key: cord-0937446-6cn1pmy5
authors: Liu, Jingyu; Long, Anna; Li, Minglong; Bai, Billy
title: Cyber-bystanders’ reactions toward tourism companies’ prejudice practice during the COVID-19 pandemic
date: 2021-12-31
journal: Journal of Hospitality and Tourism Management
DOI: 10.1016/j.jhtm.2021.09.002
sha: 3087d3cfe00953410c45f7135c5e4d9ced5ab924
doc_id: 937446
cord_uid: 6cn1pmy5

Prejudice against tourists from the epidemic areas or those infected during the COVID-19 pandemic has attracted much attention. While many studies examined the influence of this prejudice on tourists themselves, little research has been conducted to identify cyber-bystanders’ reactions to tourism companies' prejudice practice. This study aims to fill this gap by revealing the process of how cyber-bystanders identify, evaluate, and respond online to prejudice practice in the context of the tourism industry. The study developed a conceptual model to examine the influential factors and their internal relationship of cyber-bystanders’ reactions. A multistage model was proposed based on the social comparison theory. An online survey was conducted in mainland China, and 558 useable questionnaires were collected. The three-process model was estimated using the Bootstrap mediation test and hierarchical regression analysis. The results indicate that cyber-bystanders’ prejudice recognizing process could impact their reactions through the influence of their evaluation behaviors. Meanwhile, the collective sentiment on social media has a moderating effect on the relationship between cyber-bystanders’ evaluating process and their reactions to prejudice practice. Both theoretical and practical implications were discussed.

Prejudice is an irrational attitude of hostility directed against individuals (Bobo, 1995) , which has been verified appearing among various stakeholders (local residents, tourists, tourism companies, and destination authorities) and posing justice issues (Jamal & Budke, 2020) . Hsu and Chen (2019) illustrated local residents' prejudice attitude toward tourists based on stigma theory. Tourists may suffer unequal treatment at ports of entry due to their cultural characteristics (Updegrove et al., 2020) , or they may be treated differently at a restaurant by their tipping abilities (Brewster & Nowak, 2019) . The COVID-19 pandemic has been deepening social and economic inequalities and increasing the discrimination phenomena in hospitality and tourism services (Jamal & Higham, 2021) . This study focuses on prejudice practice of tourism companies towards tourists because it is a comprehensive reflection of the business, local residents, and destination authorities.

Under the background of the COVID-19 pandemic, many tourism companies involved prejudice behaviors toward tourists from epidemic areas in service processes, such as refusing reservations, avoiding contact when serving, and providing limited service (Huang & Hao, 2020) . Tourism companies' prejudice practice cannot be ignored due to their potential consequences in the mobile internet era. Tourists who perceived discriminatory service would post their experience on social media with negative comments. For example, at the beginning of the COVID-19 epidemic, many tourists from Wuhan, Hubei, China (the first city where the epidemic broke out) posted their discriminatory experiences by hotels, which include forcing existing Hubei customers to leave the hotel and canceling Hubei customers' reservations. Later, the topic of # discriminate Hubei residents on Weibo had been read 3 billion times with more than 36,000 discussions (Weibo, 2020) . Since these posts draw plenty attention, it forms negative electronic-word-of-mouth (eWOM) to those tourism companies. Therefore, escalation and de-escalation of tourism companies' negative eWOM might be caused by cyber-bystanders' online reactions (such as forwarding, commenting, and giving thumbs up).

Based on cyber-bystanders' online reactions, the cyber-bystanders' role can be sorted by the degree of being supportive towards the victims into three types as reinforcer, observer, and defender (Waches, 2012) . According to Barlinska et al. (2015) , cyber-bystanders' defending reactions could be regarded as post comments to criticize the perpetrator (tourism companies that have prejudice practice during the COVID-19 epidemic period) or forward the source of negative sentiment to the prejudice practice of the perpetrator. Previous studies of cyber-bystanders' defending reactions mainly focused on preventing the occurrence of negative events. However, little research was conducted on cyber-bystanders' defending reactions from an eWOM perspective. Understanding the root of the formation of cyber-bystanders' defending reactions is important for tourism companies to de-escalate the negative eWOM. To explore the underlying mechanism, this paper builds a conceptual model to examine what factors would influence cyber-bystanders' defending reactions to victims and how the impact would spread out.

This research has contributed to the research area on tourism discrimination from various perspectives, providing relevant implications both theoretically and practically. First, as a novel angle, cyberbystanders' reactions have been explored to research tourism discrimination and eWOM. Second, a modified conceptual model on the notion of bystander intervention model has been built to explore the root of the formation of cyber-bystanders' negative eWOM. Third, this study highlights the effect of collective sentiment on cyber-bystanders' reactions to enrich the research bystander intervention model in the context of the Internet.

Based on social comparison theory, people would compare their prepared behavior with others to prevent embarrassing situations, and may lead to social inhibition effects on individual altruistic behavior (Latane & Darley, 1969) . The theory stipulates that bystander to an event must first come to some general interpretation of the situation. Then, based on the interpretation, they would choose whether to implement defending reactions to victims in order to promote social justice. The Bystander Intervention Model (BIM) was developed by Latane and Darley (1970) to explain the bystanders' decision-making process in emergent social and psycho-dynamic situations. Several previous studies used the model to describe cyber-bystanders' reactions (Barlinska et al., 2015; Van Cleemput et al., 2014) . The model posits cyber-bystanders' reactions into five steps: (1) recognizing the event; (2) judging the situation; (3) sensing the responsibility to help; (4) determining the specific action to intervention; and (5) implementing the action. Each step is a prerequisite for the next. Song and Oh (2018) indicated that the first three steps of BIM are critical stages in the decision of implementing a pro-social response (i.e., defending victims) by the cyber-bystanders, and categorized the influencing factors on cyber-bystanders' defending reactions into experiential, situational, and psychological factors to correspond with Steps I, II, and III in the BIM.

Although previous studies demonstrated the relationship between multiple variables and cyber-bystanders' defending reactions, they only examined the one-way effect from each variable to cyber-bystanders' defending reactions. According to the BIM, each step is a sufficient condition for the next step, which means that the impact path should be multistage in nature rather than one phase. To examine the multistage pathways in the BIM, this study modified the impact factors identified by Song and Oh (2018) and proposed three crucial aspects of recognizing the prejudice practice, evaluating the discrimination situation, and perceiving the collective sentiment on social media. The three aspects correspond with the first three steps of the BIM to investigate the relationship among each aspect and cyber-bystanders' defending reactions when they see discrimination occur during COVID-19.

Recognizing tourism companies' different services as prejudice practice would be the pre-condition of cyber-bystanders' defending reactions. In typology, prejudice practice is usually categorized by causes such as gender, age, and race (Yu, 2006) and manifestations such as blatant and subtle (Van Laer & Janssens, 2011) . Many causes will influence tourism companies' attitude toward tourists in the COVID-19 period and further present with different prejudices (e.g., refusing customers, undervaluing customers, and neglecting customers), making it hard for cyber-bystanders to identify the prejudice type accurately. Since prejudice practice in the tourism context is hard to measure in a standardized way, this study measured tourism companies' prejudice practice from the cyber-bystanders' subjective perspective, in other words, perceived prejudice (Ye et al., 2012; Brewster, 2013) . Recent studies that discussed cyber-bystanders' reactions to their recognition of the prejudice focused on the experiences of witnesses (Bauman et al., 2020) and perpetrator or victims (Barlinska et al., 2013; Song & Oh, 2018) , empathy (Balakrishnan & Fernandez, 2018) , intervention education (Evans et al., 2019) , and contextual information (Troopgordon et al., 2019; Luo & Bussey, 2019) . Considering the characteristic of tourism companies' prejudice practice, this study would mainly discuss the contextual information, empathy, and victimization experience in cyber-bystanders' recognizing process.

Contextual information would be the crucial factor for cyberbystanders to recognize tourism companies' prejudice practice and further identify it as tourism discrimination or not. Existing research mentioned that the contextual information might introduce bias into cyber-bystanders' evaluation and further influence their decision-making process (Neil & Marika, 2020) . Contextual information that involves different values and policies should be regarded as an important factor in recognizing the fair 'procedure' of such prejudice practice (Rastegar et al., 2021) . Therefore, the contextual information of tourism companies' prejudice practice might influence cyber-bystanders' evaluation process. Specifically, it would influence cyber-bystanders' perceived severity of tourism companies' prejudice practice.

Empathy, as an affective response, is usually being associated with cyber-bystanders' pro-social behaviors. The role of empathy in cyberbystanders' defending reactions has been demonstrated in various studies (Balakrishnan & Fernandez, 2018) . Inclusiveness and recognition have been acknowledged as critical priorities for promoting 'just' tourism (Jamal & Higham, 2021) . As the crucial condition of sharing and understanding tourists' emotional and mental states of suffering unfair treatment, it could strongly influence cyber-bystanders' recognition of tourism companies' prejudice practice. Previous studies indicated that lower empathy could predict negative reactions of cyber-bystanders, and it could be regarded as the main reason to explain the lack of defending reactions. To discuss the impact of cyber-bystanders' reactions, Brody et al. (2016) stated that the Internet's hyperspace characteristic might suppress the emotional transmission, which might weaken the effect of empathy on cyber-bystanders' defending reactions. This finding was supported by Barlinska et al. (2013) that affective empathy has no significant impact on bystander's intervention reactions. Still, cognitive empathy has a negative effect on cyber-bystanders' intervention behaviors. Compared with affective empathy, cognitive empathy is more related to the contextual information, which might influence cyber-bystanders' perceived severity of tourism companies' prejudice practice and further affect cyber-bystanders' reactions.

Previous studies illustrated that cyber-bystanders who had victimization experiences might become more sensitive when witnessing prejudice practice, which performed more easily to recognize behaviors as prejudice behaviors and further influence their perceived severity of such behaviors. Existing research showed that past experience of victims had a significant impact on cyber-bystanders to implement defending reactions. Cyber-bystanders who had the victimization experience were more likely to react positively Allison & Bussey, 2017) . On the contrary, some studies found that victims' past experiences had no relationship with cyber-bystanders' defending reactions ( Barlinska et al., 2013; Song & Oh, 2018) . However, few studies explored the relationship among cyber-bystanders' past experience of victims, the perceived severity of prejudice practice, and their defending reactions.

Overall, the existing literature indicated that past victimization experience, contextual information, and empathy could predict cyberbystanders' defending reactions or their perceived severity of tourism companies' prejudice practice. Based on the BIM, recognizing the prejudice practice should be a sufficient condition for cyber-bystanders to evaluate the discrimination situation. Judging the severity of prejudice practice as a part of evaluating process might play an intermediary role of recognizing prejudice practice on implementing reactions. Therefore, this study will examine the direct effects of cyber-bystanders' defending reactions and the indirect effects through cyber-bystanders' perceived severity of tourism companies' prejudice behavior. The following hypotheses were proposed.

Cyber-bystanders' recognition of tourism companies' prejudice practicecyber-bystanders' empathy level (a), cyber-bystanders' knowledge of contextual information (b), and cyber-bystander's victimization experience (c) -will positively impact their defending reactions to victims.

Cyber-bystanders' evaluation of the discrimination situation will have a mediation effect between their empathy level (a), knowledge of contextual information (b), victimization experience (c), and defending reactions to victims.

Based on the decision-making process, an evaluating process is required before moving ahead to the decision-making phase. After fully understanding the prejudice practice of tourism companies during COVID-19, cyber-bystanders could assess the discrimination situation by defining the nature of the event (Ybarra et al., 2012) , weighing their control ability (Tatebe et al., 2019) , clarifying the relationship with the victims or the perpetrator , and judging the severity of the behaviors (Dillon & Bushman, 2015) . Moral disengagement as a psychological factor is always discussed with cyber-bystanders' defending reactions (DeSmet et al., 2018) . Previous studies usually measured the level of moral disengagement among cyber-bystanders associated with their sense of responsibility. Hultsman (1995) proposed the concept of 'just tourism,' indicating that some generally understood and basic ethicality is the foundation to develop tourism. This concept counseled tourism companies to act in a manner that reflects ethicality, which means tourism companies' policies are expected to be created equal instead of prejudiced. Therefore, this study suggests that individual levels of moral identity might have a greater impact on assessing discrimination situations by cyber-bystanders than their sense of responsibility to defend victims. Moral justification has been regarded as the crucial aspect of moral disengagement measurement. Cyber-bystanders who have strong moral reasons are expected as low moral disengagement. Numerous studies verified that cyber-bystanders would have more defending reactions when their moral disengagement was low because they would not escape their moral responsibility by excuses (diffuse or displace responsibility, attribute to others, and distort consequences) (Luo & Bussey, 2019) . Song and Oh (2018) highlighted that the perceived severity of the prejudice practice was a critical situational element to influence cyber-bystanders' defending reactions. The defending reactions of cyber-bystanders were connected closely to the perceived severity of the prejudice practice. When they perceived higher severity, they tend to respond positively. The interaction effect with their sense of responsibility might influence their willingness to defend the victims (Obermaier et al., 2016) . Due to the unique hyperspace structure of the Internet, cyber-bystanders could not react immediately as traditional bystanders. Instead, before cyber-bystanders decide to respond positively, they may know more contextual information of tourism companies' prejudice practice through social networks and dig into the reasons behind it.

The relationship with the victim or the perpetrator has been regarded as a major factor affecting cyber-bystanders to react pro-socially. When cyber-bystanders have a close relationship with the victims, they are more willing to defend on their behalf (Forsberg et al., 2014) . Intimacy with the perpetrator might suppress cyber-bystanders' defending reactions (Machackova, 2015) . Based on social identity theory, group bias would influence cyber-bystanders' defending reactions. Cyber-bystanders are more likely to help their 'inner group' than their 'outer group' members (Levine & Crowther, 2008) . From a behavioral economic perspective, Hayashi and Tahmasbi (2020) examined the social distance with the victims from a social discounting aspect and stated that the subjective value of cyber-bystanders' defending reactions would be discounted as a closeness function to them.

Previous research also showed that when cyber-bystanders had neither a close relationship with the victims nor with the perpetrators, bystanders' popularity or social status could predict their defending reactions . In social networks, cyber-bystanders' popularity is usually expressed by the number of subscribers. More subscribers mean higher popularity in social media. Considering the information diffusion model in social media, the poster with higher popularity would diffuse the information wider and faster. In recent studies of cyber-bystanders' reactions, the number of subscribers is usually used to measure cyber-bystanders' perceived control ability on social media (Stieglitz & Linh, 2013) . Unlike the traditional bystanders, cyber-bystanders could implement defending reactions without taking any extra responsibility. It is notable that observers who have no online reactions are regarded as potential defending reactions because this could avoid information diffusion (DeSmet et al., 2018) .

To investigate whether cyber-bystanders' evaluating process of the discrimination situation would influence their defending reactions, this study will examine the direct effects of cyber-bystanders' moral disengagement, perceived control ability on social media, perceived severity of prejudice practice, and relationship with the victims to their defending reactions to victims. Therefore, we proposed the following hypothesis.

Cyber-bystanders' evaluation -cyber-bystanders' perceived severity (a), cyber-bystanders' perceived control ability (b), cyberbystanders' relationship with victim (c), and cyber-bystanders' moral disengagement levelcyber-bystanders' relationship with victim (d)will positively impact their defending reactions to victims.

Unlike the usual decision-making process, the BIM emphasizes that bystanders' sense of responsibility may impact their reactions after evaluating the discrimination situation. The dual-process theory of moral judgment indicates that both affective and cognitive aspects can affect the final moral judgment. Different scenarios convey different messages to influence an individual's mentation and their moral judgment (Greene, 2007) . Many studies showed that personal moral judgment is influenced not only by subjective factors such as personal experience, cognitive level, and psychological characteristics, but also by public opinions and attitudes. Buzinski and Kitchens (2017) proposed that public opinions could change individuals' attitudes to a given event by affecting their psychological state and moral judgment. In pro-social scenarios, bystanders are more likely to defend victims.

In recent years, social media has experienced tremendous growth in user base and has become one of the most popular ways for communication with various options such as information sharing and instant messages (Stieglitz & Linh, 2013) . Online public opinion has gradually become an indispensable part of social opinion that influences cyber-bystanders' behaviors on social media. The content of social media often conveys not only the information but also the posters' emotional state with a personal attitude. The judgment of a certain topic or the intended emotional communication has been referred to as sentiment. Unlike public opinion, collective sentiment on social media has a clear emotional state, which might strengthen the impact of public opinions on cyber-bystanders' defending reactions. Previous research also indicated that an online communication environment is more likely to form a herd effect of sentiment. As the collective sentiment grew stronger, the cyber-bystanders were more likely to forward the original micro-blog (Stieglitz & Linh, 2013) .

Previous researchers devoted themselves to studying the impact of collective sentiment on social media between information diffusion and cyber-bystanders' reactions. They demonstrated that the collective sentiment on social media could affect cyber-bystanders' emotions and influence their online behaviors. However, there is rarely a consideration of the possibility of multivariate relationships of cyber-bystanders' defending reactions, evaluating the process of the situation and the collective sentiment of the online scenario. Concerning the decisionmaking process of cyber-bystanders has been intertwined with many factors, this study aims to examine not only the direct effect of collective sentiment but also the interaction effect to fully understand cyber-bystanders' defending reactions. Therefore, the following hypotheses were proposed.

H4. Collective sentiment on social media will positively impact cyberbystanders' defending reactions to victims.

Collective sentiment on social media will have a moderation effect between cyber-bystanders' evaluation of the discrimination situationperceived severity (a), perceived control ability (b), relationship with victim (c), moral disengagement level (d) -and their defending reactions to victims.

A conceptual model is proposed as follows (Fig. 1 ).

Before the main collection, a pilot test was conducted with 98 participants. Based on the feedback, some items that cannot explain the construct well (e.g., 'Some people deserve to be treated like animals' with the standardized factor loading being less than 0.5) were deleted in the revised questionnaire. In the main collection, an online-based survey was conducted through www.wjx.cn, an online crowd-sourcing platform in Mainland China. To determine the appropriate sample size, a power analysis was conducted. The formula of Tabachnick and Fidel (2001) was used, and the result shows that the minimum sample size should be 138 to achieve the power of 80%, which is smaller than the sample size in this study (558).

To ensure the population of this study was composed of adults who had seen the news related to tourism companies' prejudice practice on social media, participants were asked to answer a screening question at the beginning of the survey: 'Have you ever seen the related posts of tourism companies' prejudice practice on social media?' In order to eliminate participants' subjective bias on identifying tourism companies' prejudice practice, the survey instrument adopted a specific trending topic #Hotel rejects Hubei tourists to check-in to ensure the consistency of measurement of companies' prejudice practice. A total of 605 online questionnaires were collected after the peak of the COVID-19 pandemic in China from April 20 to April 30, 2020. Excluding 47 participants who reported that they had not seen the news related to tourism companies' prejudice practice on social media, the data of the remaining 558 participants were analyzed. Regarding the characteristics of the respondents, the sample distribution of this study is similar with Chinese internet users that was reported by China Internet Network Information Center (China Internet Network Information Center, 2020) (See Table 1 ).

To identify cyber-bystanders' reactions, a modified measurement scale based on the Participant Role Scale (Song & Oh, 2018) was used. The measurement categorized each cyber-bystander into one of three types (observer, defender, and reinforcer), which consisted of the items 'I forwarded or gave thumbs up the news with positive comments to the tourist (defender),' 'I watched the situation without taking any actions (observer),' 'I forwarded or gave thumbs up the news with positive comments to the tourism companies (reinforcer),' 'I forwarded or gave thumbs up the news with no comments (reinforcer).' Participants were asked to choose the closest behavioral descriptor for their online reactions.

Moral disengagement was measured through the Mechanism of Moral Disengagement Scale (MMDS), which was developed by Bandura et al. (1996) . It includes eight categories: moral justification, euphemistic labeling, advantageous comparison, disregard or distortion of consequences, displacement of responsibility, diffusion of responsibility, dehumanization, and attribution of blame. Due to the information complexity on social media, cyber-bystanders might easily ignore the consequences of tourism companies' prejudice practice and also be harder to recognize the 'euphemistic labeling' in MMDS. The sample items such as 'Teasing someone does not really hurt them' were selected into the modified questionnaire. Considering that Chinese people prefer to be neutral (responding with a mid-point of 'no option'), which may undermine the quality of survey data (Ares, 2018; Lee et al., 2002; Wong et al., 2011) , the items were measured by a four-point Likert scale ranging from 1 (strongly disagree) to 4 (strongly agree) instead of the original five-point Likert scale. The internal reliability (Cronbach's α) of moral disengagement for this study was 0.840.

To measure empathy, this study utilized the empathy scale (E-scale) of Leibetseder et al. (2007) . The E-scale is divided into four parts, includ ing emotional-sensitivity, cognitive-sensitivity, emotional-concern, and cognitive concern. The dimension of 'sensitivity' represents the empathy with fictitious situations, and the 'concern' refers the empathy with real-life situations. Given the focus of this study on cyber-bystanders' empathy, questions were selected based on the 'sensitivity' dimension. Sample items as 'I think it is exaggerated to get completely wrapped up in a book or movie' (emotional sensitivity) and 'If I see a movie, I often try to imagine how I would feel in the person's place' (cognitive sensitivity) were asked to measure personal perception of empathy. The answers were also measured using a four-point Likert scale. The Cronbach's α values obtained in this study was 0.878. Table 2 presents the results of the reliability test of moral disengagement and empathy. All the fit indexes are within the acceptable range (Cronbach's α > 0.7, CR > 0.7, and AVE >0.5).

Data was analyzed by using SPSS 25.0 and Amos 24.0. Results showed that the cyber-bystanders with no online reactions were the majority group (289 participants, 51.8%), followed by the cyberbystanders who had defending reactions such as forwarding or giving thumbs up with positive comments to tourists (187 participants, 33.5%), while the rest were cyber-bystanders who forwarded or gave thumbs up without any comments (82 participants, 14.7%). No cyber-bystanders were found to forward or give thumbs up with positive comments to tourism companies. A Chi-square test examined the difference of cyberbystanders' defending reactions in gender, age, and educational level. The results showed that there were significant differences between cyber-bystanders' gender (χ 2 = 22.756, p = 0.000, Cramer's V = 0.202), age (χ 2 = 82.739, p = 0.000, Cramer's V = 0.272), and educational level (χ 2 = 35.135, p = 0.000, Cramer's V = 0.177) on their defending reactions to tourism companies' prejudice practice. Therefore, gender, age, and educational level were entered as control variables in the subsequent regression model. Table 3 showed the relationship between cyber-bystanders' defending reactions and the independent variables. cyber-bystanders' defending reactions to victims showed a significant positive correlation with empathy (r = 0.104, p = 0.014), contextual information (r = 0.127, p = 0.003), perceived control ability (r = 0.111, p = 0.009) and perceived severity (r = 0.261, p < 0.001). In other words, cyberbystanders with high empathy, more contextual information knowledge, stronger perceived control ability, and higher perceived severity were more likely to defend victims who suffered from tourism companies' prejudice practice. cyber-bystanders' perceived severity of the tourism companies' prejudice practice showed a strong positive correlation with their knowledge of contextual information (r = 0.859, p < 0.001). Similar results have been illustrated in previous studies Ye et al., 2012) . This means that when cyber-bystanders have more knowledge of the contextual information, cyber-bystanders tend to perceive a higher severity of tourism companies' prejudice practice.

A bootstrap estimate was used to examine the mediation effect of cyber-bystanders' perceived severity of tourism companies' prejudice practice. Comparing with contextual information (β = 0.978, p < 0.001), the victimization experience (β = 0.029, p = 0.027) and empathy (β = 0.023, p = 0.004) have a relatively low effect on cyberbystanders' perceived severity, which may possibly be suppressed in terms of transmissions due to Internet's hyperspace characteristic (Brody et al., 2016) . Specifically, since cyber-bystanders are not on-site to participate in the event, the emotional transmission may be weakened. In the further analysis of mediating effect, cyber-bystanders' perceived severity has a significant indirect effect between their empathy level (95%CI: 0.0056, 0.0352), knowledge of contextual information (95%CI: 0.3457, 0.6636), victimization experience (95%CI: 0.009, 0.0799), and defending reactions to victims (see Table 4 ). Thus, H2a, H2b, and H2c were supported. The result demonstrated that cyber-bystanders' perceived severity of tourism companies' prejudice practice acts as a completed mediation in both the relationship between the empathy level to their defending experience and victimization experience to their defending experience. Cyber-bystanders' knowledge of contextual information has a negative impact on their defending reactions, and their perceived severity has a suppression effect. Hierarchical regression analysis was employed to test H1, H3, H4, and H5 (See Table 5 ). In the first step of the hierarchical optimal scaling regression model, eight independent variables with three control variables were entered. These variables explained 35.2% of cyber-bystanders' defending reactions. Contextual information (β = − 0.597, p < 0.001), moral disengagement (β = − 0.087, p = 0.002), perceived severity (β = 0.848, p < 0.001), perceived control ability (β = 0.127, p < 0.001), and collective sentiment (β = 0.208, p < 0.001) had a significant effect on cyber-bystanders' defending reactions. Thus, H1b, H3a, H3b, H3d, and H4 were supported. It implied that cyber-bystanders with low moral disengagement, high perceived severity, strong perceived control ability, and those who perceived the collective defending sentiment were more likely to react pro-socially. When interaction variables were entered in the second step, the collective sentiment showed a significant moderating effect with cyber-bystanders' relationship with the victim (β = − 0.111, p = 0.002) and their personal moral disengagement level (β = − 0.097, p = 0.038). Thus, H5c and H5d were supported. These results indicate that the collective sentiment on social media may weaken the impact of personal moral disengagement level and the relationship with victims to cyber-bystanders' defending reactions. Therefore, H5c and H5d were supported. Note: *p < 0.05, **p < 0.01, CS= Collective sentiment; RV = Relationship with victims; PS= Perceived severity; PC= Perceived control ability; MD = Moral disengagement.

As showed in Fig. 2 , cyber-bystanders with low moral disengagement levels were more sensitive to the collective sentiment on social media, which means cyber-bystanders with low moral disengagement levels tended to implement defending reactions to victims when the collective sentiment on social media was strong. On the other hand, cyberbystanders who had an alienated relationship with the victim were more likely to react pro-socially when the collective sentiment on social media was strong. However, cyber-bystanders who had a close relationship with the victim presented a very different tendency (See Fig. 3) .

The study developed a multistage model to illustrate what factors would influence cyber-bystanders' defending reactions to victims when they saw tourism companies' prejudice behavior through social media and how the impact would spread out. Based on the results, the degree of moral disengagement, perceived control ability, and perceived severity were identified as significant individual factors, and the contextual information and collective sentiment on social media as significant external factors to influence cyber-bystanders' defending reactions. The mediation and moderation effects were significant in revealing the relationship among cyber-bystanders' recognition of tourism companies' prejudice practice, evaluation of the discrimination situation, perceived collective sentiment on social media, and their defending reactions to victims.

The results of internal impact factors of cyber-bystanders' defending reactions showed that lower moral disengagement, stronger perceived control ability, and higher perceived severity could significantly predict their defending reactions. Cyber-bystanders with low moral disengagement would not disregard or distort the consequences of tourism companies' prejudice practice. From a justice perspective, they tend to assess tourism companies' prejudice practice during COVID-19 as defenders required issue; thereby, they take more defending reactions. Cyber-bystanders' perceived control ability could predict their defending reactions because once cyber-bystanders' perceive themselves are having strong control ability of the situation, they would have more confidence to influence the developments (Thornberg et al., 2017) . In other words, cyber-bystanders prefer to implement defending reactions when they believe their reactions could achieve a desirable outcome. From a social support perspective, when cyber-bystanders consider the prejudice practice more severely, they believe that they will receive higher social rewards once they defend for victims. Therefore, cyber-bystanders were more likely to defend victims when they perceived tourism companies' prejudice practice more severe, and this finding is consistent with Bastiaensens et al. (2014) . Comparing with previous studies that examined Chinese cyber-bystanders' behaviors to negative events (Han, 2016; Xin & Lai, 2015) , the findings of cyber-bystanders' defending reactions are different. Specifically, the percentage of cyber-bystanders' supporting comments to victims in the case of tourism companies' prejudice practice during COVID-19 is 33.5%, which is lower than other kinds of negative events (70% supporting comments to victims) (Huang & Kang, 2019) . This difference may be explained by social identity theory that cyber-bystanders' reaction is consistent with their perceived group identity (Barlinska, 2013) . General negative events discussed on social media are relatively far away from cyber-bystanders' daily lives than the current COVID-19 pandemic. Therefore, cyber-bystanders always regard Fig. 2 . Difference in cyber-bystanders' reactions between their high and low moral disengagement depending on collective sentiment. Fig. 3 . Difference in cyber-bystanders' reactions between their close and alienated relationship with victim depending on collective sentiment. themselves as 'outer group' in general negative events and easily make irrational judgments based on their own emotions and exhibit extreme online behaviors (Huang & Kang, 2019) . However, cyber-bystanders' perceived severity of tourism companies' prejudice practice is not only related to the severity of prejudice practice but also the range of victimization (e.g., the number of the victims). From a social identity perspective, once the diffusion range increases, the relationship between cyber-bystanders and victims may change from 'outer group' to 'inner group.' In this case, no one can be certain that they will not become another victim or perpetrator with the development of the COVID-19 epidemic. Bastiaensens et al. (2015) mentioned that bystanders would be more likely to defend themselves when victims belong to their inner group. Besides, cyber-bystanders may change their social identities from cyber-bystanders to potential victims, which leads them to react positively to protect their own benefits.

In terms of external impact factors, contextual information and collective sentiment on social media significantly influenced cyber-bystanders' defending reactions. When cyber-bystanders have more knowledge of the contextual information, they tend not to defend the victims. From the perspective of human nature, McGregor's theory-X stipulates that people are naturally indolent and reluctant to take responsibility (McGregor & Cutcher-Gershenfeld, 1960) . Different from the effect of contextual information knowledge, stronger collective sentiment towards victims on social media could significantly predict cyber-bystanders' defending reactions. This finding is consistent with the previous studies that cyber-bystander might neglect its own standpoint to obey the collective standpoint due to the herd effect (Mo & Song, 2020; Zhang & Jiang, 2021) . Through the analysis of cyber-bystanders' behaviors with group polarization, Dong (2014) found that Chinese cyber-bystanders would be more likely to comply with the behaviors of other cyber-bystanders because of the complexity of the communication on social media.

The study results also suggest that cyber-bystanders' perceived severity of tourism companies' prejudice practice fully mediated the relationship from personal empathy and victimization experience to cyber-bystanders' defending reactions. This finding is similar to the previous studies that empathy and victimization experience can influence cyber-bystanders' defending reactions (Song & Oh, 2018; Barlinska et al., 2013) . In addition, unlike previous research, our finding further explains the indirect impact of cyber-bystanders' perceived severity as the underlying mechanism that has not been discussed before. Specifically, the high empathy or victimized experience will not directly influence cyber-bystanders' willingness to take defending reactions but can increase their desire to defend for victims through perceived severity. It is worth noting that cyber-bystanders' perceived severity has a suppression effect instead of a mediation effect between their contextual information knowledge and defending reactions. It can be plausibly argued that when cyber-bystanders have more knowledge of the contextual information, they perceive tourism companies' prejudice behavior more severe. As a consequence, the odds of non-defending reactions decreased . Therefore, cyber-bystanders' perceived severity of tourism companies' prejudice practice would suppress the effect of their contextual information knowledge on their defending reactions. It is worth noting that cyber-bystanders' perceived severity has a suppression effect instead of a mediation effect between their contextual information knowledge and defending reactions. It can be plausibly argued that when cyber-bystanders have more knowledge of the contextual information, they perceive tourism companies' prejudice behavior more severe. As a consequence, the odds of non-defending reactions decreased . Therefore, cyber-bystanders' perceived severity of tourism companies' prejudice practice would suppress the effect of their contextual information knowledge on their defending reactions.

In particular, this study found a moderation effect between cyberbystanders' evaluation process of tourism companies' prejudice behavior and their perceived collective sentiment on social media.

Cyber-bystanders with low moral disengagement were more sensitive to the collective sentiment on social media. This finding is consistent with the arousal/cost-reward model. Cyber-bystanders would compare their cost and benefit of their reactions before they determine to react or not (Song & Oh, 2018) . Social reward, as a benefit for implementing moral behavior, is constantly chased by individuals with low moral disengagement (Vozzola, 2014) . Consequently, cyber-bystanders with low moral disengagement showed higher defending reactions when the collective sentiment on social media was strong. Cyber-bystanders who had alienated relationships with victims were more likely to implement defending reactions. There was a significant difference when they were aware of the collective sentiment on social media. These results are different from the previous studies . It might be because cyber-bystanders who have close relationships with the victims can privately defend them instead of speaking up on social media. Therefore, cyber-bystanders who have close relationships with the victims are less likely to support online.

This study contributes to the hospitality and tourism literature theoretically in two main ways. First, it adds to the literature of tourism discrimination and eWOM by looking at a novel angle -the cyber-bystanders' reactions. Past eWOM and tourism discrimination studies in hospitality and tourism focused on customers' perceptions of discrimination (Ye et al., 2012) , customers' satisfaction (Luoh & Tsaur, 2011; Madera et al., 2020) , performances (Brewster & Rusche, 2017) , and service failures (Min & Kim, 2019) . While previous research explored people's intrinsic perspective within the situation, this study adopted an extrinsic perspective. It expanded the research scope beyond the direct parties in discrimination situations and empirically examined cyber-bystanders' perceptions.

Moreover, the study also introduced the BIM, a substantially validated model from educational psychology, to the hospitality and tourism literature. Through building a modified conceptual model on the notion of BIM, the current study delves into the root of the formation of cyber-bystanders' negative eWOM. Therefore, it gives eWOM research a new and robust theoretical support. Additionally, this study investigated how online public opinion influences cyber-bystanders' reactions. It demonstrated that cyber-bystanders' perceived collective sentiment on social media towards victims exacerbate their defending reactions to victims. This discrepancy of cyber-bystanders' reactions to tourism companies' prejudice practice provides significant research implications in assessing the eWOM of tourism companies. Depending on the collective sentiment on social media (negative or positive), the outcome may vary due to a herd effect. For example, a recent study by Sann et al. (2020) posits that guests' online complaining behaviors are influenced by cultural background and that Asian guests are more likely to complain about services. Although they introduce various control variables, their results might have been affected by guest perceived collective sentiment on social media.

Prejudice practice of tourism companies should be dealt with. Although the behaviors of tourism companies may not be intentionally biased against customers, this phenomenon is not likely to disappear in the foreseeable future, as the COVID-19 epidemic has now become a global pandemic. The study offers important practical recommendations to tourism companies from eliminating the perceived prejudice practice and reducing the negative impact of eWOM. Unlike the other kinds of prejudice practice related to personal characteristics, tourism companies' prejudice practice during COVID-19 are more related to isolation policies. To better combat the pandemic, the isolation of tourists from highinfected areas is not only related to the management of tourism companies but more rely on the policies of local governments or the advice of WHO. According to the procedural justice theory (Thibaut & Walker, 1975) , even if the outcome is unfavorable, they would not perceive it as prejudice if the decision-making process is fair. Considering that local governments and WHO are believed to be authoritative neutral institutions, tourists may perceive tourism companies' decision-making process as objective, unbiased and fair if their behaviors are consistent with local governments and WHO. Therefore, tourism companies are encouraged to provide more contextual information about their isolation policies toward tourists from high-infected areas to demonstrate that their behaviors are procedurally fair and further eliminate the perceived prejudice, including the information about the virus, the infected areas, the recommendations by WHO, the policies by the local government, the process how the company reached the management measures and the benefits the customers can obtain, and so on. For example, companies can publish their isolation policies on social media with the guidelines from WHO or local governments' epidemic prevention.

Collective sentiment on social media can significantly influence cyberbystanders' reactions then further influence tourism companies' eWOM. Thus, tourism companies are encouraged to establish online review monitoring mechanisms to recognize tourists' perceived prejudice practice. For example, a regular related word check (e.g., prejudice, discrimination) on social media can be implemented. Once recognizing these comments, tourism companies should provide timely feedback and investigate the reasons for such occurrences. Follow-up reports on the investigation should be posted on social media to avoid the formation of collective negative sentiment, which may escalate the negative impact of eWOM.

Since no standard measurement of tourism companies' prejudice practice, cyber-bystanders can only define the severity of prejudice practice based on their subjective perceptions, regardless of the objective reasons of the behaviors. A justice framework proposed by an authoritative neutral institution can provide a relative object criterion for tourists and cyber-bystanders to evaluate whether tourism companies' behaviors are prejudiced while guiding and constraining tourism companies' justice behaviors. UNWTO (The World Tourism Organization) and WHO, as a worldwide authoritative neutral organization, are suggested establishing a sustainable and justice framework to guide tourism companies' behaviors amid and after the COVID-19 pandemic. Based on cyber-bystanders' reactions in this study, the framework should cover at least three aspects: recognizing the victims, evaluating the situation through due process, and implementing responsibility and recovery actions. The justice framework proposed by Rastegar et al. (2021) is recommended as a reference. Considering the differences between risk of COVID and cognition of prejudice, it needs to be further refined culturally and geographically.

This study is not free from limitations. First, the collected data was cross-sectional and subjective in nature (self-reports). Other research designs should be implemented in future research. For example, an experimental design may help to present various scenarios of prejudice practice. Participants can be allocated to different stimuli randomly so that the results can be more objective and valid. Also, longitudinal data with objective measures such as tourism companies' service quality can be collected to substantiate the study findings.

Second, this study was conducted in Chinese hotel-based scenarios. Different from hotels, prejudice practice in tourism transportation are related to other customers who travel together. For example, it was reported that passengers refused to board the same flight with Wuhan residents in Nagoya and forced China Southern Airlines to alter their tickets (Andrew, 2020) . Therefore, future research should continue to explore how cyber-bystanders react to tourism companies' prejudice practice in more diverse tourism-related scenarios. Since the participants in this study were from Mainland China, more studies based on samples from databases in other locations (e.g., UK) are needed for comparison, considering the potential influence of cultural differences.

Next, other factors may cause tourism companies' prejudice practice (e.g., cultural differences and sexism), which were not included in this study. For example, is the fear of COVID-19 the main cause of tourism companies' prejudice practice during the COVID-19 pandemic or just a trigger? More generally, future studies should continue to advance the understanding of the cause of tourism companies' prejudice practice.

Cyber-bystanding in context: A review of the literature on witnesses' responses to cyberbullying

China Southern Passengers in Japan refuse to fly with Wuhan residents. Simple Flying

Methodological issues in cross-cultural sensory and consumer research

Self-esteem, empathy and their impacts on cyberbullying among young adults

Mechanisms of moral disengagement in the exercise of moral agency

Cyberbullying among adolescent bystanders: Role of the communication medium, form of violence, and empathy

The role of short-and long-term cognitive empathy activation in preventing cyberbystander reinforcing cyberbullying behavior

Cyberbullying on social network sites. An experimental study into bystanders' behavioural intentions to help the victim or reinforce the bully

Experiences of adolescent witnesses to peer victimization: The bystander effect

Surveying racial discrimination: Analyses from a multiethnic labor britain

The effects of restaurant servers' perceptions of customers' tipping behaviors on service discrimination

Racial prejudices, racialized workplaces, and restaurant servers' hyperbolic perceptions of black-white tipping differences

The effects of racialized workplace discourse on race-based service in full-service restaurants

Bystander intervention in cyberbullying

Self-regulation and social pressure reduce prejudiced responding and increase the motivation to be non-prejudiced

The 46 th China statistical report on internet development

The efficacy of the friendly attac serious digital game to promote prosocial bystander behavior in cyberbullying among young adolescents: A cluster-randomized controlled trial

Determinants of self-reported bystander behavior in cyberbullying incidents amongst adolescents

Unresponsive or un-noticed? Cyberbystander intervention in an experimental cyberbullying context

The causes and countermeasures of the polarization of internet groups from the perspective of new media

Examining the efficacy of bystander sexual violence interventions for first-year college students: A systematic review

Bystanders to bullying: Fourth-to seventh-grade students' perspectives on their reactions

Why are vmpfc patients more utilitarian? A dual-process theory of moral judgment explains

Decision-making process underlying bystanders' helping cyberbullying victims: A behavioral economic analysis of role of social discounting

Resident attribution and tourist stereotypes

The dilemma hotel industry and the Hubei nationality who was refused to check in

Characteristics and mechanism of group polarization in the era of mobile internet: A content analysis based on media texts and netizen comments on Jiang Ge case

Just tourism: An ethical framework

Tourism in a world with pandemics: Local-global responsibility and action

Justice and ethics: Towards a new platform for tourism and sustainability

The unresponsive bystander: Why doesn't he help

Cultural differences in responses to a Likert scale

Structural analysis of the E-scale

The responsive bystander: How social group membership and group size can encourage as well as inhibit bystander intervention

The selectivity of moral disengagement in defenders of cyberbullying: Contextual moral disengagement

Customers' perceptions of service quality: Do servers' age stereotypes matter?

Brief report: The bystander effect in cyberbullying incidents

Customer service evaluations of employees with disabilities: The roles of perceived competence and service failure

The human side of enterprise

When service failure is interpreted as discrimination: Emotion, power, and voice

Moral disengagement and identity transformation: A study of bystanders in social network bullying

Bystanding or standing by? How the number of bystanders affects the intention to intervene in cyberbullying

COVID-19 and a justice framework to guide tourism recovery

Online complaining behavior: Does cultural background and hotel class matter?

Factors influencing bystanders' behavioral reactions in cyberbullying situations

Emotions and information diffusion in social mediasentiment of microblogs and sharing behavior

Using multivariate statistics

Empowering bystanders to intervene: Trauma responders unify to empower (TRUE) communities

Procedural justice: A psychological analysis

Classroom relationship qualities and social-cognitive correlates of defending and passive bystanding in school bullying in Sweden: A multilevel analysis

Predicting the development of pro-bullying bystander behavior: A short-term longitudinal analysis

If they notice I'm Mexican": Narratives of perceived discrimination from individuals who crossed the US-Mexico border at ports of entry

Personal characteristics and contextual factors that determine "helping

Ethnic minority professionals' experiences with subtle discrimination in the workplace

Moral development: Theory and applications

Moral disengagement and emotional and social difficulties in bullying and cyberbullying: Differences by participant role. Emotional & Behavioural Difficulties

Discriminate Hubei residents

Differences between odd number and even number response formats: Evidence from mainland Chinese respondents

Perceptions of middle school youth about school bullying

Perceived discrimination in the context of high and low interactions -evidence from medical and general tourists

Social psychology

Research on the dynamic evolution model of public opinion communication in mobile social network integrated with user behaviour