Emerging Infectious Outbreak Inhibits Pain Empathy Mediated Prosocial

***THIS IS A WORKING PAPER THAT HAS NOT BEEN PEER-REVIEWED***

Emerging infectious outbreak inhibits pain empathy

mediated prosocial behaviour

Siqi Cao1,2, Yanyan Qi3, Qi Huang4, Yuanchen Wang5, Xinchen Han6, Xun Liu1,2*, Haiyan Wu1,2*

1CAS Key Laboratory of Behavioral Science, Institute of Psychology, Beijing, China

2Department of Psychology, University of Chinese Academy of Sciences, Beijing, China

3Department of Psychology, School of Education, Zhengzhou University, Zhengzhou, China

4College of Education and Science, Henan University, Kaifeng, China

5Department of Brain and Cognitive Science, University of Rochester, Rochester, NY, United States

6School of Astronomy and Space Sciences, University of Chinese Academy of Sciences, Beijing, China

Corresponding author

Please address correspondence to Xun Liu ([email protected]) or Haiyan Wu

([email protected])

Disclaimer: This is preliminary scientific work that has not been peer reviewed and requires replication. We share it here to inform other scientists conducting research on this topic, but the reported findings should not be used as a basis for policy or practice.

Abstract

People differ in experienced anxiety, empathy, and prosocial willingness during the coronavirus outbreak. Although increased empathy has been associated with prosocial behaviour, little is known about how does the pandemic change people’s empathy and prosocial willingness. We conducted a study with 1,190 participants before (N=520) and after (N=570) the COVID-19 outbreak, with measures of empathy trait, pain empathy and prosocial willingness. Here we show that the prosocial willingness decreased significantly during the COVID-19 outbreak, in accordance with compassion fatigue theory. Trait empathy could affect the prosocial willingness through empathy level. Through distance from the epicentre (i.e. Wuhan) moderated trait empathy and empathy level, anxiety generated by this outbreak impaired prosocial willingness. Further, news discriminability also played a role in trait empathy change. Given how emergency health events influence emotion and the effects of personal traits on prosocial willingness, social media users and providers should understand the negative effects of information over-exposure on mental health during the outbreak.

Introduction

The impact of the current global health emergency, termed COVID-19 by the World

Health Organization (WHO), has significantly changed people's lives worldwide, casting a shadow over the mind of an enormous population. Although clarity and immediacy guaranteed the right to be informed, people are bombarded by various information and their moods enter a cycle of complexity (Thompson, Jones, Holman,

& Silver, 2019). Encouraging slogan and unequivocal instruction can stabilize mood, while aversive news induces emotional responses to pain in others such as sadness and anxiety (Batson et al., 1991). For example, when we see the pain or even death of a patient and the desperation of the health care workers, a heart-breaking feeling hits us and tears well up. The capacity associated with feeling and evaluating the pain level of other as well as further understanding others is specifically referred to as pain empathy (Fitzgibbon, Giummarra, Georgiou-Karistianis, Enticott, & Bradshaw, 2010), which is a fundamental feature of our innate capacity (Van der Graaff et al., 2014).

Watching others’ pain experiences may activate one’s own sensory region of the brain

(Lamm, Decety, & Singer, 2011) as if it were a personal experience (Carr, Iacoboni,

Dubeau, Mazziotta, & Lenzi, 2003).

For decades, research have shown empathy is a potential motivator for prosocial behaviours. From rescuing behaviour in ants (Nowbahari, Scohier, Durand, & Hollis,

2009) to the feeding each other in bats, prosocial behaviour is observed throughout the non-human species (Wascher, Scheiber, & Kotrschal, 2008). Decety et.al (2016) emphasized the mediation effect of empathy on pro-sociality in conspecifics, when one is sensitive to others’ stress and well-being. A spontaneous motivation to alleviate the

pain of others is generated when people feel others’ suffering (Inagaki & Eisenberger,

2016; Roberts, Strayer, & Denham, 2014), and the implementation of altruistic behaviour to a certain extent relieves negative feelings (Singer et al., 2006), or even physical pain on oneself (Y. Wang, Ge, Zhang, Wang, & Xie, 2020). Although empathy is an aspect of social responsiveness, compassion fatigue (i.e., physical and emotional exhaustion and a profound decrease in empathy) happens when one’s empathy reaches its limit (Joinson, 1992), or caused by a short-term, intense, or numbness and depression exposure to unfortunate situation (Figley, 2002). With the COVID-19 outbreak and massive news updating, global people experience heavy exposure to victims’ pain or negative news (McAlonan et al., 2007). Here we tested whether people’s empathy and prosocial tendency varies before and during this outbreak.

At the early outbreak stage, Wuhan was widely reported as the epicentre of

COVID-19, which captured global attention and concern. There are two contrary effects would indicate about psychological effect from epicentre: 1) the ripple effect suggests that the impact of an unfortunate event decays gradually as ripples spread outward from the epicentre (Slovic, 1987), while 2) the ‘Psychological Typhoon Eye’ effect (PTE), posit the closer one is to the epicentre of a devastated area, the lower a resident's concern/anxiety about health and safety. Both of the two models place emphasis on the distance from the outbreak epicentre on human emotions, and raise hypothesis regarding the influencing direction of the COVID-19 outbreak (i.e. facilitating or decreasing concern/anxiety levels along with further distance to Wuhan).

Given the importance of empathy and prosocial behaviour for collaboration during a crisis, it is crucial to assess the relationship between empathy level and prosocial willingness across individuals near to or far from the epicentre. Moreover, several

psychological factors may contribute to this link between empathy and pro-sociality, like exposure to adverse life events, information seeking or news reading, anxiety etc, should be taken into consideration. During a life-threatening period, altruistic behaviours would be selected to facilitate cooperation and enable group members to collectively survive various crises, such as food shortages and natural disasters (Fehr

& Fischbacher, 2003). However, the impact of quarantines strongly influences approach motives (Brooks et al., 2020), and people immersed in emergency-related information could lead to compassion fatigue and decrease prosocial tendency. We asked whether people would be numb to other’s pain and hypothesized that (i) as the

COVID-19 outbreak and news reports increase, empathy-induced prosocial willingness (pain-sharing willingness) would decrease during the outbreak, and this prosocial willingness shift would be most strongly evidenced by empathy levels (pain ratings of the performer) decreasing in our online task; (ii) the overexposure to various information influences empathy; (iii) outbreak-generated anxiety may influence empathy and prosocial behaviour in terms of distance to the epicentre (Wuhan).

Methods

Participants

The first dataset (539 participants, 274 females, Mage = 28.43, SD = 7.84) was collected before the emerging infectious disease outbreak of COVID-19

(2019.11.21~2019.11.23), while the second dataset (570 participants, 366 females,

Mage = 25.08, SD = 8.75) was collected during the peak period of public concern about

COVID-19 (2020.2.23~2020.2.24). We distributed the task and questionnaires on an online platform. As Table 1 shows, in pre- and post-outbreak, the groups were unevenly distributed in terms of age (t(1088)=6.64, P<0.001) and gender (χ2(df=1,

N=1090)=14.89, P<0.001). Hence, we controlled for age and gender as possible confounding variables when we compared the response variables cross the timepoints.

Stimuli

The stimuli consisted of 12 video clips portraying actors telling either the truth or lies about the pain induced by electric shocks. Twelve participants in Beijing (six females and six males, Mage = 22.4) were recruited as actors for video recording. All actors were asked to either express their true feelings about pain (electric shock pain) or lie when they were not really experiencing pain.

In the real pain scenario (i.e. telling the truth about their pain), participants were asked to express and describe their genuine feelings of pain evoked by the electric shock induction approach. The electric shock approach is a commonly used pain induction method, in which an electric shock is delivered to the participants fingers.

In the fake pain scenario (i.e. lying about the pain), the participants were asked to pretend to have been shocked and to describe those fake feelings. The actors in the fake pain scenario had been sitting in the room and watched videos of people getting electric shocks, but they were not shocked personally. The research paradigm was approved by the China Academy of Science ethics committee, and all performers in the video recording gave informed consent to participate in the study.

Online study task

Figure 1 illustrates the present study components in both the first and second online study tasks.

Fig.1. Online data components

Video rating task

Twelve video clips (six genuine pain and six vicarious pain) were presented to the participants in randomized order. After each video clip played, the participants judged whether the actors actually received electric shocks (choose ‘1’ if it is true, choose ‘2’ if it is not). Then, the participants were asked to rate (choose form ‘1’ = ‘None’ to ‘7’ =

‘very much’ ) ‘how much pain do you think the person felt?’ , followed by ‘how much pain you feel when you are watching this video clip?’, and ‘how much would you like to share the pain of this person?' There was no time limit to provide feedback, and the next video was played after all the ratings were completed. After watching and rating all 12 videos, participants completed the questionnaire session.

Questionnaires

The participants completed the Interpersonal Reactivity Index (IRI) at the end of

the online task. The IRI is a widely used scale to measure individual differences in empathy. It captures four separate aspects: Perspective Taking, Fantasy, Empathic

Concern, and Personal Distress(Davis, 1983). Participants rated how well each of the

28 statements in the IRI described their feelings on a scale ranging from ‘1 = totally not agree’ to ‘5 = totally agree’. The Cronbach's coefficient of the questionnaire in this study was 0.71.

Further, regarding the effect of quarantine, participants were asked to complete the Autism-spectrum Quotient (AQ) (Baron-Cohen, Hoekstra, Knickmeyer, &

Wheelwright, 2006; Baron-Cohen, Wheelwright, Skinner, Martin, & Clubley, 2001), which has been used to assess autism spectrum disorder in adults(Ashwood et al.,

2016). The Cronbach's coefficient of the questionnaire in this study was 0.38.

Note that there were some differences in the questionnaires that were administered in the second study given that it took place during the peak period of public concern about the coronavirus. Specifically, the second online task included the

State-Trait Anxiety Inventory (STAI) to capture both the participants’ trait anxiety and state anxiety components(Spielberger, 2010). The Cronbach's coefficient of the questionnaire in this study was 0.94. Moreover, an 18-item questionnaire about the emerging COVID-19 outbreak was included (Fig. 5, Fig. S10, more details in supplementary materials).

Demographic information

After the questionnaire session, participants were asked to provide demographic information including location, age, gender, etc.

Statistical analysis

A total of 520 and 570 participants in the first and second dataset, respectively,

were included in the final analysis. Nineteen participants with missing values of demographic information (three participants did not provide age data and 16 participants did not provide distance information calculated by IP) were removed from the final analysis.

Data were analysed using SPSS 25.0 statistical software (IBM Corp., Armonk,

NY) and R. Based on the hypotheses of the research, descriptive analysis were first conducted in SPSS. For the data characteristics comparison between the two datasets

(pre vs. post-outbreak), group comparisons of nominal variables were analysed using the chi-square test, and independent samples t-tests were used to compare group differences (pre vs. post groups). Linear regression analysis exploration and correlation analysis were executed using R function lm() and cor.test(). As for model construction, mediation and moderation models were computed with PROCESS 3.4 in SPSS and R packages MeMoBootR for serial moderated mediation model.

Results

Video authenticity judgment

Video authenticity judgment significantly affected perceived pain in others (i.e. empathy level), self-pain (vicarious experience), and shared pain (i.e. prosocial willingness, supplementary materials Table 1). Therefore, it was controlled in the further model analysis as a possible confounding variable.

Empathy levels and prosocial willingness pre- and post-outbreak

To determine the effect of the coronavirus outbreak on empathy and prosocial behaviour, t-tests were conducted to compare the differences in perceived pain in others (i.e. empathy level) and shared pain (prosocial willingness) between pre- and post-outbreak. The results showed both empathy levels (pre: M=3.89, SD=1.55; post:

M=3.60, SD=1.53; t=4.27 , P<0.001) and prosocial willingness (pre: M=3.73, SD=1.67; post: M=3.50, SD=1.63; t=3.16 , P<0.01) significantly decreased during the outbreak

(Fig. 2).

Trait empathy: pre- and post-outbreak

Since we were interested in the potential influence of the epidemic disease on trait empathy, a t-test was also conducted on IRI subscales, including perspective- taking, trait empathy, personal distress, and fantasy. The results showed that when the outbreak occurred, a negative fluctuation was witnessed in the individuals’ empathy-related traits (Fig. 2). Except for personal distress (pre: M=23.53, SD=4.07; post: M= 4.14, SD= 3.50; t=-2.62, P<0.01) and fantasy (pre: M=24.54, SD=4.68; post:

M=23.98, SD=4.49; t=2.03 , P<0.05), all other IRI subscale scores were significantly decreased in the post-outbreak: trait empathy (pre: M=25.05, SD=3.50; post: M=24.56,

SD=3.03; t=2.48, P<0.05) , and perspective taking (pre: M=24.14, SD=2.97; post:

M=23.43, SD=4.06; t=5.71 , P<0.001). However, the AQ seems to be stable in the two stages (pre: M=23.47 SD=3.27; post: M=23.52, SD=2.86; t=-0.27, P=0.79). The aforementioned results showed that empathy levels, trait empathy, perspective-taking, fantasy, and prosocial willingness decreased during the coronavirus outbreak.

The mediation effect of empathy level

We further investigated whether empathy level would mediate the relationship between individual differences in trait empathy and prosocial willingness by establishing three mediation models with trait empathy (empathy, perspective-taking, and fantasy) as the respective predictor, empathy level as a mediator, and prosocial willingness as the dependent variable. Data were standardized in the analysis of the moderation and mediation models and we used PROCESS in SPSS for the mediation

and moderation analysis(Hayes, 2012).

The results showed a significant mediating model for trait empathy and perspective-taking, but not for fantasy. Trait empathy positively predicted prosocial willingness (β=0.052, t(2095)=3.16, P=0.0016; path C’) and empathy level in a positive manner (β=0.057, t(2098)=3.49, P< 0.001; path A). Furthermore, empathy level robustly predicted prosocial willingness (β=0.70, t(2095)=21.97, P<0.001; path B) (Fig.

3).

Perspective taking significantly positively predicted prosocial willingness

(β=0.096, t(2095)=5.85, P<0.001; path C’), and empathy level as well (β=0.043, t(2098)=2.64, P=0.008; path A). Similarly, perspective taking significantly predicted prosocial willingness (β=0.62, t(2095)=22.07, P<0.001; path B) (Fig. 3).

Finally, the same model was not proved valid for fantasy. Fantasy did not predict empathy level (β=0.03, t(2098)=1.96, P=0.05; path A); although, empathy level robustly predicted prosocial willingness (β=0.62, t(2095)=21.98, P<0.001; path B). In addition, fantasy did not predict prosocial willingness (β=0.01, t(2095)=0.69, P=0.49; path C’).

Our results indicated that empathy level could only mediate the predictive relationship from empathy and perspective taking to prosocial willingness, but not from fantasy. We further included a binary variable of time (pre- and post-outbreak) as a moderator to test whether the coronavirus would have a moderating effect (model 14 in PROCESS). These results showed that the moderating effect happened on an empathy level predicting prosocial willingness (β=-0.083, P = 0.012 < 0.05, 95% CI: [-

0.15,-0.02], path B), suggesting that the predictive extent from empathy to prosocial behaviour decreased after the coronavirus outbreak (Fig. S5).

Anxiety and distance from the outbreak epicentre

To test the cognitive empathy-driven prosocial behaviour changes in terms of distance under the emergency according to the PTE, a moderation model (model 1 in

PROCESS) was used. The variable ‘distance to Wuhan’ is a binary value (near/far) separated by the mean of the distance calculated by the IP address (see supplementary materials Method). There was no evidence that distance influenced how empathy level correlated with prosocial willingness (Fig. S5).

We then investigated the effect of anxiety on this process, considering that anxiety would have a negative influence on empathy and prosocial behaviours. To investigate this possibility, we established a serial two-part mediation model (model 6 in the

PROCESS). The results showed that anxiety could influence prosocial willingness through trait empathy and empathy level (Fig. 4). Regarding the current mental state development, the State-Trait Anxiety Inventory (STAI) was used to evaluate post- outbreak anxiety levels (Table 1), along with some direct questions about how people felt during the outbreak in terms of anxiety, depression, and fear (see more in supplementary materials). Although state anxiety did not directly predict prosocial willingness, but according to the suppressing effect the effect of trait empathy and empathy level was valid (MacKinnon, Cheong, & Pirlott, 2012) (β=0.096, t(2095)=5.85,

P<0.001; path C’).

Finally, we explored the effect of distance according to the PTE. We calculated each participant’s distance to the disease epicentre, which was based on IP address

(near and far) by the mean of the distance (see supplementary material Method,

Distance to Wuhan calculation from IP address). The interaction of state anxiety and distance to Wuhan was significant (β=0.17, P=0.007). In the post-outbreak stage,

compared to people near the outbreak epicentre, the state anxiety of people in the far region did not sensitively predict trait empathy (Fig. S10). Therefore, we continued to analyse the possibility that distance could play a role in the serial mediation chain. Our result indicated that distance moderated the influence that state anxiety has on trait empathy and further affected prosocial willingness (Fig. 4).

Fig.2. Changes from pre- to post-outbreak

Fig.3. The mediation effect of empathy level a, Time functioned as a moderator in the path B

(empathy level to prosocial willingness) (pre: 95% CI = [1.5%,5.6%], 44.12% mediated; post: 95%

CI = [1.2%, 4.8%], 38.88% mediated). b, time still had moderating effect in the path B (pre, 95%

CI = [0.6%, 4.6%], 22.08% mediated; post, 95% CI = [0.5%, 4.1%], 19.00% mediated).

Fig.4. A serial two-part mediation model (post-outbreak). a, State anxiety did not predict overall prosocial willingness, β = -0.03, t(1119) = -0.964, P = 0.335 (path C). State anxiety predicted trait empathy β = -0.16, t(1119) = -5.78, P < 0.000 (path A1). State anxiety did not predict empathy level,

β = -0.015, t(1118) = -0.67, P = 0.504 (path A2), while trait empathy predicted empathy level, β =

0.08, t(1118) = 3.33, P < 0.001 (path DA1). Trait empathy predicted overall prosocial willingness,

β = 0.06, t(1117) = 2.595, P < 0.001 (path B1). As in any other models, empathy level predicted overall prosocial willingness, β = 0.52, t(1117) = 0.52, P < 0.001 (path B2). The direct effect of state anxiety could be suppressed that state anxiety did not predict overall prosocial willingness, β <

0.01, t(1117) = 0.07, P = 0.944 (path C’). The interaction between distance to Wuhan and state anxiety was significant, β = 0.17, t(1117) = 2.74, P < 0.01, which suggested there was a moderating effect of distance in the serial two-part mediation model. Trait empathy indirect effect: β = - 0.008,

95% CI = [-1.86%, - 0.06%], 23.57% mediated. Trait empathy and empathy level indirect effect:

β = - 0.008, 95% CI = [-1.5%, - 0.1%], 23.57% mediated. b, To figure out how anxiety and distance impact trait empathy. Similarly, a simple moderation model (model 1 in PROCESS) was used to

explore the idea. The result showed that state anxiety had a negatively predicting effect of trait empathy, β = -0.23, t(1120)=-6.27, P < 0.001. Distance to Wuhan did not predict trait empathy, β =

0.11, t(1120) = 1.82, P = 0.07. The interaction of state anxiety and distance to Wuhan was significant, β = 0.17, t(1120) = 2.72, P = 0.007 < 0.01. In the post-outbreak period, the state anxiety of people far from epicentre did not sensitively predict empathy compared to near population.

Fig.5. Results of linear regression predicting state anxiety from measures of the STAI dimension in selected 11-item out of total 18-item COVID-19 related questions, with validation in a subsample of 25% of participants. The left panel represents the discovery dataset and the right panel represents results from the validation dataset.

News discriminability, anxiety and empathy

Excessive information exposure makes people become confused about news authenticity. Again, a moderation model (model 1 in PROCESS) was used to examine how news discriminability induced anxiety-affected trait empathy (IRI subscale).

Fig.6. Trait empathy changes across state anxiety moderated by news discriminability

News discriminability moderated how state anxiety negatively predicted trait empathy (STAI subscale). State anxiety in this model predicted empathy (β=-0.17, t(1119)=-5.62, P<0.001). News discriminability did not significantly predict trait empathy (t(1119)=0.02, P=0.98). The more anxious people felt, the less empathy people displayed. The interaction of news discriminability and state anxiety was significant (β=-0.07, t(1119)=-2.38, P=0.02). In the post-outbreak period, people became confused about overwhelming news which imposed a bad influence on trait empathy.

Discussion

The salient public COVID-19 threat triggers government intervention and psychological responses to public health problems. The goal of this study was to investigate the effect of an emerging infectious outbreak on people’s empathy levels and prosocial willingness and further explore the underlying mechanisms of the relationship between these variables. Importantly, our study provides several moderation and mediation models to examine the effects of empathy on prosocial behaviour before and after the COVID-19 outbreak, with the consideration of the

distance to the epicentre. We explored the influence of potential impactors like emotional states, particularly anxiety on the link between empathy and prosocial willingness.

General empathy-driven kindness and its decrease during COVID-19 outbreak

Our data first extend previous work by illustrating the general empathy-driven prosocial disposition (Masten, Morelli, & Eisenberger, 2011; Roberts et al., 2014; Y.

Wang et al., 2020), that the correlation of empathy and prosocial willingness even during the outbreak of infectious emergence. From the evolutionary perspective, when people find others at a disadvantage such as physical pain and mental setback experiences, a motivation to offer help is generated for both individual and collective benefit (Majdandžić, Amashaufer, Hummer, Windischberger, & Lamm, 2016). In fact, our results suggest the willingness to behave pro-socially would decrease as empathy levels (feeling other’s pain) decrease in the post-outbreak period. Moreover, we found trait empathy could influence sharing tendencies, which were mediated by the perception of others’ pain. ‘Collateral damage from trauma’, that is, those who do not experience misfortune directly, can also create a similar psychological impact when witnessing others’ pain experience (Gallese, 2003). Staub and Vollhardt (2008) put forward the ‘suffering altruism’ (altruism born of suffering) concept (Rao et al., 2011), and people may gain more social support and thus generate a sense of gratitude that enhances prosocial inclination (W. Wang, Wu, & Tian, 2018). Nevertheless, instead of enhancing the possibility of benevolence, less kindness was observed in the present study. First of all, the overall feeling of others’ pain significantly decreased after the outbreak. Emotion contagion as the primary stage of empathy, is widely seen in daily life, for example, yawning, tickling, and feeling distress, which represents certain

sensitivity to what others are experiencing (Zheng et al., 2020). Our results also showed a decreasing tendency of perspective taking ability in the post-outbreak that people less stood the perspective of others to feel others’ pain feelings during the outbreak period. Considering the possible large energy costs associated with empathy, empathy is never easy, and requires the involvement of emotional and cognitive aspects(Cameron et al., 2019). This may explain why empathy avoidance may be a preferable choice as people only have limited one’s inner resources.

Anxiety, information uncertainty and distance to the epicentre mediate the empathy level and prosociality

The finding of the mediation model with personal trait, emotion state and distance to epicentre may explain decreased pain empathy or prosocial tendency after the outbreak. Our data indicate the salience of trait empathy was markedly inhibited by anxiety level, variations in distance difference (near and far regions), and different news authenticity discriminability.

As one of the disruptive impacts on society, the COVID-19 makes people over- stressed and anxious, which may lead to the post-traumatic stress disorder years after the event (McAlonan et al., 2007). Since anxious feeling has been indicated to inhibit the neural underpinnings (i.e. mirror neuron system) of empathy (Gallese, 2003), we further explored how people’s feelings and trust are affected by anxiety. We found that people who currently felt anxious (state anxiety) and more confused about the authenticity of the news showed a lower level of empathy. It is worth noting that state anxiety, in addition to being directly predictive of trait empathy, can also be moderated by information uncertainty. Understandably, when one’s own survival is at stake, one can no longer provide help or emotional support to others.

Besides, other than the direct predictive effect, state anxiety also predicted trait empathy in terms of distance to the outbreak epicentre. Contrary to our hypothesis according to the PTE, an infectious pandemic disease indeed frustrated people near the epicentre more than people in regions farther away. Nevertheless, the moderation analysis indicated that distance predicted trait empathy post-outbreak. Although people far from the outbreak epicentre did not show different prosocial inclinations from the near population, the COVID-19 questions in dataset 2 (post-outbreak) suggested that people in the near region reported patients to receive more stigmatization and rejection from others. Stigma from others was a major theme throughout the literature (Lee, Chan, Chau, Kwok, & Kleinman, 2005), and it may be a long-lasting issue, even when the outbreak is under control. Therefore, compared to the far distance regions, people may experience a heavier psychological stress in the near distance region. In terms of the effects that anxiety imposed on empathy during the outbreak, the direction was inverse to what is suggested by the PTE or ripple effect.

The trait empathy in people near the epicentre declined faster as state anxiety grew, while in people outside the epicentre trait empathy did not change significantly when personal anxiety increased.

Whenever a public health event occurs, all kinds of information can be mixed, with novel, eye-catching information with some degree of uncertainty. Indeed, the nature of the media will make this type of information get more attention in more exaggerated ways (Thompson et al., 2019). When watching a lot of news regarding topics such as disasters, illness, and homelessness, people might experience empathy first and then symptoms similar to compassion fatigue, such as depression and anxiety to varying degrees (Kinnick, Krugman, & Cameron, 1996). Under the current barrage of

information, we also verified compassion fatigue in our study, protecting people from burnout (Figley, 2013). If empathy and compassion produce motivation, but there is no practical way of action, people would inevitably feel more anxious and miserable, which will further influence pro-sociality (Back & Arnold, 2014). At a time when compassion fatigue is emerging, we may need to re-examine our psychological capacity and the environment around us.

The trait empathy predicts empathy and prosocial willingness

Individual differences in trait empathy (measured by the IRI) exist, which largely influences behaviour tendency (Decety et al., 2016). In accordance to previous studies, we considered the ‘feeling of others’ pain’ ratings as indicators of empathy levels

(Danziger, Prkachin, & Willer, 2006; Fitzgibbon et al., 2010). Our findings indicat that empathy level was significantly predicted by the trait empathy, suggesting an effective induction of empathy in the study. We were primarily concerned with how trait empathy affected prosocial behaviour and the role of empathy level in the process. As expected, a robust mediation effect was established when the effect of COVID-19 (comparison between two-time stages) was taken into account. We found that trait empathy could predict prosocial willingness directly, but also indirectly through empathy level, which varied in pre- and post-timepoint. Perspective taking, the critical part of cognitive empathy, had similar mediating effects, referring to some extent to the theory of mind: an important aspect of empathy is the ability to put oneself in other people's shoes

(Singer, 2006).

Limitations

The present study has several limitations. First, data were obtained by self-report from a sample that was not fully randomly distribute. Second, the participants were in

isolation period during the second assessment. Third, we could not formally evaluate the specific psychometric properties of our questionnaire on the psychological effects of COVID-19. Fourth, due to the anonymous nature of the data collection, we could not pair individuals at the two-time intervals. Fifth, using a dichotomous distance variable, the possible influence of a continuous geographical position was not explored.

Sixth, even though a large sample was used in the study, the difference in IRI subscales between the two periods might have been caused by the unpaired sample that was actually systematically different. Besides, the slightly difference components of the data made it difficult to further examine the contrast between the two time stages.

To some extent, our data indicate how individuals’ prosocial willingness changed pre- and post-outbreak and provides a new perspective on how trait empathy can decrease pro-sociality behaviours. Empathy allows us to engage in more prosocial behaviours. However, being over-exposed to information generates anxiety, which dampens our prosocial motivation. Caring for disadvantaged people is conducive to resolving social contradictions, manifesting morality, easing conflict situations, and promoting a harmony society (Masten et al., 2011). Although kindness is desirable, the anxiety and over-exposure to negative information may lead to indifference, especially for care workers experiencing the overconsumption of their own empathy and feeling much more stressed because of close contact with patients’ death over and over (McAlonan et al., 2007). Thus, mental health services should be readily available for certain medical care workers worldwide.

To sum up, we aimed to investigate the effect of being chronically worried about the threat of diseases on engagement in trait empathy, adaptive empathy, or prosocial behavioural tendency, especially when diseases are particularly salient such as an

outbreak. The preconditions for intervention lay on the knowledge of changes in psychological states and the potential mechanisms. For example, it is necessary to raise awareness about the potential damage to mental health that over-exposure to information during the outbreak can cause. Thus, it is important to distract oneself from the anxiety-provoking pictures and headlines. Future interventions to help people manage their emotions during public health crises are worthy of attention.

Data and code availability

The data and code supporting the findings of this study are available from the https://osf.io/9bgkt/.

Acknowledgements

This work was supported by the National Natural Science Foundation of China [grant number: U1736125, 31400963], and CAS Key Laboratory of Behavioral Science,

Institute of Psychology to HW.

Author Contributions

H.W. conceived of the presented idea. S.C. and Q. H. implemented the study, S.C.,

Y.Q. and H.W. performed the statics and results discussion. Y. W., X. H. and H.W. verified the distance analytical methods. X.L. and H.W encouraged S.C. to investigate empathy fatigue and supervised the findings of this work. S.C. lead the manuscript writing, and all authors discussed the results and contributed to the final manuscript.

Competing Interests

All of the authors declare no competing interest.

References

Ashwood, K., Gillan, N., Horder, J., Hayward, H., Woodhouse, E., McEwen, F., . . . Wilson, C. (2016). Predicting the diagnosis of autism in adults using the Autism-Spectrum Quotient (AQ) questionnaire. Psychological medicine, 46(12), 2595-2604. Back, A. L., & Arnold, R. M. (2014). “Yes It's Sad, But What Should I Do?”: Moving from Empathy to Action in Discussing Goals of Care. Journal of palliative medicine, 17(2), 141-144. Baron-Cohen, S., Hoekstra, R. A., Knickmeyer, R., & Wheelwright, S. (2006). The autism-spectrum quotient (AQ)—adolescent version. Journal of autism and developmental disorders, 36(3), 343. Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001). The autism-spectrum quotient (AQ): Evidence from asperger syndrome/high-functioning autism, malesand females, scientists and mathematicians. Journal of autism and developmental disorders, 31(1), 5-17. Batson, C. D., Batson, J. G., Slingsby, J. K., Harrell, K. L., Peekna, H. M., & Todd, R. M. (1991). Empathic joy and the empathy-altruism hypothesis. Journal of personality and social psychology, 61(3), 413. Brooks, S. K., Webster, R. K., Smith, L. E., Woodland, L., Wessely, S., Greenberg, N., & Rubin, G. J. (2020). The psychological impact of quarantine and how to reduce it: rapid review of the evidence. The Lancet. Cameron, C. D., Hutcherson, C. A., Ferguson, A. M., Scheffer, J. A., Hadjiandreou, E., & Inzlicht, M. (2019). Empathy is hard work: People choose to avoid empathy because of its cognitive costs. Journal of Experimental Psychology: General, 148(6), 962-976. doi:10.1037/xge0000595 Carr, L., Iacoboni, M., Dubeau, M.-C., Mazziotta, J. C., & Lenzi, G. L. (2003). Neural mechanisms of empathy in humans: a relay from neural systems for imitation to limbic areas. Proceedings of the National Academy of Sciences, 100(9), 5497-5502. Danziger, N., Prkachin, K. M., & Willer, J.-C. (2006). Is pain the price of empathy? The perception of others' pain in patients with congenital insensitivity to pain. Brain, 129(9), 2494- 2507. doi:10.1093/brain/awl155 Davis, M. H. (1983). Measuring individual differences in empathy: Evidence for a multidimensional approach. Journal of personality and social psychology, 44(1), 113. Decety, J., Bartal, I. B.-A., Uzefovsky, F., & Knafo-Noam, A. (2016). Empathy as a driver of prosocial behaviour: highly conserved neurobehavioural mechanisms across species. Philosophical transactions of the Royal Society of London. Series B, Biological sciences, 371(1686), 20150077-20150077. doi:10.1098/rstb.2015.0077 Fehr, E., & Fischbacher, U. (2003). The nature of human altruism. Nature, 425(6960), 785-791. Figley, C. R. (2002). Treating compassion fatigue: Routledge. Figley, C. R. (2013). Compassion fatigue: Coping with secondary traumatic stress disorder in those who treat the traumatized: Routledge. Fitzgibbon, B. M., Giummarra, M. J., Georgiou-Karistianis, N., Enticott, P. G., & Bradshaw, J. L. (2010). Shared pain: from empathy to synaesthesia. Neuroscience & Biobehavioral Reviews, 34(4), 500-512. Gallese, V. (2003). The Roots of Empathy: The Shared Manifold Hypothesis and the Neural Basis of Intersubjectivity. Psychopathology, 36(4), 171-180. doi:10.1159/000072786 Hayes, A. F. (2012). PROCESS: A versatile computational tool for observed variable mediation, moderation, and conditional process modeling. In: University of Kansas, KS. Inagaki, T. K., & Eisenberger, N. I. (2016). Giving support to others reduces sympathetic nervous system‐related responses to stress. Psychophysiology, 53(4), 427-435.

Joinson, C. (1992). Coping with compassion fatigue. Nursing, 22(4), 116, 118. Kinnick, K. N., Krugman, D. M., & Cameron, G. T. (1996). Compassion Fatigue: Communication and Burnout toward Social Problems. Journalism & Mass Communication Quarterly, 73(3), 687-707. doi:10.1177/107769909607300314 Lamm, C., Decety, J., & Singer, T. (2011). Meta-analytic evidence for common and distinct neural networks associated with directly experienced pain and empathy for pain. Neuroimage, 54(3), 2492-2502. Lee, S., Chan, L. Y., Chau, A. M., Kwok, K. P., & Kleinman, A. (2005). The experience of SARS-related stigma at Amoy Gardens. Social science & medicine, 61(9), 2038-2046. MacKinnon, D. P., Cheong, J., & Pirlott, A. G. (2012). Statistical mediation analysis. Majdandžić, J., Amashaufer, S., Hummer, A., Windischberger, C., & Lamm, C. (2016). The selfless mind: How prefrontal involvement in mentalizing with similar and dissimilar others shapes empathy and prosocial behavior. Cognition, 157, 24-38. doi:https://doi.org/10.1016/j.cognition.2016.08.003 Masten, C. L., Morelli, S. A., & Eisenberger, N. I. (2011). An fMRI investigation of empathy for ‘social pain’and subsequent prosocial behavior. Neuroimage, 55(1), 381-388. McAlonan, G. M., Lee, A. M., Cheung, V., Cheung, C., Tsang, K. W., Sham, P. C., . . . Wong, J. G. (2007). Immediate and sustained psychological impact of an emerging infectious disease outbreak on health care workers. The Canadian Journal of Psychiatry, 52(4), 241-247. Nowbahari, E., Scohier, A., Durand, J.-L., & Hollis, K. L. (2009). Ants, Cataglyphis cursor, use precisely directed rescue behavior to free entrapped relatives. PloS one, 4(8). Rao, L.-L., Han, R., Ren, X.-P., Bai, X.-W., Zheng, R., Liu, H., . . . Li, S. (2011). Disadvantage and prosocial behavior: The effects of the Wenchuan earthquake. Evolution and Human Behavior, 32(1), 63-69. Roberts, W., Strayer, J., & Denham, S. (2014). Empathy, anger, guilt: Emotions and prosocial behaviour. Canadian Journal of Behavioural Science/Revue canadienne des sciences du comportement, 46(4), 465. Singer, T. (2006). The empathic brain: how, when and why? TRENDS in cognitive sciences, 10(10), 435-441. Singer, T., Seymour, B., O'Doherty, J. P., Stephan, K. E., Dolan, R. J., & Frith, C. D. (2006). Empathic neural responses are modulated by the perceived fairness of others. Nature, 439(7075), 466-469. Slovic, P. (1987). Perception of risk. 236(4799), 280-285. Spielberger, C. D. (2010). State‐Trait anxiety inventory. The Corsini encyclopedia of psychology, 1-1. Staub, E., & Vollhardt, J. (2008). Altruism born of suffering: The roots of caring and helping after victimization and other trauma. American Journal of Orthopsychiatry, 78(3), 267- 280. Thompson, R. R., Jones, N. M., Holman, E. A., & Silver, R. C. (2019). Media exposure to mass violence events can fuel a cycle of distress. Science advances, 5(4), eaav3502. Van der Graaff, J., Branje, S., De Wied, M., Hawk, S., Van Lier, P., & Meeus, W. (2014). Perspective taking and empathic concern in adolescence: Gender differences in developmental changes. Developmental psychology, 50(3), 881. Wang, W., Wu, X., & Tian, Y. (2018). Mediating roles of gratitude and social support in the relation between survivor guilt and posttraumatic stress disorder, posttraumatic growth among adolescents after the Ya’an Earthquake. Frontiers in psychology, 9, 2131. Wang, Y., Ge, J., Zhang, H., Wang, H., & Xie, X. (2020). Altruistic behaviors relieve physical pain. Proceedings of the National Academy of Sciences, 117, 950-958. Wascher, C. A., Scheiber, I. B., & Kotrschal, K. (2008). Heart rate modulation in bystanding geese watching social and non-social events. Proceedings of the Royal Society B: Biological Sciences, 275(1643), 1653-1659. Zheng, C., Huang, Y., Bo, B., Wei, L., Liang, Z., & Wang, Z. (2020). Projection from the 24

Anterior Cingulate Cortex to the Lateral Part of Mediodorsal Thalamus Modulates Vicarious Freezing Behavior. Neuroscience bulletin, 36(3), 217-229.

Figure legends

Fig. 1. Online data components

Fig. 2. Changes from pre- to post-outbreak

Fig. 3. The mediation effect of empathy level a, Time functioned as a moderator in the path B (empathy level to prosocial willingness) (pre: 95% CI = [1.5%,5.6%], 44.12% mediated; post: 95% CI = [1.2%, 4.8%], 38.88% mediated). b, time still had moderating effect in the path B (pre, 95% CI = [0.6%, 4.6%], 22.08% mediated; post, 95% CI =

[0.5%, 4.1%], 19.00% mediated).

Fig. 4. A serial two-part mediation model (post-outbreak). a, State anxiety did not predict overall prosocial willingness, β = -0.03, t(1119) = -0.964, P = 0.335 (path C).

State anxiety predicted trait empathy β = -0.16, t(1119) = -5.78, P < 0.000 (path A1).

State anxiety did not predict empathy level, β = -0.015, t(1118) = -0.67, P = 0.504 (path

A2), while trait empathy predicted empathy level, β = 0.08, t(1118) = 3.33, P < 0.001

(path DA1). Trait empathy predicted overall prosocial willingness, β = 0.06, t(1117) =

2.595, P < 0.001 (path B1). As in any other models, empathy level predicted overall prosocial willingness, β = 0.52, t(1117) = 0.52, P < 0.001 (path B2). The direct effect of state anxiety could be suppressed that state anxiety did not predict overall prosocial willingness, β < 0.01, t(1117) = 0.07, P = 0.944 (path C’). The interaction between distance to Wuhan and state anxiety was significant, β = 0.17, t(1117) = 2.74, P < 0.01, which suggested there was a moderating effect of distance in the serial two-part mediation model. Trait empathy indirect effect: β = - 0.008, 95% CI = [-1.86%, - 0.06%],

23.57% mediated. Trait empathy and empathy level indirect effect: β = - 0.008, 95%

CI = [-1.5%, - 0.1%], 23.57% mediated. b, To figure out how anxiety and distance impact trait empathy. Similarly, a simple moderation model (model 1 in PROCESS)

was used to explore the idea. The result showed that state anxiety had a negatively predicting effect of trait empathy, β = -0.23, t(1120)=-6.27, P < 0.001. Distance to

Wuhan did not predict trait empathy, β = 0.11, t(1120) = 1.82, P = 0.07. The interaction of state anxiety and distance to Wuhan was significant, β = 0.17, t(1120) = 2.72, P =

0.007 < 0.01. In the post-outbreak period, the state anxiety of people far from epicentre did not sensitively predict empathy compared to near population.

Fig. 6. Trait empathy changes across state anxiety moderated by news discriminability

Tables

Table 1 Study sample characteristics (M (SD))

P pre (n = 520) post (n =570) t values

Demographic Age (years) 28.43 (7.84) 25.08 (8.74) 6.64 ***

Sex(female) 53% 64% χ2 = 14.89 ______

IRI

Perspective taking 24.14 (2.97) 23.43 (2.81) 4.06 ***

Trait empathy 25.05 (3.50) 24.56 (3.03) 2.48 *

Fantasy 24.14 (4.68) 24.54 (4.49) 2.02 *

Personal distress 23.53 (4.07) 24.14 (3.50) -2.62 **

AQ 23.47 (3.28) 23.52 (2.86) -0.27 0.79

STAI

State anxiety ______40.31 (10.13) ______

Trait anxiety ______43.89 (8.59) ______

Correct rate 0.51 (0.14) 0.50 (0.12) 9.97 **

Empathy level 3.89 (1.55) 3.60 (1.53) 4.27 ***

Vicarious experience 3.16 (1.50) 2.78 (1.33) 6.20 ***

Prosocial willingness 3.73 (1.67) 3.50 (1.63) 3.16 **

Note: *** P < 0.001, ** P < 0.01, * P < 0.05.

M: mean, SD: standard deviation

Supplementary materials.

Fig.S1. Additional information about participants status.

Fig.S2. Video judgement task protocol. To validate these materials, all recorded 12 video clips (around 50 seconds for each one) were shown, in random order, to the other independent 30 participants (15 men and 15 women, Mage = 22.92), which was not overlapped with dataset1 and dataset2. They were told to judge whether the person in the video is telling truth or lying. There was a 10 second pause between videos, and then the next video was shown. The mean accuracy was 52.7%

(Maccuracy = 52.17%) for all videos. It indicated that people actually cannot accurately identify real or vicarious pain feelings from the video clips. The study was approved by the China Academy of Science ethics committee, and all participants in the investigation gave informed consent to the study.

Supplementary Method

Distance to Wuhan calculation from IP address

Firstly, we extracted the address information in https://ipinfo.io/

Fig.S3. IP extraction. It is notable, that the accuracy of the calculation varies among cities. For example, if an IP belongs to Kunshan, it will show Kunshan instead of Suzhou, but if an IP belongs to Cangnan (a county which belongs to Wenzhou), it will show Wenzhou instead. Additionally, we encountered two phenomena about the data. In some cases, two different records may share the same IP address. Maybe that’s because they were using the same Wi-Fi network at that time, so we cannot neglect them arbitrarily. Another one is that there was no IP address which exists in both pre- and post-dataset, a possible explanation could be the usage of dynamic IP address. As for distance calculation, we assumed that the earth is a uniform sphere with the radius r=6400km. We assumed that there’s point A which latitude and longitude are (x1,y1), and point B which latitude and longitude are (x2,y2). So, the spherical distance between A and B is:

dist=r*arccos(sin(x1)*sin(x2)+cos(x1)*cos(x2)*cos(y1-y2))

Table.S1 Video session data (M ± SD)

p true (n = 1091) faked(n =1018) values

Empathy level 4.72 ± 0.96 2.67 ± 1.34 ***

Vicarious experience 3.74 ± 1.23 2.12 ± 1.11 ***

Prosocial willingness 4.40 ± 1.38 2.76 ± 1.48 ***

Note: *** P < 0.001, ** P < 0.01, * P < 0.05.

M: mean, SD: standard deviation

Fig S4. Video authenticity judgement influences empathy level, vicarious

experience and prosocial willingness 32

Fig.S5. Time moderated empathy level predicting prosocial willingness.

Empathy level significantly predicted prosocial willingness, β = 0.62, t(2096) = 22.03,

P < 0.001. Time does not predict prosocial willingness, t(2096) = -0.16, P = 0.87. The interaction of time and empathy level was significant, which indicated the valid moderation model (Fig.3). Time negatively influenced the pathway that empathy level predicted prosocial willingness, β = -0.083, P = 0.012 < 0.05, 95% CI: [-0.15,-0.02]. In the post-outbreak period, even though people felt the same the pain of others, their willingness to do prosocial willingness (share pain) was decreasing.

Fig.S6. Empathy level changes across trait anxiety in two distance

condition.

Fig. S7. Empathy level significantly predicted prosocial willingness, β = 0.60, t(2096) = 23.96, P < 0.001 (path B). Distance from Wuhan did not predict empathy level, β = -0.03 t(2096) = -0.95, P = 0.34. The interaction of distance from Wuhan and other pain was not significant, t(2096) = -1.58, P = 0.11, 95% 95% CI = [-0.12,0.01], which indicated no moderating effect of distance.

Fig.S8. Pre- and post- outbreak items networks of AQ.

Fig.S9. Pre- and post-outbreak IRI items relationship. a, IRI subscales correlation matrix. b, c, Pre- and post-outbreak item network of IRI.

Fig.S10. The COVID-19 related questions illustration.

Fig.S11. State anxiety correlated post COVID-19 questions. a,b, Among the

18-item COVID-19 questionnaire, two questions were positively correlated with the state anxiety score in STAI, (a)‘When do you think is the turning point of this increasing trend of newly confirmed cases each day?’ (r = 0.15 , P < 0.001).(b)‘The unpredictability of ’(r = 0.10, P < 0.001). The current mental state assessment was assessed —— ‘Your psychological status within the recent ten days:’ (c) anxiety (r =

0.34, P < 0.001), (d) fear (r = 0.33, P < 0.001), (e) anger (r = 0.20, P < 0.001) and (f) happiness (r = -0.35, P < 0.001).

COVID-19 questionnaire

1. Have you physically gone back to work/study? (choose one that applies) - Yes - No 2. The relationship between you and Wuhan (choose one that applies): (1) Myself or my relatives currently living in Wuhan (2) Working and studying in Wuhan (3) Used to live in Wuhan for a long time (4) Have been travelling to Wuhan (before the coronavirus outbreak) (5) Never been to Wuhan 3. Based on hospitals and official department, do people around you have any confirmed or suspected cases of COVID-19? (choose one that applies) (1) Yourself (2) Relatives (3) People you know (4) People in your local community, school, or company (5) No known confirmed cases 4. Rate the following properties regarding the COVID-19: (enter a number from 1 to 10; 1: very low; 10: very high) - Infectiousness - Death rate - Your possibility of getting infected 5. Do you think the supply listed below in Wuhan is sufficient? (enter a number from 1 to 10; 1: severely deficient; 10: very abundant) - medical professionals and medical supply - life supply - psychological support 6. Your psychological status within the recent ten days (enter a number from 1 to 10; 1: very low; 10: very high) - Anxiety - Depression - Fear - Anger - Happiness 7. Amount of time spent daily paying attention to the disease (choose one that applies) (1) less than 10 minutes (2) 11 – 30 minutes

(3) 31 – 60 minutes (4) 61 – 120 minutes (5) more than 120 minutes 8. When do you think is the turning point of this increasing trend of newly confirmed cases each day? (choose one that applies) (1) within 7 days (2) 8 – 14 days (3) 15 – 30 days (4) 31 – 60 days (5) more than 60 days 9. The major pathway you use to get information about the disease: (choose one that applies) - TV news - WeChat moments - WeChat official accounts - Weibo - Tiktok - Toutiao - Alipay - Baidu - Zhihu - Other (please specify) 10. On a scale of 1 to 10, how much do you agree with the following statement (1: completely agree; 7: completely disagree) (1) Patients have been treated in time for COVID-19 (2) Patients with COVID-19 have been socially discriminated during the disease outbreak (3) The coronavirus outbreak will be under complete control in the near future (4) As soon as there is medical prevention (e.g. vaccine) or treatment appearing in the market, I will purchase them immediately (5) I will not publish or repost any information that is not officially confirmed (6) The unpredictability of the coronavirus outbreak has severely affected my life (7) If I cannot get any information or news regarding this coronavirus outbreak, I will feel nervous or frustrated (8) I feel confident in overcoming this disease (9) I can accurately distinguish true and false messages related to this disease

Table.S2 COVID-19 related question in dataset2 (M (SD))

near (n = 357) far (n =213) P values

Patients with COVID-19 have been socially discriminated 4.94 (2.77) 4.46 (2.61) ** during the disease outbreak.

News discriminability: I can accurately distinguish true 5.04 (2.76) 4.54 (2.51) * and false messages related to this disease.

Note: *** P < 0.001, ** P < 0.01, * P < 0.05.