1. Introduction
Urbanization has reached an exceptional speed and scale. According to the Global Report on Human Settlements by UN-Habitat (2022), the energy consumption of cities is over two-thirds of the world’s energy and accounts for more than 75% of global CO
2 emissions as a result of economic growth and rising population [
1,
2,
3]. In addition, it is expected that global energy-related CO
2 emissions will increase from 71% in 2006 to 76% in 2030 [
4]. Moreover, about half of the anthropogenic CO
2 emissions released to the atmosphere since the industrial revolution have occurred in the last 40 years. Today, even if anthropogenic GHG emissions were reduced to zero, researchers believe that changes in the climate system and their potential impacts would remain as a result of the burning of fossil fuels such as coal, oil, natural gas, and land-use change such as deforestation, agriculture, etc. [
5,
6,
7,
8,
9,
10,
11,
12].
The first severe global energy crisis is currently affecting the world. The International Energy Agency (IEA) has stated that the world has never seen such an extreme energy crisis before (IEA, 2022 [
13]). Moreover, recent research has suggested that 31 million Europeans lived in energy poverty in 2021 due to the COVID-19 recovery and Russia’s invasion of Ukraine [
14]. In 2022, The European Commission announced its REPowerEU plan, which provides the primary steps to avoid Europe’s energy dependence [
15]. Among OECD countries, Turkey has one of the highest energy demand ratios due to the population growth and increasing energy consumption patterns on the end-user side. According to the energy statistics of the IEA, Turkey’s industry accounts for 36% of the nation’s total final energy consumption, followed by transport (27%), housing (20%), and services (17%) (IEA, 2020) [
4]. However, the 2023 Turkish Energy Policy aims to reduce energy dependency by up to 30% by 2023 [
16] by developing new policies and standards to regulate energy use, improving energy efficiency, and lowering energy consumption. The research conducted by the Ministry of Energy shows that the energy saving potential of Turkey is considerably high; 30% in the building sector, 20% in the industrial sector, and 15% in the transportation sector in 2020 [
17,
18].
Within this perspective, climate change, global warming, and greenhouse gas emissions are driven by human behavior and thus could be reduced via greener behavioral and lifestyle changes. In the same vein, the IPCC’s sixth assessment report concluded that global emissions can be reduced by 40–70% by 2050 and global energy demand can decrease as a result of behavioral change [
19]. In addition, the European Commission’s REPowerEU plan aims at rapidly reducing energy dependence by 2030 via three critical components: behavioral changes, diversified energy resources, and a clean energy transition [
15]. Therefore, changing individual behavior to reduce energy consumption and demand could be the most cost-effective strategy to reach the goal of a sustainable, affordable, equitable, and secure energy supply in Turkey [
20,
21,
22]. In light of this goal, there have been many attempts to rebuild cities and communities in the context of a huge transition from an agricultural and industrial economy to a knowledge-based economy (such as the wired city, informational city, virtual city, smart city, intelligent city, sharing city, etc.) [
23,
24,
25]. As one example, smart cities and communities initiatives are more than just a matter of putting new technologies into place; instead, they are an attempt to understand how people use technology to solve their problems in more innovative ways in the information age. Moreover, smart cities and communities are utilizing technology to empower citizens to take control of their lifestyles more productively and to encourage them to participate actively and to cooperate with all stakeholders [
26,
27,
28,
29,
30,
31]. So, the real smart city needs to enable the use of the Internet of Things (IoT), virtual reality (VR), artificial intelligence (AI), and augmented reality (AU) approaches to increase participatory planning. Most crucially, within the scope of the crowdsourcing IoT (Crowd-IoT) paradigm, the government needs to encourage its citizens to collaborate to reduce energy consumption and carbon emissions for a sustainable urban lifestyle [
32,
33,
34,
35].
According to the wide range of studies, behavioral changes have enormous environmental benefits, as much as ecological consciousness restrictions and practices. For example, changing transportation choices such as using a bicycle or public transportation instead of a private car has greater ecological impacts than car-sharing or higher parking fees. Moreover, changing buying behavior has more positive effects on the environment than using recycled products [
36,
37]. Additionally, related works show that there is huge potential to improve energy conservation in public areas, public transportation, and dwellings by making use of the IoT, meters, and sensors [
32,
33,
34,
35]. However, these actions are related to various behavioral antecedents [
36,
38]. As a crucial part of the conceptual framework of the study, the Theory of Planned behavior (TPB) has been used to explain and predict a variety of human behaviors from different disciplines of science, but it is rarely applied in the area of energy conservation behavior in the context of smart cities and communities [
39,
40,
41,
42,
43,
44,
45]. Therefore, one of the main aims of the study is to investigate the energy conservation behavior of individuals ‘
to minimize the negative impact of one’s actions on the natural and built world’, which has been conceptualized as pro-environmental behavior [
46].
In this context, the well-known guides of Ajzek (2006) and Francis et al. (2004) for operational models of the Theory of Planned Behavior (Icek Ajzen, 1985) have been adapted for this study regarding the energy conservation behavior in the light of previous studies in the pro-environmental behavior literature [
47,
48]. According to the Theory of Planned Behavior, although there is not always a positive correlation between behavioral intention and actual behavior, an individual’s intentions are the first precursor to performing a behavior. Moreover, as can be seen in
Figure 1, behavioral intention depends on three main variables: (1) attitudes toward the behavior, (2) subjective norms, and (3) perceived behavioral control. As the first variables of TPB, attitudes toward the behavior refer to the degree of a person’s favorable or unfavorable evaluation of the behavior. Attitudes are based on two components, which are ‘behavioral beliefs’ (beliefs about consequences of the behavior), e.g., ‘reducing energy consumption will increase saving money’, and ‘outcome evaluations’ (advantages and disadvantages judgments about of the outcome of the behavior), e.g., ‘decreasing contributing to the protection of the natural resources is … desirable/undesirable’. Secondly, subjective norms are determined by the perceived social pressure to perform or not perform a behavior. ‘Normative beliefs’ (the perceived behavioral expectations of other people), e.g., ‘I feel under pressure of social media to reduce my energy consumption’) and ‘Motivation to comply’ (positive or negative evaluations about each normative beliefs) are the two supportive components to measure the subjective norms dimension. Thirdly, perceived behavioral control reflects people’s beliefs that they are capable of performing the behavior. Additionally, it can be directly measured by evaluating the individual’s self-efficacy and beliefs regarding the behavior’s controllability. It has two indirect measures, which are control beliefs (individual’s beliefs about the presence or absence of facilitators or barriers to performing the behavior), e.g., ‘the decision to reduce my energy consumption is beyond my control’, and influence behavior (perceived power of control beliefs to perform a behavior), e.g., ‘I am confident that I could reduce my energy consumption if I wanted to’ [
47,
48,
49,
50,
51]. In addition, as can be seen in
Figure 1, socio-cultural, demographic, environmental, and personal factors might be influential on behavioral, normative, and control beliefs of individuals about to perform a target behavior.
From an interdisciplinary perspective (urban planning, cognitive science, and information and communication science), this paper would like to make a contribution to the effectiveness of feedback and intervention mechanisms on energy conservation behavior towards sustainable energy communities. In this context, the impacts of energy feedback mechanisms on energy consumption behavior will be examined in the neighborhoods of the Kadikoy District in Istanbul, Turkey in 2019 (This paper is part of an EU-ERANET Co-fund (smart city) consortium project titled ‘Community Data-Loops for energy-efficient urban lifestyles (CODALoop)’ and supported by the Scientific and Technological Research Council of Turkey (TUBITAK), 116K011) Among 39 districts in Istanbul, Kadikoy has been selected as a case study area because of its diversified socio-economic structure and the initiatives of the local authority, such as building regulations and recycling policies, that aim to reduce the district’s carbon footprint and energy use. Therefore, the following sections of this paper will: (i) describe the design and implementation of the methodology of the study, including the construction of the survey, the selection of the case study area volunteer groups, data collection, and feedback and intervention mechanisms; (ii) analyze the effect of feedback and interventions on the energy consumption behavior of 100 volunteers; and (iii) discuss the potential of feedback and intervention mechanisms to encourage energy conservation behavior for sustainable and energy-efficient communities in smart cities.
3. Results and Findings
A paired sample
t-test was conducted to evaluate the impact of the feedback and interventions on energy conservation behavior of 100 volunteers (
Table 9). In this context, the null hypothesis (
) is that the average difference in multidimensional variables of energy conservation behavior scores is 0 from t1 (intention:
M = 5.76,
SD = 1.30; attitude:
M = 6.21,
SD = 1.14, and indirect measurement
M = 5.24,
SD = 0.72; subjective norm:
M = 3.61,
SD = 1.13, and indirect measurement
M = 5.13,
SD = 0.99; perceived behavioral control:
M = 5.20,
SD = 1.26, and indirect measurement
M = 4.79,
SD = 0.73; neighborhood belonging:
M = 4.22,
SD = 1.44) to t2 (intention:
M = 6.07,
SD = 1.03; attitude
M = 6.40,
SD = 0.97, and indirect measurement
M = 5.09,
SD = 0.63; subjective norm:
M = 4.08,
SD = 1.21, and indirect measurement
M = 5.25,
SD = 0.96; perceived behavioral control:
M = 5.28,
SD = 1.22, and indirect measurement
M = 4.48,
SD = 0.51; neighborhood belonging:
M = 4.42,
SD = 1.43). However, prior to conducting the analysis, the assumption of normally distributed difference scores was examined. Then, the Shapiro–Wilk test was performed, showing no evidence of non-normality (
W(100) = 0.96, intention:
skewness = 0.75,
kurtosis = 1.20; attitude (direct measurement):
skewness = 0.20,
kurtosis = 0.87 and attitude (indirect measurement):
skewness = −0.18,
kurtosis = 0.29; subjective norm (direct measurement)
skewness = 0.16,
kurtosis = 0.59 and subjective norm (indirect measurement):
skewness = 0.26,
kurtosis = −0.05; perceived behavioral control (direct measurement):
skewness = 0.38,
kurtosis = 0.56 and perceived behavioral control (indirect measurement):
skewness = −0.30,
kurtosis = 0.97; neighborhood belonging:
skewness = 0.13,
kurtosis = 0.44). According to Hair et al. (2010) [
75], values for skewness or kurtosis less than ± 1.0 indicate that the skewness or kurtosis for the distribution can be considered normal. However, Tabachnick and Fidell (2013) [
76] conclude that skewness or kurtosis values between ±1.5 are, in many cases, also acceptable and can be considered normal. Based on these outcomes and after visual examination of the histogram and the
QQ plot, the assumption was considered satisfied, and a paired sample
t-test was considered appropriate in this case.
A paired sample t-test was conducted to see the effects of feedback and intervention on the energy conservation behavior of the 100 volunteers from before the feedback and intervention program to after the feedback and intervention program. The results indicate that the null hypothesis was rejected for the intention score (t(99) = −2.75, p = 0.00), the attitude (indirect measurements) score (t(99) = 2.29, p = 0.02), the subjective norm (direct measurements) score (t(99) = −4.17, p = 0.00), and the perceived behavioral control (indirect measurement) score (t(99) = 3.60, p = 0.00). Therefore, the energy conservation behavior scores of the volunteers after the feedback and interventions (intention: M = 6.07, SD = 1.03; subjective norm: direct measurement M = 4.08, SD = 1.21; perceived behavioral control: indirect measurement M = 4.48, SD = 0.51) were statistically significantly higher than the energy conservation behavior scores of the volunteers before the feedback and interventions (intention: M = 5.76, SD = 1.30; subjective norm: M = 3.61, SD = 1.13; perceived behavioral control: indirect measurement M = 4.48, SD = 0.51). Moreover, the attitude (indirect measurement) variable of the energy conservation behavior scores of the volunteers after the feedback and interventions (M = 5.09, SD = 0.63) were statistically significantly lower than the attitude (indirect measurement) variable of the energy conservation behavior scores before the feedback and interventions (M = 5.24, SD = 0.72). Consequently, there is enough evidence to support the claim that the feedback and interventions affected the energy conservation behavioral scores of the volunteers in the following dimensions: intention, attitude (indirect measurement: behavioral beliefs and outcome evaluations), subjective norm (direct measurement), and perceived behavioral control (indirect measurement: control beliefs and influence behavior).
Interestingly, other results indicate that the null hypothesis failed to reject for the following psychological variables of energy conservation behavior: attitude (direct measurements) score:
t(99) = −1.69,
p = 0.09; subjective norm (indirect measurements) score:
t(99) = −1.17,
p = 0.24; perceived behavioral control (direct measurement) score:
t(99) = −0.67,
p = 0.50; and neighborhood belonging scores:
t(99) = −1.45,
p = 0.14. So, there was not a significant difference in the energy conservation behavior scores of volunteers after the feedback and interventions (attitude (direct measurements) score:
M = 6.40,
SD = 0.97; subjective norm (indirect measurements) score:
M = 5.25,
SD = 0.96; perceived behavioral control (direct measurement) score:
M = 5.28,
SD = 0.51; and neighborhood belonging score:
M = 4.42,
SD = 1.43) and before the feedback and interventions (attitude (direct measurements) score:
M = 6.21,
SD = 1.14; subjective norm (indirect measurements) score:
M = 5.13,
SD = 0.99; perceived behavioral control score (direct measurement):
M = 5.20,
SD = 0.12; and neighborhood belonging:
M = 4.22,
SD = 1.44). Correspondingly, as can be seen in
Table 10, there is not enough evidence to support the claim that there would be an effect of the feedback and interventions on 100 volunteers’ energy conservation behavioral scores for the following dimensions: attitude (direct measurements), subjective norm (indirect measurements), perceived behavioral control (direct measurement), and neighborhood belonging scores. In addition, a graphical representation of the means and adjusted 95 % confidence intervals (CI) is displayed in
Table 10.
There is strong evidence that the feedback and interventions program improves behavioral intentions (
t(99) = −2.75,
p = 0.00), attitudes (behavioral beliefs and outcome evaluations) (
t(99) = 2.29,
p = 0.02), subjective norms (
t(99) = −4.07,
p = 0.00), perceived behavioral control (control beliefs and influence behavior) (
t(99) = 3.60,
p = 0.00) variables of energy conservation behavior (
Table 10). Alternatively, this can be described as an effect size given by the absolute value of the difference in means (behavioral intentions (
M = −0.30); attitudes (behavioral beliefs and outcome evaluations) (
M = 0.15); subjective norms (
M = −0.47); perceived behavioral control (control beliefs and influence behavior) (
M = 0.30)) divided by the standard deviation (behavioral intentions (
SD = 1.10; attitudes (behavioral beliefs and outcome evaluations) (
SD = 0.68); subjective norms (
SD = 1.15); perceived behavioral control (control beliefs and influence behavior) (
SD = 0.85)), which is approximately 0.27 (this is classified as a ‘small’ effect size) for behavioral intention, 0.22 (classified as a ‘small’ effect size) for attitudes (behavioral beliefs and outcome evaluations), 0.40 (‘medium’ effect size) for subjective norms, and 0.35 (‘medium’ effect size) for perceived behavioral control (control beliefs and influence behavior), as shown in
Table 11 [
77].
4. Discussion
It is expected that the global energy demand will continue to rise by 2050 while natural resources will dramatically decrease every day. However, the goals of decarbonization, net-zero emissions targets, climate emergency calls (1.5 °C global warming limit), smart environmental transformation, and energy transition efforts bring hope for fundamental changes in climate action all around the world, as can bee seen in the objectives of the Paris Agreement and the Kyoto Protocol. All of these efforts have proven that ‘change’ is not going to happen overnight, as it will take time, effort, and bottom-up collective actions to reach a green energy transition for all.
Among OECD countries, Turkey has had the fastest-growing energy demand ratio over the last twenty years. Thus, the current Turkish energy policies aim to reduce energy dependency and energy consumption through increased energy efficiency and energy conservation, as a smart energy transition not only involves low-carbon technologies, decentralized energy systems, infrastructure, policies, and standards, but also improving energy efficiency, adopting energy-saving techniques, changing consumption patterns in households, and urban mobility choices in a sustainable urban energy system [
2,
78]. According to the study conducted by the Ministry of Energy, it has been found that the energy-saving potential is considerable at 30% in the building sector, 20% in the industrial sector, and 15% in the transportation sector [
17,
18,
79]. The findings indicate that encouraging energy conservation based on behavioral measures at an individual level could be a key strategy in the energy transition.
This study aims to understand the multidimensional dynamics of energy conservation behavior through feedback and interventions mechanism. In this context, the impacts of energy feedback mechanisms on energy conservation behavior are examined in the neighborhoods of the Kadikoy District in Istanbul, Turkey. Among the 39 districts in Istanbul, Kadikoy has been selected as a case study area because of its diversified socio-economic structure and due to the local authority’s initiatives, such as building regulations and recycling policies, aimed at reducing the district’s carbon footprint and energy use. According to the selection criteria of the study, 100 residents volunteered to participate in the research for a period of eight months. It is important to mention that all of the volunteers received comparative, historical, and goal-setting feedback, as well as face-to-face interventions (as a part of the feedback and intervention plan of the study) during the research period. Since the study was limited to measuring the effects of the feedback and intervention on energy conservation behavior, it was not possible to measure which feedback and interventions were more effective for volunteers. Another uncontrolled factor is whether the volunteers received the online interventions, such as emails and social media posts, or read the energy reports or energy bulletin. In addition, participation in face-to-face activities was optional. Moreover, the 100 volunteers filled out the ’Energy-Saving Behavioral Questionnaire’, specifically constructed in order to measure changing patterns of energy conservation behavior, two times: at the beginning (t1) and at the end of the experiment (t2). By the end of the survey period, energy conservation behavioral data had been collected from 100 volunteers and centralized in a database. To test the hypothesis, the present study used the following tests: reliability, normality, and paired sample t-test (using IBM SPSS Statistics V.26.) to test the effect of the feedback and interventions program on the change in multidimensional variables of energy conservation behavior at t1 and t2.
When comparing the two results, it can be seen that the energy conservation behavior scores of volunteers after the feedback and interventions (intention: M = 6.07, SD = 1.03; subjective norm: direct measurement M = 4.08, SD = 1.21; perceived behavioral control: indirect measurement M = 4.48, SD = 0.51) are statistically significantly higher than the energy conservation behavior scores of volunteers before the feedback and interventions (intention: M = 5.76, SD = 1.30; subjective norm: direct measurement M = 3.61, SD = 1.13; perceived behavioral control: indirect measurement M = 4.48, SD = 0.51). Moreover; the attitude (indirect measurement) variable of the energy conservation behavior scores of volunteers after the feedback and interventions (M = 5.09, SD = 0.63) were statistically significantly lower than the attitude (indirect measurement) variable of the energy conservation behavior scores before the feedback and interventions (M = 5.24, SD = 0.72). In summary, these results show that as a result of the feedback and intervention, volunteers reflect a relatively high behavioral intention as a first precursor to perform a behavior, which consists of expectations, wants, and decisions, in favor of energy conservation. Moreover, the subjective norm reflects the perceived social pressure of the participant’s social network groups, such as family, friends, neighbors, or the government. After feedback and interventions, 100 participants perceived slightly high social pressure about reducing their energy consumption levels.
Additionally, the effects of the feedback and intervention on the perceived behavioral control (control beliefs and influence behavior) score can be summarized as participants feeling more in control and feeling likeable to achieve reducing their energy consumption within the eight months. Another consequence of the effect of the feedback and interventions program is that the attitude (behavioral beliefs and outcome evaluations) score of the participants reflects a weak to moderate positive attitude toward reducing their energy consumption level at the end of the experiment. It is important to highlight that the participant’s attitude scores are positive at both t1 and t2. This can be explained by the fact that a positive (+) score means, overall, the participant is in favor of reducing their energy consumption level. However, there are several possible explanations for why the direct measurement of the attitude score of the participants is lower after the feedback and interventions, such as feedback frequency, external factors in the experimentation process, complexity of the tasks, etc. [
66,
68,
72]. Further research should be undertaken to investigate when and how feedback about energy usage is more effective via a metadata analysis.
The empirical findings in this study provide a new understanding of empowering citizen participation in a smart city’s energy transition process through a data-driven feedback loop. The evidence from this study suggests that feedback and intervention mechanisms can boost or reduce an individual’s energy conservation behavior as the first level of the bottom-up energy transition approach in smart cities. Moreover, the study results can be used to develop targeted feedback and intervention mechanism aimed at dimensions of energy conservation behavior. In parallel with this, a special focus should be given to smart city applications, which are powerful tools for local governments to enable such a feedback and intervention mechanisms within the smart energy domain. However, considerably more work will need to be performed to determine the relationship between behavioral intention as the dependent variable and attitude, subjective norms, and perceived behavioral control as the predictor variables through a multiple regression procedure. Moreover, further research could clarify the link between actual energy consumption data and energy conservation behavior. This bottom-up energy transition approach will provide useful insights for the local government in empowering citizen participation and data-driven feedback loops.