler, 2015). The calculated indices of internal consistency per construct alongside the number of items per construct are given in Table 4. All the indices were calculated on the overall sample. Out of nine constructs, six have satisfactory scale reliability with Alpha, AVE, and rho A above or marginally close to the defined thresholds. However, three constructs draw attention to low construct reliability: EE, HM, and PV.

The measured reliability can be quite low for constructs with just two items. Namely, Cronbach’s alpha is sensitive to the number of items per scale, whereas a low number of items might lead to a low Cronbach’s alpha (Gliem & Gliem, 2003). Having in mind that EE and HM, have two and three items per scale, we can conditionally accept the reliability of EE and HM. However, the measured reliability of PV clearly indicates that changes have to be made to this construct. The change to the model that we introduced was the inclusion of the variables PV1 and PV2 individually.

The next analysis is related to the discriminant validity of the constructs. To do so, we used the Fornell–Larcker criterion which compares the amount of variance captured by the construct’s AVE and the shared variance with other constructs (Fornell & Larcker, 1981). The discriminant validity compares the correlations between each pair of constructs and the value of the square root of the AVE of each construct. In other words, the discriminant validity assessment has the goal of ensuring that a reflective construct has the strongest relationships with its own indicators (e.g. in comparison with any other construct) in the PLS path model.

The results of the discriminant validity analysis are given in Table 5. The values in the table which are bold represent the square roots of AVE, while all other values indicate the correlation among constructs. The Fornell–Larcker criterion is satisfied when the value on the diagonal is greater than the correlations in the same row and column. The presented results indicate that the discriminant validity is confirmed in the proposed model.

Relevant literature in the field suggests that multicollinearity can significantly impact the quality of the SEM model (Jagpal, 1982). Multicollinearity is a statistical phenomenon which occurs when there is a high level or correlation among the variables in the model. High level of correlation among the variables leads to biased estimates, high standard errors, and inflated values of the t-statistics. Therefore, the potential presence of multicollinearity should be inspected. To assess the presence of multicollinearity, we used the variance inflation factor (VIF). If the value of VIF is above 5, there is multicollinearity in the model. Otherwise, there is no multicollinearity in the model. The obtained VIF per item is provided in Table 6. It can be concluded that there is no multicollinearity in the observed model.

4.3 PLS-SEM Results

In the paragraphs that follow, we present the results of the PLS-MGA. We conducted four MGA, one per each categorical variable we previously defined. The explored and compared models are the models per gender category, per age category, per type of disability, and per educational attainment. The statistical procedure for all four models was the same and is as follows. Among the several MGA comparison tests, we opted for the Bootstrap MGA.