Abstract
This study examines how flashback nudging—a reminder of past pro-environmental actions—delivered through a chatbot survey over 6 weeks in a quasi-experimental design, can facilitate pro-environmental behavior spillover among tourists visiting nature-based destinations. Integrating conversational artificial intelligence allows for personalized, timely, and scalable interventions, making the delivery of flashback nudges more efficient and engaging than traditional methods. Flashback nudging effectively strengthens the impact of environmental self-identity on behavioral change. Although nudges take time to work, the multigroup analysis revealed significant differences in self-reported past pro-environmental behaviors between the treatment and control groups. The findings demonstrate the importance of deliberate, technology-enabled interventions in enhancing meaning-making processes and promoting transformative tourism experiences. Artificial intelligence advances behavior change efforts by enabling innovative, lasting environmental engagement. This study emphasizes the potential of conversational artificial intelligence to drive tourism’s contributions to a sustainable future, advancing the discourse on leveraging technology for behavioral interventions in tourism.
Keywords
Introduction
A recent Tourism Panel on Climate Change report suggested that the tourism industry may miss its target of reducing emissions by 50% by 2030, as outlined in the Glasgow Declaration of Climate Action in Tourism (TPCC, 2023). This global environmental concern has lent further urgency to research efforts that seek to learn how tourism can derive positive pro-environmental behavior spillover effects. In general, “pro-environmental behaviors” refer to actions intentionally carried out to reduce environmental impact (Kollmuss & Agyeman, 2002), and “spillovers” occur when the performance of certain pro-environmental behaviors leads to the performance of the same or similar pro-environmental behaviors by individuals (Van der Werff et al., 2014). Previous studies have shown that tourism has the potential to induce long-term behavior change among tourists who visit pro-environmental destinations (Ballantyne, Packer, & Falk, 2011; Ballantyne, Packer, & Sutherland, 2011; Hehir et al., 2021; Hughes et al., 2011). However, more research is needed to improve its practical application and to replicate such findings across multiple contexts at a large scale.
There is a limited understanding of the mechanism driving pro-environmental behavior spillovers from tourism. Wu et al. (2021) reported that spillover effects were bounded by the same contexts and subject to the availability of infrastructure. Other studies observed that strong pro-environmental behavior intentions expressed on-site often did not materialize into actual pro-environmental behavior (Hughes, 2013; Wu et al., 2015). Other scholars, meanwhile, have advocated the importance of using post-visit action resources (Bueddefeld & van Winkle, 2017; Hughes et al., 2011), reflective engagement (Ballantyne, Packer, & Falk, 2011), and interpretive experiences (Ballantyne, Packer, & Sutherland, 2011) to encourage tourists’ adoption of long-term pro-environmental behavior. These studies show that self-reflection and thoughtful intervention are key to unlocking tourism’s potential to induce positive pro-environmental behavior spillovers.
For tourists who have engaged in pro-environmental behaviors during their visit to nature-based destinations, a reminder of their previous relevant behavior would serve as a thoughtful intervention that can trigger self-reflection. Indeed, a study by Van der Werff et al. (2014) showed that doing so strengthened individuals’ environmental self-identity, ultimately leading to their engagement in other pro-environmental behaviors—demonstrating a spillover effect. This thoughtful intervention is here described as “nudging,” defined as any attempt to change people’s behavior without significantly changing their economic incentives (Thaler & Sunstein, 2008). With recent advancements in artificial intelligence technology, such reminders can be delivered at scale through technology such as conversational artificial intelligence (Majid et al., 2023). Chatbots, a form of conversational artificial intelligence (Ling et al., 2023), have been explored in recent years for their ability to assist in changing human behavior, such as helping manage stress or encouraging more physical exercise (Zhang et al., 2020). Extending this line of inquiry, it is interesting to understand how chatbots can deliver nudges, facilitate self-reflection, and enable pro-environmental behavior spillovers.
Wu et al. (2021) posit that longitudinal research is the only viable way to study pro-environmental behavior spillovers, especially in tourism. Thus, this study set out a 6-week investigation into how chatbots can be used to deliver educative nudges to facilitate pro-environmental behavior spillover. The Gili Islands in Lombok, Indonesia, were selected as the setting for the field experiment because tourists on the islands are allowed to use only sustainable travel options, such as walking, cycling, e-biking, or hiring traditional horse carriages (González-Rodríguez & Tussyadiah, 2022). Majid, Tussyadiah, and Kim (2024) conceptualized an ideal scenario where tourists who return from the Gili Islands would continue to travel sustainably by using environmentally friendly transport more often in their daily lives after interacting with a chatbot developed to enable the desired spillover effects. Majid, Tussyadiah, Kim, and Chen (2024) later validated the proposed concept through a scenario-based experimental survey, finding that people’s intentions to use the proposed chatbot technology were strongly correlated with their intentions to use environmentally friendly transport more often in their daily lives.
This study aims to provide empirical evidence from real-life settings to the two aforementioned studies by comparing the behavior of two distinct groups over time. The treatment group consisted of tourists returning from the Gili Islands, whereas the control group contained participants from six different regions of Indonesia who had never been to the Gili Islands. This study investigates how pro-environmental behavior spillovers are enabled through flashback nudging and discusses how other tourist destinations may benefit from following similar approaches. Furthermore, given the recent technological advancements in generative conversational artificial intelligence (Shin et al., 2023), this study offers timely insights into how advanced technologies can be empowered to drive sustainability (Xiang et al., 2021). Additional theoretical contributions, managerial implications, research limitations, and future research directions are also discussed.
Literature Review and Hypothesis Development
Conceptualizing “Flashback Nudging” as Stimuli in the Stimuli-organism-response Framework
Nudging, as popularized by Thaler and Sunstein (2008), refers to subtle interventions that guide individuals toward desirable behaviors without restricting their choices. Traditional nudging strategies, such as framing, default options, and informational prompts, have proven effective at influencing immediate decision-making (Acquisti et al., 2017; Hagman et al., 2015), but these strategies often overlook the potential impact of past actions and their influence on future behaviors. This study introduces “flashback nudging,” a novel approach that leverages reminders of individuals’ past pro-environmental behaviors to strengthen environmental self-identity, thereby fostering behavioral consistency and spillover effects over time (Van der Werff et al., 2014). Unlike conventional nudging, flashback nudging focuses on connecting past behaviors with future intentions through reflective processes, thereby addressing a gap in the research on how interventions targeting past actions influence long-term identity reinforcement and behavioral spillover.
Flashback nudging draws on self-perception theory (Bem, 1972), which posits that individuals infer their own attitudes and identities by observing their past actions. By reminding individuals of their previous pro-environmental behaviors, flashback nudging reinforces their perception of being environmentally responsible, thus enhancing their environmental self-identity—a key driver of pro-environmental behavior spillover. This approach also aligns with the theory of planned behavior (Ajzen, 1991) in that it positively influences attitudes, subjective norms, and perceived behavioral control, thereby strengthening intentions and capabilities to engage in future pro-environmental behaviors. Furthermore, the stimuli-organism-response framework (Mehrabian & Russell, 1974) provides a theoretical foundation for understanding how reminders of past behaviors (stimuli) affect psychological states (organism) such as environmental guilt, concern, and self-identity, which in turn drive behavioral responses (Jacoby, 2002; Vieira, 2013).
The concept of flashback nudging is particularly relevant in the context of Kahneman’s (2011) dual-process theory of decision-making, which distinguishes between system 1 thinking (fast and intuitive) and system 2 thinking (slow and deliberative). Sunstein (2016) highlights the effectiveness of system 2 nudges, such as educational campaigns, in promoting environmental conservation. Flashback nudging acts as an educative nudge by combining reminders of past pro-environmental behaviors with educational information about the positive impacts of those behaviors. This approach aims to trigger self-reflection (Kirillova et al., 2017; Soulard et al., 2023), nurturing value orientations such as biospheric and altruistic values (Gatersleben, 2014; Schwartz, 2006; Stern et al., 1993). However, further longitudinal research is needed to examine how flashback nudging influences psychological processes and assess whether it leads to sustained spillover effects.
Organismic Factors
Environmental Guilt
In essence, guilt is an emotion that stems from a feeling of responsibility for harmful actions (Izard, 1977). When contextualized into pro-environmental behavior, environmental guilt is an emotional result of one’s inability to meet the behavioral ideal of what is perceived as “good” for the environment. Although past research used various terminologies, such as eco-guilt (Mallett, 2012), when referring to this construct, research has often found that environmental guilt is an important driver of pro-environmental behavior (Eom et al., 2021). For instance, United States residents who felt strong guilt about their country’s contribution to environmental degradation and climate problems showed a stronger willingness to behave more pro-environmentally (Mallett, 2012).
However, we have only limited understanding of how environmental guilt as a factor would behave in a longitudinal observation. Lacasse (2016) suggested that in a cross-sectional context, the feeling of guilt is alleviated by pro-environmental behavior. However, because pro-environmental behavior spillover is a process of adopting a new behavior in which the respective individuals often have not fully internalized as part of their habit, there may be occasions where the performance of pro-environmental behavior remains inconstant at the start (Gardner et al., 2012). As such, how environmental guilt may change over time and affect the adoption of pro-environmental behavior warrants observation in a longitudinal study.
Environmental Concern
A universal definition of environmental concern is people’s expressions of concern about environmental issues (Dunlap & Jones, 2002). Environmental concern is a widely studied concept often found to be a significant predictor of pro-environmental behavior, whether directly (Fujii, 2006) or indirectly via other constructs (Bamberg, 2003). Past research has also operationalized the environmental concern as a single construct (Fujii, 2006) or a measurement scale comprising multiple underlying factors (Cruz & Manata, 2020; Weigel & Weigel, 1978). In defining the terminology of environmental concern, some scholars also equated the concept with environmental attitudes (Cruz & Manata, 2020; Schultz, 2001).
Despite these differences, environmental concern has consistently received substantial consideration in research on pro-environmental behavior, with recent studies viewing the construct through longitudinal lenses (e.g., Qiao & Dowell, 2022). For instance, using cross-lagged models, McBride et al. (2021) found that environmental concern predicted a slight residual increase in psychological distress within the same individuals over time. Meanwhile, in a longitudinal analysis of a British cohort study from 1991 to 2008, Melis et al. (2014) found that the mean level of people’s environmental concern decreased over time. Their conclusion indicated the broader political divisions in the United Kingdom at the time, which were fueled by the financial crises. Thus, uncovering how environmental concern would behave over time in a tourism-context study on pro-environmental behavior spillover is important to enrich our understanding of the concept.
Environmental Self-identity
Environmental self-identity describes the extent to which a person identifies as someone whose actions are environmentally friendly (Van der Werff et al., 2014). Previous literature on pro-environmental behavior spillover has consistently found that environmental self-identity is the strongest predictor of the adoption of the spillover effects (Chaiken & Baldwin, 1981; Lacasse, 2016; Truelove et al., 2021; Van der Werff et al., 2013). This is because environmental self-identity, at its core, aligns well with the traditional self-perception theory, suggesting that people form their self-identities by observing the implications of their own behaviors (Bem, 1972). Therefore, positive spillovers were found when mechanisms to strengthen people’s environmental self-identity were introduced to induce pro-environmental behavior spillovers, such as through reminders of past behavior (Van der Werff et al., 2014) or labeling (Lacasse, 2016).
As tourists become increasingly environmentally conscious, the above mechanism can be highly applicable to a tourism context. However, to the best of our knowledge, this hypothesis has not yet been explored in previous research. Tourist destinations implementing pro-environmental regulations allow tourists, traditionally identified as hedonistic consumers of the industry (Dolnicar, 2020), to behave pro-environmentally (e.g., González-Rodríguez & Tussyadiah, 2022). Therefore, studying the interplay of the mechanism we refer to as “flashback nudging” in a longitudinal experiment will reveal how environmental self-identity as a factor will influence the potential pro-environmental behavior spillover.
Response: Pro-environmental Behavior Intention and Self-reported Past Behavior
Existing research on pro-environmental behavior spillover in the tourism context has investigated behavior intention and actual behavior to evaluate spillover effects (e.g., Wu et al., 2021). One challenge in researching pro-environmental behavior is social desirability bias, which likely affects behavior intention, rendering it a less reliable predictor of actual behavior and often creating the intention–behavior gap (e.g., Godin et al., 2005). As a solution, scholars have looked at both factors across different time points to draw conclusions (Dolnicar et al., 2024). However, in the absence of interventions, examining behaviors across different time points without accounting for the transition between home and holiday contexts that tourists experience (Wu et al., 2021) overlooks an important aspect. Conversely, when tourists exhibit behavior changes following specific interventions, this response suggests that such interventions may be effective, which aligns with past research indicating that individuals’ past actions are strong predictors of their future behavior (e.g., Bentler & Speckart, 1981).
To facilitate pro-environmental behavior spillover, previous literature argues that post-visit action resources are needed as an intentional intervention (e.g., Wu et al., 2015). When integrated into a longitudinal study that captures both behavior intention and actual behavior, the intervention allows researchers to evaluate any degree of behavior change that the intervention may cause. Dolnicar et al. (2024) recently found that such measures, known as “quasi-experimental research designs,” constitute only 1% of quantitative approaches applied in tourism and hospitality research hitherto. As such, this study considers both behavior intentions and self-reported past behaviors as the R components that will be investigated in the longitudinal experiment. Figure 1 presents the conceptual model developed using the stimuli-organism-response framework.
Conceptual model developed using the stimuli-organism-response framework.
Methods
Study Design
This study adopted a deductive approach to generate questionnaire items to design the proposed stimuli-organism-response model. After all constructs were identified, measurement items were derived from the literature. Three items for environmental guilt were adopted from Truelove et al. (2021). Three items for environmental concern were derived from Fujii (2006). Three items for environmental self-identity and one for pro-environmental behavior intention were borrowed from Van der Werff et al. (2013). Lastly, one item (i.e., how often did you use environmentally friendly transport in the past week?) was adapted from Chang and Krosnick (2003) to measure self-reported past pro-environmental behavior. Regarding the environmentally friendly transport options provided in the survey, this study considers four travel modes most relevant to the Indonesian context: (a) active travel such as walking and cycling, (b) public transport, (c) online ride-hailing services, and (d) electric vehicles. All items were measured on a five-point Likert scale (see Appendix Table A1). For self-reported past behavior, participants responded to the given question using the following scale: 1 = never, 2 = sometimes, 3 = about half the time, 4 = most of the time, and 5 = always.
This study employed a mixed within-subjects and between-subjects design to examine the effects of nudging over time (within-subjects) and across different participant groups (between-subjects). To achieve these objectives, two groups of survey participants were recruited, whose comparison is relevant to the between-subjects design. Meanwhile, the within-subjects design in this study has a time dimension where participants serve as their own controls, with their pre-intervention behavior serving as the baseline measurement. Instead of having a group of participants who were not exposed to the intervention (as in the case of between-subjects design), the within-subjects design in this study compares each individual’s behavior before the educative nudges to their behavior after the educative nudges. Thus, all the research variables, measured repeatedly over 6 weeks—including environmental guilt, environmental concern, environmental self-identity, pro-environmental behavior intention, and self-reported past behavior—were used to conduct within-subjects and between-subjects observations to uncover the underlying mechanisms of how nudges—educative and flashback—operate. The 6-week duration was based on previous research examining behavior change over similar periods (e.g., De Backer et al., 2018; Liang & Willemsen, 2022).
Sample and Recruitment
The following two groups were recruited to ensure that this study’s sample was theoretically justified (Charness et al., 2012). The first group consisted of local Indonesian tourists from the Gili Islands, recruited as they were about to leave the island of Gili Trawangan (referred to as the “Gili group”). The second group (the “non-Gili group”) consisted of Indonesians who had never visited the Gili Islands. These participants were residents of six major cities/regions in Indonesia: Bali, Bandung, Jakarta, Solo, Surabaya, and Yogyakarta. Recruitment for the non-Gili group was conducted through the professional networks of Indonesian academics, with 10 participants sourced from each of the 6 locations.
Each group initially included 60 participants, for a total of 120 participants. Participants were spread in ten different WhatsApp groups, each containing 10 to 15 participants. All participants were given monetary compensation upon completing the survey: 200,000 Indonesian rupiah for the Gili group and 100,000 Indonesian rupiah for the non-Gili group (about 13.00 and 6.50 United States dollars, respectively). The survey was conducted over seven weekends, beginning September 23 and ending November 6, 2023. Participants received a weekly survey link every Saturday morning, with a deadline for responses set at Sunday midnight. Late responses submitted by Monday morning were accepted, but those submitted past this extension were excluded from the analysis. Figure 2 illustrates the research design and the longitudinal timeline.
Research design and longitudinal timeline.
At the end of the survey, six participants from the Gili group dropped out, and two participants (one from each group) missed a few weeks of responses. Thus, 112 participants provided the complete responses that were used for further data analysis (see Table 1). The demographic profile of the participants, predominantly young adults, aligns with the typical visitor profile for the Gili Islands, as noted in previous research (Majid, Tussyadiah, & Kim, 2024; Majid, Tussyadiah, Kim, & Chen, 2024). However, future research should explore a more diverse demographic sample and employ additional controls to reduce selection bias and better assess the generalizability of the findings.
Descriptive Statistics.
Survey and Intervention Design
Educative nudges were developed into conversational formats for chatbot surveys (see Appendix Table A2). A chatbot survey was selected because it enables research participants to engage conversationally with the intervention messages while submitting their survey responses. Additionally, previous research has shown that chatbot surveys can yield higher-quality data than regular surveys (S. Kim et al., 2019). The educative nudges contained information on climate-related disasters and environmental issues gathered from academic and gray literature, including news sources and government press releases.
The chatbot survey consisted of two halves. The first half contained educative nudges providing information on climate-related issues, and the second half included questionnaire items. The Gili group received a few more lines (i.e., flashback nudges) reminding them of their past pro-environmental experiences during their visit to the Gili islands before the questionnaire items (see Appendix Table A3). This intervention design is the key aspect of this research, which may have contributed to differential effects between the groups. Participants took about 5 min to complete the survey. Appendix Figure A1 illustrates the human–chatbot conversation facilitated by the chatbot survey.
Translation and Expert Review
Because the survey targeted Indonesians, all materials were rigorously translated into Indonesian to ensure conceptual equivalence. Back-translation was also employed, wherein materials were first translated into Indonesian and then independently translated back into English by a bilingual academic familiar with the subject matter. The back-translation was compared with the original to identify and address any discrepancies in meaning. Furthermore, an expert review was conducted in August 2023 with eight Indonesian doctoral students and faculty members with expertise in social science subjects, such as tourism, psychology, and business management. Their feedback was used to refine and improve the survey presentation. Once finalized, the survey materials were imported into the Landbot survey platform. All collected data were securely hosted on Google Sheets.
Mapping Hypotheses to Methods
•
•
•
Justification for the Sample Sizes and Mitigation Strategies
Considering the relatively small sample sizes in this study—with 112 participants completing the longitudinal survey—several steps were taken to ensure the validity and robustness of the findings. First, partial least squares structural equation modeling (PLS-SEM) was employed because it is well suited for exploratory research and small sample sizes (Hair et al., 2019). PLS-SEM is particularly advantageous when the primary goal is theory development because it allows for the estimation of complex models while minimizing requirements for normal data distribution and large sample sizes (Hair et al., 2019). Furthermore, to mitigate any potential biases introduced by the limited sample size, the data collection process included rigorous checks for missing data, and a bootstrapping approach with 5,000 resamples was applied to generate robust standard errors and confidence intervals.
In addition to the methodological justification for PLS-SEM, efforts were made to enhance the generalizability of the findings. To this end, the study design utilized a longitudinal approach with seven waves of data collection to increase the reliability of observed patterns over time. This approach allowed for robust within-subjects and between-subjects comparisons, minimizing the risk of idiosyncratic results caused by small sample sizes. Additionally, the sample was purposefully diversified to include participants from varied regions of Indonesia, thereby ensuring a mix of environmental experiences and attitudes that reflect broader societal trends.
Finally, the decision to use smaller sample sizes was also informed by resource constraints inherent to longitudinal studies. The trade-offs between sample size and survey length were carefully considered to balance participant engagement and response quality. Although larger sample sizes would strengthen statistical power, the longitudinal nature of this research, combined with the advanced capabilities of PLS-SEM, provides a sufficient foundation for exploring the proposed relationships and testing the novel concept of flashback nudging. Future studies are encouraged to replicate this research using larger and more diverse samples to further validate these findings.
Results and Discussion
A series of statistical analyses were conducted to investigate the impact of educative nudges and flashback nudging. Specifically, a paired-sample t-test and Mann–Kendall non-parametric test were conducted to evaluate the effects of educative nudges to inform the within-subjects observation. Meanwhile, Hotelling’s T-squared test, independent samples t-test, longitudinal structural equation modeling, and multigroup analysis were conducted to analyze the impact of flashback nudging to inform the between-subjects observation. Descriptive statistics of all measurement items are provided in Appendix Table A4.
An a priori power analysis was conducted using G*Power version 3.1.9.7 (Faul et al., 2007) to determine the minimum sample size required to conduct the statistical tests. Results indicated that the required sample size to achieve 80% power for detecting a medium effect, at a significance criterion of α = .05, was N = 34 for the paired-sample t-test and N = 102 for the independent samples t-test (51 per group). For Hotelling’s T-squared test, a sample size of N = 86 (43 per group) was required to achieve 80% power for detecting a large effect at a significance criterion of α = .05. Thus, the obtained sample size of N = 112 with N = 53 for one group and N = 59 for the other is adequate to run the tests. Appendix Table A5 presents the results of measures of the sampling adequacy (MSA), showing that N = 53 for the Gili group (MSA = 0.618) and N = 59 for the non-Gili group (MSA = 0.662) are within the acceptable threshold (Hair et al., 2019).
Analysis of the Impact of Educative Nudges
A paired-sample t-test using SPSS was conducted to evaluate the impact of the educative nudges on the self-reported past pro-environmental behavior of the participants in both the Gili and non-Gili groups. The results showed significant increases in self-reported past pro-environmental behavior scores for both groups. For the Gili group, the self-reported past behavior scores
A Mann–Kendall non-parametric test using the trend package (Pohlert, 2023) in R, a statistical computing language, was conducted to assess whether there is an increasing or decreasing trend in the data values of each factor over time, as well as whether the trend is statistically significant in either direction. There were substantial increasing trends in the self-reported past behavior of both groups: Gili (S = 19, p < 0.01) and non-Gili (S = 16, p < .05). Significant increasing trends were also found in the behavior intention of both groups: Gili (S = 14, p < .05) and non-Gili (S = 12, p < .05). Another significant increasing trend was found in the environmental guilt of the non-Gili group (S = 15, p < .01), whereas the Gili group showed only a partially significant increasing trend (S = 11, p < .1). Insignificant increasing trends were shown in the environmental self-identity of both groups: Gili (S = 7, p > .1) and non-Gili (S = 5, p > 0.1). Meanwhile, the environmental concern of both groups showed insignificant decreasing trends: Gili (S = −4, p > .1) and non-Gili (S = -6, p > .1). Appendix Table A6 presents the results of the Mann–Kendall test.
These two tests demonstrate that the educative nudges shared with the two experimental groups effectively encouraged participants to adopt more pro-environmental behaviors. Results from the paired-sample t-test showed that educative nudges triggered immediate behavior change (see Appendix Table A7). This is evident from the significant differences between baseline self-reported past pro-environmental behavior and self-reported past pro-environmental behavior after participants received the nudges in the first week. The baseline mean level for the Gili group was higher than that of the non-Gili group, as the former had engaged in pro-environmental behavior the week before while still on the Gili Islands. This finding was further supported by the Mann–Kendall test, which showed significantly increasing trends in both behavior intention and self-reported pro-environmental behavior for both groups.
When investigated more closely, the significance values of these two parameters from the Gili group were higher than those of the non-Gili group, indicating that the given nudges worked more effectively for the Gili group. This difference can be further explained by the more significant increase in environmental guilt among the non-Gili group. According to past research studying environmental guilt as the driver of pro-environmental behavior (e.g., Eom et al., 2021; Mallett, 2012), when people perceive that they can do more for the environment but still fall short of meeting their own ideals, they will feel guilty. Weaker feelings of guilt were found among the Gili participants because they increased their self-reported past behavior more than their counterparts.
Meanwhile, the difference in the direction of the trends between environmental self-identity (positive) and environmental concern (negative), albeit insignificant, is another interesting finding from this study. As people start embracing more pro-environmental behavior, their expressed concern is converted into real actions that increase their self-perception as someone who is environmentally friendly. This finding may help explain the decline in environmental concerns found in the longitudinal study by Melis et al. (2014), suggesting that a decrease in environmental concerns could also represent the increased importance of climate challenges as people start to take more concrete pro-environmental actions. However, it should be noted that the Gili group’s results are also influenced by the combined effects of both educative and flashback nudges starting in Week 1, rather than educative nudges alone. Therefore, the significant results observed in the non-Gili group alone can already demonstrate the effectiveness of educative nudges.
Analysis of the Impact of Flashback Nudging
Hotelling’s T-squared test in SPSS was conducted to assess the difference in data values of each measured factor between the Gili and non-Gili groups. The results showed a significant difference between the self-reported past pro-environmental behavior of these groups (M = 3.720, R = .268, T-square = 24.477, df = 1, 6; p < .005). The behavior intentions (M = 4.046, R = .341, T-square = 69.846, df = 1, 5; p < .001), as well as the environmental self-identity (M = 4.153, R = .418, T-square = 245.325, df = 1, 5; p < .001) of these two groups, were also significantly different. The environmental guilt of these groups was also significantly different, despite having lower scores (M = 3.504, R = .160, T-square = 7.961, df = 1, 5; p < .05). Meanwhile, the environmental concerns of these groups were insignificantly different (M = 4.807, R = .026, T-square = 1.633, df = 1, 5; p > .1). This is due to the “ceiling effects” (Wang et al., 2008), meaning that participants in both groups generally already stated that their environmental concern was high, with minimal fluctuations. Appendix Table A8 provides the results of Hotelling’s T-squared test.
An independent samples t-test was conducted to compare the general behavior of all measured factors across all time points in Gili and non-Gili groups. For self-reported past pro-environmental behavior, the significant difference in the scores for Gili (M = 4.15, SD = 1.008) and non-Gili (M = 3.73, SD = 1.201) was found only in Week 6; t(110) = 2.003, p = .048 (two-tailed). The magnitude of the differences in the means (mean difference = 0.422, 95% CI: [0.004, 0.840]) was large, with a Cohen’s d of 1.114. In Week 5, the difference was partially significant: Gili (M = 4.19, SD = 0.962), non-Gili (M = 3.81, SD = 1.210); t(110) = 1.802, p = .074 (two-tailed). The magnitude of the differences in the means (mean difference = 0.375, 95% CI: [−0.037, 0.788]) was large, with a Cohen’s d of 1.100.
For pro-environmental behavior intention, the most notable finding was only a partial significant difference in the scores for Gili (M = 4.34, SD = 0.960) and non-Gili (M = 3.86, SD = 0.937) in Week 4; t(110) = 2.649, p = .099 (two-tailed). The magnitude of the differences in the means (mean difference = 0.314, 95% CI: [0.120, 0.831]) was large, with a Cohen’s d of 0.948. For environmental guilt, Week 4 was also the only time where a partial significant difference in the scores for Gili (M = 3.81, SD = 1.051) and non-Gili (M = 3.49, SD = 0.947) was found; t(110) = 1.663, p = .099 (two-tailed). The magnitude of the differences in the means (mean difference = 0.314, 95% CI: [−0.060, 0.688]) was large, with a Cohen’s d of 0.997. Interestingly, although no significant differences were found in the scores for the environmental concern of the Gili and non-Gili groups, the scores for their environmental self-identity were always significantly different across all time points (all p values < .01). Appendix Table A9 provides the full results from the independent samples t-test.
These two tests illustrate that flashback nudging, given only to the Gili group, could be an effective catalyst for the different behaviors observed in (almost) all factors. Essentially, all research participants share the same demographic background, except that this experiment tapped into the Gili group’s pro-environmental behavior experience through the use of flashback nudging. The significant differences in the two groups’ self-reported past pro-environmental behavior, pro-environmental behavior intention, environmental self-identity, and environmental guilt found in Hotelling’s T-squared test can, therefore, be attributed to the flashback nudging. This finding was further supported by the independent samples t-test results, showing that the two groups’ environmental self-identity scores were consistently significantly different. This demonstrates the potential effectiveness of flashback nudging in strengthening the environmental self-identity of the tourists who visit the Gili Islands. In a closer observation, the significant difference in the self-reported past pro-environmental behavior among the two groups was found only in Week 6, with behavior intention and environmental guilt starting to show a sign of difference in Week 4. This observation indicates that flashback nudging takes time to achieve its full effectiveness. Lastly, due to ceiling effects (Wang et al., 2008), there were no specific, meaningful insights that can be observed from environmental concerns in these two tests.
Longitudinal Structural Equation Modeling and Multigroup Analysis
Further longitudinal analyses were needed to evaluate the relationships among observed factors and better understand how the self-reported performance of pro-environmental behavior was eventually achieved, as well as that they differed between these two groups. First, a base model that best explains the relationships among all measured factors was formulated. There were three proposed models, and the fit of each of their models was evaluated. In Model 1—the most complex model—environmental guilt, environmental self-identity, and environmental concern are connected to pro-environmental behavior intention and self-reported past behavior; the past week’s self-reported pro-environmental behavior is connected to the current self-reported past behavior. Model 2 still includes the past week’s self-reported pro-environmental behavior, whereas environmental guilt, environmental self-identity, and environmental concern are not connected to self-reported past pro-environmental behavior. Model 3 does not have the past week’s self-reported pro-environmental behavior.
The PLS-SEM algorithm test was conducted in SmartPLS 4 using the data from Week 6 on all three proposed models. The overall fit of the measurement model was evaluated by looking at several parameters, such as the values of standardized root mean square residual (SRMR ≤ 0.08), normed fit index (NFI ≥ 0.90), and Bayesian information criterion (BIC; Hair et al., 2019). The results showed that Model 2 was the best-performing model with the lowest BIC values (self-reported past pro-environmental behavior = −127.751, behavior intention = −108.63) and other parameters within the threshold (SRMR = 0.058, NFI = 0.916).
Once the base model was established, data from the other weeks (Week 1 to Week 5) were analyzed to ensure consistent results. The results demonstrated that the scores were consistently within the range of acceptable values for model fit (SRMR: [0.043, 0.072], NFI: [0.799, 0.92]). Although one NFI value was slightly below the acceptable 0.800 for exploratory research (Time 1 = 0.799), the model is considered acceptable and can be used for further analysis because the NFI values in other time points were always close to the ideal 0.900. The individual base models from different time points were combined to form the longitudinal model. A PLS-SEM algorithm test was carried out to check the model fit, finding that the SRMR value was 0.099, above the recommended 0.080. After removing problematic items from the model (i.e., EG3 from Weeks 5 and 6 due to high VIF values), the model fit was improved to an acceptable level (SRMR = 0.078; Roemer, 2016). Appendix Table A10 summarizes the results of the comparison of base model options, the evaluation of the base model across all time points, and the evaluation of model fit for the longitudinal model.
The construct validity was measured by evaluating convergent and discriminant validity. The standardized factor loadings of all items were greater than the suggested 0.5 cutoffs (see Appendix Table A11). They ranged from 0.583 to 0.998, except for one item slightly below the threshold yet still considered acceptable: EG1t3 (0.494). Internal consistency reliability was established because all Cronbach’s alpha values were above the threshold of .6 (Hair et al., 2019). Convergent validity was established because all average variance extracted (AVE) values were above the threshold of 0.5. Composite reliability (CR) values were all greater than the threshold value of .7. The variance inflation factor (VIF) values were all less than 10, which showed no critical multicollinearity issues. Across different time points, R-square values for pro-environmental behavior intention (.475, .691) and self-reported past pro-environmental behavior (.395, .716) were considered moderate to substantial (see Appendix Table A12). Discriminant validity was confirmed by evaluating the results of heterotrait–monotrait (HTMT) and the Fornell–Larcker criterion (see Appendix Table A13). All values were within an acceptable threshold, except for one HTMT score slightly above the threshold of 0.900 (EI5–EI6 = 0.903). Because no issues were found in other parameters for these two factors, no changes were made to the composition of the overall model.
Further analysis was conducted by estimating the path coefficients via bias-corrected bootstrapping using 5,000 subsamples. Results showed that the effects of environmental guilt on pro-environmental behavior intention fluctuated (see Figure 3), with a significant effect found only in Week 2. Furthermore, the effects were only partially significant in Weeks 3, 4, and 6. The environmental guilt effects on pro-environmental behavior intentions were insignificant in Weeks 1 and 5. The environmental self-identity effects on pro-environmental behavior intentions were significant across the different weeks, except for Week 2. Environmental concern significantly affected pro-environmental behavior intention only in Weeks 3 and 4. Self-reported pro-environmental behavior in the previous week consistently and significantly predicted pro-environmental behavior intention for the next week. Pro-environmental behavior intention failed to predict self-reported pro-environmental behavior twice, namely in Weeks 4 and 6. Finally, self-reported pro-environmental behavior in the past week always had significant effects on the self-reported pro-environmental behavior the week after, with only one being partially significant, from Week 1 to Week 2 (p = .064). The positive changes in carry-over effects were significant for self-reported pro-environmental behavior from t3 to t4 and from t5 to t6. Figure 3 summarizes the results of the longitudinal structural equation modeling analysis (see Appendix Table A14 for the complete results).
Structural model.
PLS-SEM multigroup analysis was performed in SmartPLS 4 using the bias-corrected and accelerated bootstrapping method with 5,000 subsamples. In multigroup analysis, environmental guilt and environmental concern were dropped due to errors encountered in the analysis. These errors may be attributable to the small sample size or the somewhat inconsistent performance of these constructs compared to environmental self-identity, which was more consistent in its significance. When the PLS-SEM algorithm was run to check the model fit, both SRMR (0.053) and NFI (0.833) were within the threshold.
As such, relationships between environmental self-identity (EI), pro-environmental behavior intention (IP), and self-reported past pro-environmental behavior (SP) over time across these two groups were analyzed. Multigroup analysis results showed that a significant difference was found only in the relationship between EIt6 and IPt6 (see Appendix Table A15 for the complete multigroup analysis results). Meanwhile, the differences between SPt3 and IPt4, SPt5 and IPt6, IPt6 and SPt6, and SPt5 and SPt6 in these groups were partially significant. In short, significant differences were found only in the final week, showing that nudges take time to work effectively. Another notable result was the difference in the coefficient between environmental self-identity and pro-environmental behavior intention in these two groups, which increased from Week 3 until a significant difference was achieved in Week 6. The relationship between environmental self-identity and pro-environmental behavior intention was stronger in the Gili group. This finding successfully demonstrated the role of flashback nudging in facilitating the pro-environmental behavior spillover in the Gili group. Figures 4 and 5 display the longitudinal structural equation modeling analysis results of the Gili and non-Gili groups, respectively. To enhance the clarity of the discussion of the research findings, Table 2 presents a summary of the hypotheses tested in this study, their levels of support, and a detailed discussion of the corresponding results.
Gili group.
Non-Gili group.
Summary of Hypotheses Testing Results.
Taking a Closer Look into Flashback Nudging Mechanisms Through Moderation Analysis
As presented in Table 1, socio-demographic information such as gender, age, education, and income was collected from all research participants. Because the primary focus of this study was to understand the mechanisms behind flashback nudging, a moderation analysis of the Gili group data, incorporating the four socio-demographic factors, was conducted using SmartPLS 4.0 based on the longitudinal model. Although most moderator relationships did not yield significant results, a few interesting observations with (partially) significant findings offer valuable insights for future research (see Table 3).
Summary of (Partially) Significant Results Found in the Moderation Analysis.
Income positively influenced self-reported past behavior in Week 2 (β = .206, t = 2.063), but its effect weakened in Week 4 (β = .192, t = 1.823), indicating a diminishing impact over time. This observation is further supported by the partially significant negative moderating effect of income on the relationship between environmental self-identity and pro-environmental behavior intention in Week 6 (β = −.203, t = 1.807). This finding suggests that individuals with strong environmental self-identity may eventually perceive their income as less critical in shaping their pro-environmental behavior intentions, highlighting the catalytic potential of flashback nudging to transform tourists into pro-environmental individuals by reminding them of their past pro-environmental behavior.
Education demonstrated mixed effects. Despite being partially significant, education negatively moderated the relationship between pro-environmental behavior intention in Week 1 and self-reported past behavior in Week 2 (β = −.181, t = 1.738). This result may suggest that strong early intentions, coupled with higher education, could result in lower immediate action. Conversely, at a later stage, education positively moderated the relationship between pro-environmental behavior intention in Week 4 and self-reported past behavior in Week 5 (β = .268, t = 2.394), indicating that education may eventually help translate intentions into sustained behavior. However, there was one instance where education had a partially significant negative direct effect on pro-environmental behavior intention, namely in Week 5 (β = −.149, t = 1.807). Because this finding occurred only once, it may reflect potential barriers such as time constraints, financial pressures or lack of access to environmentally friendly transport options that could ultimately hinder the performance of pro-environmental behavior.
Age also played a role, albeit only a partially significant one, as it negatively moderated the relationship between environmental self-identity and pro-environmental behavior intention in Week 6 (β = −.203, t = 1.866). Older individuals with strong environmental self-identity may face challenges in translating their values into actions, likely due to established routines and resource constraints. Although these findings reveal intriguing dynamics, caution is warranted in their interpretation because the partially significant results were observed only at specific time points rather than consistently across all measurements. Building on these indicative findings, future interventions could enhance the application of flashback nudging by incorporating facilitation measures that address income, education, and perceived age-related accessibility. This could be achieved through improved transport efficiency and government support, aligning with Majid, Tussyadiah, Kim, and Chen’s (2024) recommendations regarding the role of socio-demographic factors as potential moderators in pro-environmental behavior spillover mechanisms.
Conclusion and Implications
This study responds to calls for further research to understand better how artificial intelligence technology, coupled with flashback nudging as a behavior change intervention, can facilitate pro-environmental behavior spillover in the tourism context through a longitudinal experiment (Dolnicar et al., 2024; Majid, Tussyadiah, & Kim, 2024: Majid, Tussyadiah, Kim, & Chen, 2024; Wu et al., 2021). The dynamics of environmental guilt, environmental concern, and environmental self-identity as latent factors were observed in their relationships with people’s pro-environmental behavior intentions and self-reported past pro-environmental behaviors. The impact of educative nudges and flashback nudging on research participants has also been examined.
Theoretical Implications
The longitudinal observations of environmental guilt, environmental concern, and environmental self-identity in this study reveal important theoretical implications. The results indicate that environmental guilt is a relatively unstable construct in a longitudinal context. The fluctuation of significance levels in the longitudinal structural equation modeling demonstrates the nature of human feelings of guilt. The performance of pro-environmental behavior causes environmental guilt to decline, whereas the absence of pro-environmental behavior performance increases feelings of guilt. Such fluctuations may be normal when adopting a new pro-environmental behavior. This finding further contextualizes previous literature reporting that environmental guilt significantly predicts pro-environmental behavior (Mallett, 2012). Although it might be the case in a one-off cross-sectional survey, when the observation continues, the feelings of guilt may fluctuate until the individually fully adopts the pro-environmental behavior as their identity. Therefore, the finding in this study corroborates the proposition by Truelove et al. (2021) that guilt and identity are important determinants of pro-environmental behavior and can be stimulated.
With respect to environmental concern, the findings in this study echo three important observations from previous literature. First, the nature of the survey items developed by Fujii (2006) may cause the values to reach ceiling effects, which restricts the number of possible insights into the significance of the factor’s influence on behavior intention and self-reported past behavior. Second, in a closer investigation, the declining trend in environmental concerns extends our understanding that, as Melis et al. (2014) pointed out, this can indicate positive change toward the increased importance of climate concerns. This leads to the third observation: Environmental concern, as predicted by Bamberg (2003), may not be a direct predictor of pro-environmental behavior. In our study, the decline in environmental concerns may be related to the increase in environmental self-identity. As people begin to translate their concerns into actions, they start perceiving themselves as having more environmental self-identity. The opposing directions of these two factors warrant more scholarly attention because it may help us better understand how to operate these two underlying factors.
As evidenced in the longitudinal structural equation modeling and multigroup analysis results, environmental self-identity is the stable construct and as such should remain the focus of future research on pro-environmental behavior spillover. Due to challenges in conducting longitudinal experiments to observe environmental self-identity as a factor (Wu et al., 2021), scholars appear to have missed the opportunity to develop thoughtful, intentional interventions that can capitalize on this particular underlying construct (Dolnicar et al., 2024). Although how flashback nudging affects environmental self-identity in general terms is well understood in cross-sectional contexts (e.g., Lacasse, 2016; Van der Werff et al., 2014), it is worth highlighting how this construct became the catalyst of pro-environmental behavior intention from the third week and eventually caused a significant difference in the two groups’ self-reported past performance of pro-environmental behavior in the final week. This finding provides further support for the self-perception theory (Bem, 1972) such that it can be further operationalized into the development of interventions for pro-environmental behavior spillover.
Regarding the interaction between pro-environmental behavior intention and self-reported past pro-environmental behavior, this study’s findings differ from those of previous studies that reported an intention–behavior gap (Wu et al., 2021). In this study, although pro-environmental behavior intention did not consistently significantly predict self-reported past pro-environmental behavior, when the coefficient was negative (indicating an intention–behavior gap), the relationships between the constructs were insignificant, lending little support for a conclusion toward the identification of a gap. The explanation is due either to the provision of intervention or to the broad nature of the item measuring self-reported past pro-environmental behavior. Interestingly, self-reported pro-environmental behavior in the past week may consistently affect both pro-environmental behavior intention for the upcoming week as well as the upcoming week’s self-reported pro-environmental behavior. It is worth noting that the relationship between the past week’s self-reported pro-environmental behavior and pro-environmental behavior intention may stem from the fact that they were measured simultaneously, which causes recall bias (Althubaiti, 2016). However, the effect of the past week’s self-reported pro-environmental behavior on the upcoming week’s self-reported pro-environmental behavior corroborates a popular hypothesis about human psychology, explained below.
Scholars have often found that people’s past actions are a good predictor of their future behaviors (e.g., Bentler & Speckart, 1981). In the current study, however, this postulation is expanded by an observation that it is the fact that people can change their behavior following the given nudges in the first week (partially significant from SPt1 to SPt2) that serves as a precedent for the future possibility of continued behavior change (see Figure 3). This finding also confirms the proposition from Dolnicar et al. (2024), suggesting the importance of capturing people’s actual pro-environmental behavior in research, rather than relying on behavior intention. A few significant carry-over effects found in self-reported pro-environmental behavior at later time points (from t3 to t4 and from t5 to t6) further suggest that this construct is stabilized over time (Roemer, 2016). Thus, this study illuminates how scholars can theoretically understand the underlying mechanisms behind pro-environmental behavior spillovers from the tourism context, specifically concerning the increased adoption of environmentally friendly transportation (M. J. Kim et al., 2023). In addition, this study contributes to the literature on the stimuli-organism-response framework by applying it to the longitudinal experiment, responding to calls in previous tourism research (e.g., Hsiao & Tang, 2021; Sun et al., 2021).
Managerial Implications
An interesting finding from the control group in this study indicates that educative nudges trigger behavior change among individuals. Given the ubiquity of technology that can nudge people at scale, such as conversational artificial intelligence, policymakers can benefit from using such technologies to begin educating society about the importance of transitioning into a more sustainable way of living. Among tourists who have engaged in pro-environmental behavior at destinations where pro-environmental regulations apply, flashback nudging strengthens their environmental self-identity and further establishes pro-environmental behavior adoption. With that finding, this study simultaneously advances the recommendations from the previous literature about the importance of providing post-visit action resources to tourists returning from pro-environmental destinations (e.g., Hughes et al., 2011; Wu et al., 2015).
Furthermore, thanks to advances in artificial intelligence, the concept proposed in this study can accommodate other suggestions, such as the need for reflective engagement (Ballantyne, Packer, & Falk, 2011) and interpretive experiences (Ballantyne, Packer, & Sutherland, 2011). This study presents a novel application of human-AI interactions in the post-visit stage, expanding beyond the traditional use of AI for trip planning (e.g., Hrankai & Mak, 2025; Kang et al., 2024). Conversational artificial intelligence can engage users in self-reflection and help them make meaning of their past experiences by giving them alternative ways to interpret those experiences. Especially important is that this study identifies the critical role of the past performance of pro-environmental tourism behavior and sending flashback nudging to leverage that memory (Y. Kim et al., 2022). Previous studies (e.g., Hughes et al., 2011) often involved tourists who only witnessed pro-environmental phenomena without necessarily performing pro-environmental behavior. Therefore, the managerial implications offered by this study are more substantial because people’s self-perceptions are built based on their true past experiences. Referring to the thesis provided by Kahneman (2011) and Sunstein (2016) on system 2 nudges that target deliberative processing, the provision of flashback nudging can eventually result in a more lasting behavior change.
That said, policymakers and managers of tourist destinations with pro-environmental regulations in place can begin leveraging those locations as catalysts for transformation (Soulard et al., 2021). These stakeholders can develop targeted interventions to bridge different contexts that tourists will have to go through upon leaving the tourist sites. This measure would align with the principles behind the emerging “regenerative tourism” concept, where tourism initiatives are expected to produce net positive effects for its larger socio-ecological systems (Bellato & Pollock, 2023). Conversational artificial intelligence, such as chatbots, can carry out these interventions to facilitate pro-environmental behavior spillovers through digital nudging (Purohit & Holzer, 2019). Such investments and measures will advance the AI4GoodTourism framework into practice by leveraging the power of artificial intelligence to help humans live more sustainably (Majid et al., 2023).
Limitations and Future Research
Some limitations of this study should be addressed in future research. Firstly, the measurement of past pro-environmental behavior relied on self-report measures. Although these measures provided valuable insights, they are also potentially subject to biases, such as social desirability bias—particularly when research involves socially sensitive topics such as pro-environmental behavior. It was necessary to use self-reported measures due to feasibility constraints, including limited resources and the longitudinal design. Similar studies (e.g., Kaida & Kaida, 2019; Prati et al., 2015; Unanue et al., 2016) that also relied on self-reports have yielded important insights. However, future research could benefit from employing objective measures, including automatic tracking tools (e.g., electronic records of public transportation use or the global positioning systems trackers) to improve the accuracy of behavioral assessments and more reliably capture actual pro-environmental behavior.
Secondly, this study merged four different types of environmentally friendly transportation, which may have influenced the perceived ease for participants in adopting the suggested pro-environmental behavior. Future research could refine this approach by analyzing individual pro-environmental actions separately, considering the specific characteristics of each tourist destination and the availability of environmentally friendly transport options in the region. Such granular analysis would allow for more tailored interventions and nuanced insight into pro-environmental behavior spillover.
Thirdly, the 6-week study duration may have been too short to observe lasting effects or the formation of new daily habits. Longitudinal research in real-world tourism contexts is often constrained by financial, logistical, and ethical limitations, making the limited timeframe of this study an unavoidable challenge (Dolnicar et al., 2024; Wu et al., 2021). Nevertheless, this study highlights the importance of extending the study duration in future research to capture the long-term dynamics of pro-environmental behavior spillover and its potential to translate into sustained habits. This suggestion aligns with previous calls for more research investigating the long-term effects of nudges (Beshears & Kosowsky, 2020). Moreover, this study faced potential group equivalence issues. The Gili group’s higher responsiveness to the intervention may have been influenced by pre-existing pro-environmental tendencies beyond those controlled for at baseline, rather than solely by the intervention itself. Future research could incorporate a broader range of baseline control variables related to pro-environmental behaviors to further isolate the intervention’s specific impact. Additionally, employing experimental designs with random assignment or matched comparison groups could strengthen causal inferences.
Finally, this study also confronted contextual limitations. The non-Gili group consisted primarily of undergraduate students, which may limit generalizability. However, because the spillover process transcends different settings, the use of the Gili Islands as a study context underscores the value of specific case studies in pro-environmental behavior spillover research (Majid, Tussyadiah, & Kim, 2024; Wu et al., 2021). This study calls for future research to explore diverse tourist destinations and develop intentional interventions that bridge the gap between home and holiday contexts, promoting positive behavior spillover. Such efforts align with the recommendations of Dolnicar et al. (2024), emphasizing the role of interventions in advancing sustainable development through the tourism industry. Despite the challenges they entail, quasi-experimental designs in longitudinal research remain a critical tool for understanding the tourism sector’s contributions to sustainability.
Supplemental Material
sj-docx-1-jtr-10.1177_00472875251337777 – Supplemental material for From Destination to Daily Life: A Longitudinal Study on the Effects of Flashback Nudging on Pro-environmental Behavior Spillover
Supplemental material, sj-docx-1-jtr-10.1177_00472875251337777 for From Destination to Daily Life: A Longitudinal Study on the Effects of Flashback Nudging on Pro-environmental Behavior Spillover by Gilang Maulana Majid, Iis Tussyadiah, Yoo Ri Kim and Jason Li Chen in Journal of Travel Research
Footnotes
Appendices
Additional Nudges for the Gili Group.
| Week | Content |
|---|---|
| 1 | Let’s remember your Gili experience. You didn’t use polluting motorized vehicles, did you? You behaved environmentally friendly while there. Let’s keep being environmentally friendly even though you’re no longer there! You can do it! |
| 2 | If you managed to cycle on Gili Trawangan for an hour, it could burn around 245 kilos of calories! Very good for health! (Cyclescheme, 2021). Let’s continue to behave environmentally friendly in our daily lives so that we are always healthy! |
| 3 | If you happened to walk on the Gilis for 2 hours, you surely have stepped 10,000 times a day and burned more than 300 kilos (Firman & Cohen, 2022). Your environmentally friendly behavior is certainly good for your own health. Let’s keep it up, you can stay environmentally friendly even though you’re no longer in Gili! |
| 4 | Environmentally friendly transportation regulations in Gili can save 20.77 tons of carbon emissions for a year! This amount is equal to the amount of CO2 absorbed by about 25 hectares of forest for one year. Very cool, right! (Carbon Footprint, n.d.; Subaidi, 2023). Let’s stay environmentally friendly by reducing the use of private vehicles! |
| 5 | Walking regularly for about 20 minutes every day can help you avoid getting sick up to 40% better! (Harvard Health Publishing, 2023). I’m sure you walked more often when you were on the Gilis. If you could walk more while on the Gilis, for sure you can also do this habit in your daily life! |
| 6 | You feel fresh and tranquil while on Gili, right? Come on, let’s cultivate environmentally friendly behavior by reducing the use of private motorized vehicles so that our earth is protected! |
Note. The calculation for data shown in Week 4 was conducted on the website Carbon Footprint. The assumption is that 250,000 tourists that visit Gili Islands would not use motorbike (small up to 125 cc) for at least 1 km during the day they spend on the Gili Islands. The calculation resulted in 20.80 metric tons of carbon emissions. The number of Gili visitors for a year was roughly estimated based on the data reported in Subaidi (2023).
Author Note
Also available in Indonesian. See supplemental material for details.
Author Contributions
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Center for Higher Education Funding, Ministry of Education, Culture, Research and Technology of the Republic of Indonesia, and the Indonesian Endowment Fund for Education under the scholarship grant received by the corresponding author.
Ethical Approval and Informed Consent
This research received a favorable ethics opinion from the University of Surrey Ethics Committee. All participants provided written informed consent prior to the commencement of data collection.
Data Availability
Data can be made available upon request.
Supplemental Material
Supplemental material for this article is available online.
Author Biographies
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.
