Fear Emotion Decelerates Outcomes of Climate Change Education

Emotion mediating educational outcomes has been recognized for decades. However, empirical experiments to test these predictions, particularly for climate change education, which is often mixed with various emotions, are rare. In this study, we conducted a two-week climate change education program with speci�c video clips designed to induce fear or hope in students to explore how emotions affect educational outcomes. The study involved 1,730 students from nine middle schools in three coastal cities (Xiamen, Shenzhen, and Ningbo) in China. The results demonstrated that emotional video clips are a successful stimulus for the target emotion. In the hope treatment group, emotion did not signi�cantly affect the educational outcomes, as indicated by the limited change in students’ climate change involvement, self-eciency, and mitigation behavior. However, in the fear treatment group, emotion signi�cantly decelerated students’ change in mitigation behavior compared to the lecture-only group. These decelerated behaviors are mostly located at the behavior change of low carbon life. Based on the mediation analysis, both the hope group and lecture-only group had direct effects on climate change mitigation behavior, while the fear group only had an indirect effect on climate change behavior, mediated primarily by climate change involvement. The study thus highlighted that both negative and positive emotions should not indiscriminately used in climate change education programs to safeguard signi�cant educational outcomes.


Introduction
As a conscious mental reaction and subjectively experienced feeling, emotion is often associated with behavior changes and has drawn attention from environmental educationists for decades (Nabi 2003; Chapman et al. 2017;Nabi et al. 2018).While much debate remains concerning the effects of positive and negative emotions, an early study by Iozzi (1989) suggested that an emotionally-oriented attitude is a key predictor of unconscious behavior.Negative stimuli are detected more easily at lower levels of input or exposure than positive stimuli (Dijksterhuis and Aarts 2003); thus, negative emotions may enhance participants' attention or concern (Schupp et al. 2007).On the other hand, fear arousal is de ned as an unpleasant emotional state triggered by the perception of threatening stimuli (Rogers and Deckner 1975).This emotional state involves physiological arousal as well as cognitive, affective, and behavioral responses directed toward reduction or elimination of fear; people in fear condition may perform fear control due to criticize or denying the threat presented in a message (Witte et al. 2001).In this case, a negative or counterproductive impact on fear may exist, which has been supported by some studies (Ruiter et al. 2014).The emotion of hope, however, appears to be constantly recommended for inclusion in environmental education (Ojala 2012;Stevenson et al. 2018).Hope may create goal congruence and enhance one's perceived control of a situation, leading to a more favorable attitude and increasing the intention to act (Lee et al. 2017).Positive emotions may also affect motivation and behavior for a long time (Fredrickson, 2001;Tannenbaum et al., 2015).Thus, bringing positive emotional perspectives into environmental educational practice has been increasingly adopted ( Kollmuss and Agyeman 2002).Some evidence suggested that discrete emotions alone were able to explain a large proportion of the variance (50%) in public global warming policy support, which is higher than several other variables (Smith and Leiserowitz 2014).Incorporating emotion into considerations for CCE is also ineluctable, as many communicators, governments, nongovernmental organizations, and individuals often deliver information related to climate change with strong emotions (Salama and Aboukoura 2018).Meanwhile, there is increasing debate on the effects of targeting speci c emotions on motivating or inhibiting public engagement with climate change, especially for the emotions of fear and hope (Ojala 2012; Lemanski and Villegas 2019).For example, a limited series of studies investigating actual behavior in response to fear appeals showed that the results were mixed (Hine and Gifford, 1991;Bloodhart et al. 2018).Some researchers pointed out that instilling fear in people might cause avoidance and reductions in personal e cacy, whereas making people feel hopeful caused increased e cacy and engagement (O'Neill and Nicholson-Cole 2009).Others argued that messages designed to induce feelings of hope and optimism about climate change may lower motivation to engage in mitigation efforts (Hornsey and Fielding 2016).Arguments for these paradoxical results may be due to the research design, as often the emotion may also be associated with knowledge delivery or other psychological variables; it is not always easy to distinguish the effects of emotion on other variables (Chapman et al. 2017).Very little research on manipulating different emotions to determine the pure effect of emotion on CCE has been conducted.Climate change affects mankind on a global scale, but the people most vulnerable to its impacts are children and youth in developing countries, where 85% of the world's children live (UNICEF 2014).
Although climate action is undoubtedly imminent, the brunt of climate change impacts is projected to arrive just as younger generations reach adulthood (IPCC 2014).Hence, effective climate change education for teenagers is urgent in preparing them to tackle imminent environmental issues.Prior studies have shown that CCE may build concern and enhance knowledge among adolescents (Guy et al. 2014;Stevenson et al. 2014).More recent research suggests that fear can raise teenagers' risk perceptions and concerns about climate change and promote pro-environmental behavior (Skurka et al. 2018).Although the evidence is suggestive, it is not conclusive due to the limited experimental evidence and the incomprehensive understanding of the nature or direction of the relationship between different emotions of climate change engagement (Chapman et al. 2017).Therefore, the effectiveness of emotion in climate change education remains unclear, especially for today's youth.
In this study, we designed a curriculum focusing on the factual knowledge of climatic change, coupled with a video clip intended to instill emotions of fear or hope as treatments to determine how emotion affects curriculum-based CCE.Based on a review of the previous literature, we propose the following hypotheses.
H 1 : The two emotional treatments signi cantly induce target emotions.A1): concept of climate, facts of global climatic change, main causes for climate change, and climate mitigation action.Considering that the 7th-grade students might not have primary knowledge of climate change science, we simpli ed the concepts and used several animated videos to explain the scienti c phenomenon (see Supplementary 1).Meanwhile, to decrease the emotional impact of lectures, we tried our best to apply a neutral image (line chart) and science language and nally recorded all four lectures as videos to avoid bias caused by the lecturers themselves.Each lecture video lasted 15 minutes to ensure that all procedures were completed in one lesson (40 minutes).
The emotional clips used in the experiment originated from two documentaries of climate change: "A Warming Earth" (Chinese) in 2008 and "An Inconvenient Sequel: Truth to Power" (English dialogue with Chinese subtitles) in 2019.The former lm was awarded the title of Best Science Film from the 28st Golden Rooster Award in China, and the latter has been widely promoted and adopted as educational material worldwide (David 2006;Expatica 2007;Leask 2007).By adjusting the needs of the contents, time length, and target emotion, we edited four clips for each lm (see Supplementary 2).In the "A Warming Earth," we selected "The Super Typhoon Saomai" and "Sea Level Rising" as fear clips and "Carbon Sequestration Technology" and "New Clean Energy" as hope clips.In "An Inconvenient Sequel: Truth to Power," we used "The Super Typhoon Haiyan" and "Extreme Weather" as fear clips and "Technical Leapfrog in Developing countries" and "The Pairs Agreement" as hope clips.The four fear clips were watched by the fear group students, and the hope clips were watched by the hope group students.They watched these clips before lecture videos in four classes.For example, fear group students in their rst class watched "The Super Typhoon Saomai" rst and then the lecture video "concept of climate".

Pilot studies
To con rm that all emotional clips could induce the target emotion successfully, we conducted two pilot tests before the formal study.A total of 29 adults in Menglun, Xishuangbanna, Yunnan Province, including ve environmental education specialists and 24 college students, took part in the rst pilot test.First, we asked each participant to extract the video sequence randomly.Then we played the eight emotional clips one by one, requesting that participants ll in the emotional questionnaire immediately after each of the eight videos.In this test, all the target emotions were salient, which proved that fear and hope can be induced by emotional clips in adults (see Supplementary 3).Considering that one fear clip ("Mosquito-borne virus") was di cult for children to understand, we replaced it with another familiar topic: "The Super Typhoon Haiyan."We also considered evaluate positive, negative, and neutral contents in these two emotions (see Supplementary Table A2).
The second pilot test was conducted at a junior school in Kunming, Yunnan Province with six rst-grade classes (N = 285).This pilot study aimed to 1) retest the validity of emotional clips, 2) examine the di culty and duration of lecture videos, and 3) test the reliability and validity of questionnaires in the target population.All six classes of students (N = 285) completed pre-and post-questionnaires, and four of them (N = 180) attended the education program.In this pilot study, target emotions were signi cantly higher than other emotions among junior students, which also supported the validation of the emotional clips.The reliability of the two items in the pre-test questionnaire did not meet the criteria (climate change concern, α = 0.57; climate change involvement, α = 0.46).However, after eliminating two items by con rmatory factor analysis, the reliability of all the variables was increased to an acceptable level (α > 0.60).

Study sites and procedures
Previous research found that fatalities from natural hazards, vulnerability to sea level rise, and living within oodplains can largely predict the risk perception of climate change (Brody et al. 2008;Spence et al. 2011;Demski et al. 2017).Coastal cities might be more physically vulnerable to climate change due to being situated in low-lying areas within the immediate proximity of the coast.Thus, we selected coastal cities in three coastal provinces as the sites for studies: Shenzhen, Guangdong Province, Xiamen, Fujian Province, and Ningbo, Zhejiang Province.The formal study was conducted from September 2019 to December 2019 in nine junior schools.In each city, 2-4 junior schools were selected using convenience sampling.
Data collection in this study adapted pre-and post-intervention surveys and a control group design (Fig. 1).In each school, four classes of the same grade (grade 7, corresponding to ages 11-15) participated in the study.One class was randomly chosen as the control group for which we did not provide any intervention; the other three classes were conducted as three treatments: 1) lecture-only, 2) lecture with hope-based clips, and 3) lecture with fear-based clips.For the lecture-only treatment, we provided four lectures in a 2-week period; each lecture consisted of a 15-minute presentation video (total intervention time = (15 min × 4); see supplement 1).For the two emotion treatments, we provided a 2minute emotional video prior to each lecture presentation (total intervention time = (2 min + 15 min) × 4).

Measures
Emotional responses were assessed after watching each video.Students in the emotion condition were asked to rate the strength of the 15 emotions they felt while watching the video on a scale of 1 (not at all) to 5 (extremely) (also see Nabi et al. 2018) (supplementary Table A2).Of these, three items related to fear (anxious, afraid, worried) (Cronbach's α = 0.90); three items related to hope (hopeful, encouraged, optimistic) (Cronbach's α = 0.88); three items related to sadness (sad, upset, disappointed) (Cronbach's α = 0.89), three items related to neutral emotions (careless, calm, relaxed) (Cronbach's α = 0.75); and three items related to happiness (happiness, exciting, energetical) (Cronbach's α = 0.91).Given the bland and descriptive nature of the control message, the 15 emotion items would have been incongruous and potentially counterproductive by unintentionally priming emotional responses unrelated to message content.Thus, participants in this condition were asked about general emotional intensity with four semantic differential items: This article seemed "emotional/impressive/exciting" (0 = not at all and 5 = extremely).To more closely match the time allotted to complete the threat message post-test, participants were asked several ller questions (e.g., "I learned a lot from this article," "The article seemed factual") that were not intended for analysis.
Perceptions The survey measured a range of perceptions associated with climate change, related behavior, behavioral intention, and key demographic variables.We used published scales to test these variables and prepared four to six items for each variable in the initial version.After the second pilot study, we deleted some items that had low reliability.The nal questionnaire (Supplementary Table A3) includes seven variables using a 5-point scale ranging from "strongly agree" to "strongly disagree": four Demographics The socio-demographic variables of gender, age, ethnicity and "Big Five Personality" (Rammstedt and John 2007) were included in the analysis (Supplementary Table A3).The "Big Five Personality" describes ve discrete dimensions of personality: agreeableness, conscientiousness, extroversion, neuroticism, and openness.In this study, we only found four sections through principal component analysis: agreeableness, conscientiousness, extraversion, and openness (Supplementary Table A4).The post-test survey had the same content, although the order of the items differed from the pre-test.

Data analysis
To verify the effectiveness of emotional clips, univariate statistical models with one-way analysis of variance (ANOVA) were used.
The longitudinal analysis used 1,539 adolescents with matched pre-and post-test surveys, allowing changes for all climate-related outcomes to be calculated at the individual level.Univariate statistical models and pairwise comparisons with the Tukey HSD test were used to evaluate the effectiveness of the emotion on perceptions and mitigation intention.Type III sum of squares was used to determine statistical signi cance (p < 0.05).
To understand how changes in mitigation behavior were related to the changes in perceptions within the four groups, controlling site as a random factor, multiple regression using a mixed linear model in R (version 3.5.1)was programmed.Four regression models were used: all socio-demographic variables were included in Model 1, treatment variables were obtained in Model 2, all changes in perceptual variables were involved in Model 3, and all socio-demographic and perceptual variables were included in Model 4 (see Table 1).Further, multiple-mediation analysis was constructed in Mplus (N fear = 797, N hope = 754, N lecture−only = 789) to understand the changes in mitigation behavior in the three treatments.Sociodemographic variables were included in the analysis as covariates.Bias-corrected bootstrap 95% con dence intervals (CIs) were based on 1,000 bootstrap samples for indirect effects.

The effectiveness of emotion on climate science knowledge and mitigation behavior
Compared with the control group, the lecture-only group signi cantly improved climate change involvement, self-e cacy, climate science knowledge, and mitigation behavior, while there was no signi cant difference in climate change concern and total behavioral intention (Fig. 2a-f).The fear group had signi cantly increased self-e ciency and knowledge compared to the control group (Fig. 2c, d).The hope group showed signi cantly increased knowledge and total behavior compared to the control group (Fig. 2d, e).Speci cally, the lecture-only group had higher climate science knowledge than the hope group (Fig. 2d), while the lecture-only group showed improved mitigation behavior compared to the fear group (Fig. 2e).
Considering about the two subgroups (learning and communication, low carbon life) of mitigation behavior.All three treatments displayed higher communication behaviors than the control group (Fig. 2g).For the low carbon life subgroup, only the lecture-only group showed the greatest improvement compared to the control group.The hope group had no signi cant difference compared to the lecture-only group and the fear group, while the fear group was signi cantly lower than the lecture-only group (Fig. 2h).

The direct and indirect effect of treatments and mitigation behavior
The mixed linear model showed that all three treatments and changes in perceptions were positively correlated with changes in mitigation behavior, both for considering the factors together and separately (Table 1).Changes in concern, involvement, and self-e ciency were strong predictors of changes in mitigation behavior.Agreeableness and pre-test knowledge positively improved mitigation behavior.However, other personality traits and demographic variables failed to explain the changes in mitigation behavior.The multiple-mediation model showed that the fear group indirectly affected mitigation behavior (Fig. 3).
When the indirect effects of those mediators were included in the model, the positive relationship between fear group and mitigation behavior increased (standardized direct effect, c' = 0.061, p = 0.082; standardized total effect, c = 0.088, p = 0.016); change in involvement, with a small improvement through the fear group (B = 0.077, p < 0.05) and a large contribution to mitigation behavior (B = 0.24, p < 0.001), was the most important mediator of the relationship (95% CI: 0.01, 0.04).
The hope group could increase mitigation behavior directly (standardized direct effect, c' = 0.093, p = 0.006; standardized total effect, c = 0.116, p = 0.002).Moreover, the lecture-only group improved mitigation behavior both directly and indirectly in mediation analysis (standardized direct effect, c' = 0.136, p < 0.001; standardized total effect, c = 0.193, p < 0.001); change in involvement was the most important mediator of the relationship (95% CI: 0.01, 0.05), with a large improvement in the lecture-only group (B = 0.13, p < 0.001) and a large contribution to mitigation behavior (B = 0.23, p < 0.001).The change in concern was also a signi cant mediator (95% CI: 0.01, 0.03).The three mediation models explained 8%, 13%, and 13% of the variance in mitigation behavior, respectively.

Discussion
As emotions may mediate educational outcomes, a more comprehensive understanding of their effects is needed.This study showed that fear would indeed decelerate educational outcomes, as indicated by selfreported changes in climate change mitigation behavior, particularly on the change of low carbon life.Furthermore, hope did not have a signi cant impact on educational outcomes compared to the control, fear, and lecture-only groups.Thus, bringing emotion into EEC requires more prudential consideration.
A previous study suggested that messages with emotion improve pro-environmental behavior more than natural messages, especially negative emotions (Morris et al. 2019).Negative emotions often cause more attention or concern (Schupp et al., 2007), and thus may generate more action (Skurka et al. 2018).
Consequently, science consultants advocated the use of fear and catastrophe narratives to represent the consequences of climate change (Doulton and Brown 2009;Hulme 2008).However, in this study, concern related climate change did not differ signi cantly among the treatments; there was no difference even with the control group, suggesting that students in the coastal cities in this study may already hold relatively high levels of concern for climate change.On the other hand, an increasing number of studies have suggested that the predominant negative information about climate change does not bene t the cultivation of mitigation actions (Hart and Feldman 2014; Morisson and Hat eld-Dodds 2011).A study in the environmental psychology eld (Greitemeyer 2013) indicated that fear-framed, climate changea rming lms did not have a signi cant, positive impact on concern for the environment.It may be that fear does not trigger consideration, but rather apathy for the future consequences of climate change.A recent study that tested preferences for and impacts of three negative emotions (fear, sadness, and anger) in comparison to messages framed without emotion, found that people generally preferred messages framed without emotion (Bloodhart et al. 2018).Thus, an increasing recommendation found in academic research and literature is that participants prefer non-emotional messages to modi ed emotional messages about climate change (Bloodhart et al. 2018).
Our study has con rmed that fear has a negative impact on educational outcomes compared to the lecture-only group.Fredrickson (2001) noted that on occasions when survival is threatened, negative affect prompts action that brings immediate and direct adaptive bene ts.There is also evidence that individuals with depression are more focused on the possible costs than the possible bene ts of speci c risks (Pietromonaco and Rook 1987; Yuen and Lee 2003).This suggests that individuals high in negative affect will have a more limited understanding of the natural environment, and the adoption of mitigation behavior is costly.This may lead individuals with higher negative emotions to avoid them and engage instead in behaviors that yield immediate, rather than long-term bene ts.This might explain why fear decelerates low carbon life in mitigation behavior in this study.
On the other hand, there is a call for promoting the development of hope among youth in climate change education (Stevenson et al. 2018), as hope may enhance people's self-e ciency (O'Neill and Nicholson-Cole 2009), thus leading to a more positive behavior change.However, hope emotion treatment did not contribute signi cantly to most of the psychological variables compared to the lecture-only group in this study.Notably, even the change in knowledge in the hope treatment group was lower than that of the lecture-only group.One possible reason is that hope may weaken people's motivation and cause them to become less engaged (Hornsey and Fielding 2016).These ndings are consistent with one reading of fantasy realization theory (Oettingen 2012): If people focus entirely on a desired future and do not contrast that future with a negative current reality, the effect can be counterproductive in terms of working toward that future.Another study found that "constructive hope" (being hopeful because peopleindividually and collectively-can reduce climate change) is a predictor of increased policy support and political engagement, whereas "false hope" (being hopeful because something external other than people will x the problem) predicts the opposite effect (Marlon et al. 2019).The complexity of the emotion of hope reminds us that hope-lled messages about climatic change need to be balanced with active reminders of the negative current reality.Educators also need to engage students taking personal responsibility for changing behaviors through cooperation, participation, and organization of social, political, and cultural efforts.
Our results provide not only an effective educational program for adolescents' behavior but also a theoretical model to understand the factors for changes in mitigation behavior.Hope group, lecture-only group, climate change concern, involvement, and self-e ciency were found to improve mitigation behavior directly.A prior study also found hope in youth's direct impact on mitigation behavior (Stevenson et al. 2018).Education can potentially build climate change hope by outlining strategies to mitigate climate change (Ojala 2016).The fear group in uenced mitigation behavior was fully mediated by climate change involvement, and the lecture-only group in uenced mitigation behavior was partially mediated by climate change concerns and involvement.Both fear and lecture-only groups translated involvement into a greater willingness for mitigation behavior.The elaboration likelihood model (ELM) model suggests that people increase involvement by consciously processing important information in detail and depth through the central process (Lazard and Atkinson 2015).Moreover, the appraisaltendency framework (ATF) points out that different emotions have different cognitive processes (Tiedens and Linton 2001).Fear is characterized by low levels of certainty and negative valence.People tend to process fear information more elaborately, and fear is associated with central processing.Among all three mediation models, climate change involvement proved to be a more important variable than climate change concern and self-e ciency in in uencing climate change mitigation behavior.
This study demonstrates the possibility of inducing fear and hope in adolescents through emotional videos.Prior research used lm, music, pictures, and smells to induce emotion, while lm presentations proved to be more effective in eliciting discrete emotions than music and smell (Lench et al. 2011).This research successfully induced fear and hope in adolescents through documentary clips related to climate change, showing that adolescents' emotions were indeed affected by mass media, such as news, the Internet, television, and lms.Recently, an increasing number of educators around the world have adopted lms or documentaries for use in their classrooms (Harness and Drossman 2011; Maier et al. 2014).No matter what kind of education method is utilized, emotions induced by pictures or videos exist in and out of the classroom.Therefore, educators need to be aware of the emotions that materials may evoke and their impact on education outcomes.
We note that our data are too limited to be generalized to other areas and other age groups.To increase personal involvement, people in different places need relevant stimuli (Moser, 2010).For instance, sea level rise can induce fear among teenagers residing in coastal areas, while its effectiveness on inland students is unclear.Further interventions and assessments are required for interior areas on engagement.
Emotion stimuli in neutral materials should also be considered in future research.Although we carefully selected lecture materials, using science graphs instead of disaster photos, these may still arouse Con icts Interest I declare that we have no known competing nancial interests or personal relationships that could have appeared to in uence the work reported in this paper.
Chadwick 2015; Jacobson et al. 2018; Greenaway and Fielding 2019).Climate change education (CCE) aims to promote climate change mitigation behavior.Studies have highlighted the importance of emotions, beliefs, and perceptions in CCE (Pooley and o'Connor 2000;

H 2 :
Students who receive lecture-only treatment show signi cant improvement in their knowledge of climate change, climate change mitigation behavior, and behavior intention.

H 3 :H 4 : 2 Methods 2 . 1
Fear emotion treatment enhances educational outcomes via an increase in concern and involvement.Hope emotion treatment enhances educational outcomes via an increase in self-e ciency.Lectures for CCE and emotional stimuli The lecture videos used in this study came from Stanford Middle School courseware, climate change lesson plans for years 7-11 from "Oxfam Education," "China's Perspective on Climate Change" in a massive open online course (MOOC), and other open-source educational videos on YouTube.The lectures were presented four times, including four aspects (see supplementary Table questions about climate change concern (Cronbach's α = 0.81), taken from previous research on climate change risk perception (Stevenson et al. 2015); ve questions concerning climate change involvement (α = 0.78) according to environment involvement research (Schuhwerk and Lefkoff-Hagius 1995); ve terms to test self-e ciency (α = 0.85) ve questions evaluating mitigation behavior (α = 0.84), which came from climate change engagement research among young people (Mead et al. 2012; Ojala 2012); four questions testing mitigation behavior intention (α = 0.85), adapted from Hu and Chen (2016), who primarily conducted research on Chinese students; and 15 questions on climate science knowledge, divided into two aspects: single-item in subjective knowledge and objective knowledge (true or false, single choice; α = 0.76), selected these items from previous research about climate change knowledge on teenagers (Tobler et al. 2012).We used factor analysis to nd that climate change mitigation behavior can also be divided into two subgroups (KMO = 0.82): learning and communication (α = 0.87) and low carbon life (α = 0.74).
surprise and concern among teenagers.How to avoid the possible emotional interference of knowledge videos requires further exploration.Other objective assessment methods, such as skin conductance response(Ho and Lipp, 2014) or eye tracking(Lisa et al., 2017), can be adapted to improve the accuracy of emotion data.Current information on climate change consists of many emotional messages (Salama and Aboukoura 2018).Many educational materials also contain striking emotional information, often invoking fear or sadness.Our study suggests that in order to achieve effective educational outcomes for climate change education, we should neither try to abuse negative emotion (as this can be counterproductive) nor to oversell positive emotion (because it can be overdone and needs to be balanced).Meanwhile, educators should work to boost students' involvement and level of concern to improve climate-change mitigation behavior in CCE.Particular storylines and visual materials based on local areas could increase teenagers' involvement and decrease the psychological distance of climate change (Mcdonald et al. 2015).Other educational tools to strengthen personal relevance, such as engaging in deliberative discussions and implementing school or community projects, have proven to be effective CCE strategies (Hu and Chen 2016, Monroe et al. 2017).Declarations Funding This was funded by the Chinese Academy of Sciences 135 Program (grant number 2017 XTBG-F04).

Figures
Figures

Figure 1 Design
Figure 1

Figure 2 The
Figure 2

Table 1 A
linear mixed model analysis of Big Five personality, demographic variables, knowledge, four treatments, attitude change variables and mitigation behavior change variables