Tracing the Roots of Online Hate Speech: Exploring Outgroup Prejudice

Contact Theory and Prejudice

First, contact theory suggests that direct and personal contact between members of different groups can weaken prejudice and promote more accurate perceptions of out-groups. This theory, originally proposed by Gordon Allport in 1954, has been supported by numerous studies over the decades (Cheong et al., 2011; Silverstein & Flamenbaum, 1989). According to contact theory, increased interpersonal contact is believed to reduce social distance—the emotional and psychological gap between different groups. This reduction in social distance can help mitigate the effects of negative media stereotypes, which often portray out-groups in a biased or unfavorable light (Roh, 2013; Silverstein & Flamenbaum, 1989).

For instance, when individuals have direct, personal interactions with members of an out-group, they are more likely to see them as individuals rather than as stereotypical representatives of a group. This personalization process helps in breaking down negative stereotypes and building more nuanced and accurate perceptions. Moreover, such contact can foster empathy and understanding, further reducing prejudicial attitudes. Conversely, a lack of personal contact may result in a shortage of information and experience about the other group. This lack of direct knowledge can lead to the formation of inaccurate attitudes and reinforce existing negative stereotypes (Tan et al., 1997).

The media often plays a significant role in shaping public perceptions of different groups. Negative media portrayals can create and perpetuate stereotypes, leading to increased prejudice (Ahmed et al, 2021). However, personal contact with out-group members can counteract these media-induced stereotypes (Roh, 2013; Silverstein & Flamenbaum, 1989). For example, if media frequently depicts a particular ethnic group as prone to violence, individuals who have positive personal interactions with members of that group are less likely to believe such stereotypes (Ahmed, 2017). Therefore, this study posits that direct personal contact reduces prejudice, as it allows for the humanization of out-group members and challenges the simplistic and often negative portrayals presented in the media.

• H1: Direct, personal contact with members of out-groups will be negatively associated with levels of prejudice against these groups.

However, the effects of such contact are likely to vary depending on the specific groups involved. Allport (1954) discussed that certain conditions must be met for contact to reduce prejudice: equal status among group members, common goals, and institutional support. Consequently, extensive empirical research has developed on what types of contact reinforce or weaken prejudice (Kwon, 2024; Pettigrew & Tropp, 2006). If these conditions are not met, studies have reported that increased contact between outgroups can lead to heightened anxiety, thereby reinforcing prejudice (For example: Park, 2015, p. 84; Roh, 2013). The following section will introduce conflict theory to examine how intergroup competition can strengthen prejudice and discuss its implications for contemporary South Korean society.

Intergroup Competition Theory and Prejudice

Second, intergroup competition theory can provide another lens through which to examine the dynamics of prejudice. This theory proposes that when different groups compete for limited resources and opportunities, it can lead to conflict, prejudice, and discrimination. The theory suggests that as minority groups, including immigrants, increase in number, the majority group may perceive them as a threat to their own interests and well-being (Quillian, 1995; Scheepers et al., 2002). This perceived competition can trigger negative attitudes and behaviors towards the minority group.

Intergroup competition theory is rooted in the realistic group conflict theory (Hwang et al., 2007; Joo & Im, 2022; Seok, 2023; Sherif et al., 1988), which posits that competition for tangible resources such as jobs, housing, and social services can lead to intergroup hostility. For instance, if members of a majority group believe that immigrants are taking away job opportunities, they may develop negative attitudes towards immigrants, viewing them as competitors rather than fellow community members. This perception of threat can be exacerbated during times of economic downturn or resource scarcity, where the competition for limited resources becomes more intense.

Building on this theory, this study predicts that prejudice will increase when individuals are perceived as rivals competing for limited resources and opportunities. The perceived threat to the majority group's social and economic standing can lead to heightened levels of prejudice and discrimination against the minority group. This perspective highlights the need to identify which groups are perceived as instigators of social competition or threats within that society. This study aims to determine which groups are recognized as competing in South Korean society in the 2020s through their relationship with prejudice variables.

Furthermore, the media can amplify perceptions of intergroup competition by highlighting stories of conflict and competition between groups. For instance, news reports that focus on job losses among the majority population due to the influx of immigrant workers can fuel perceptions of competition and threat. These media narratives can reinforce existing prejudices and make it more difficult for positive intergroup relations to develop.

• H2. Perceived competition for limited resources and opportunities will be positively associated with levels of prejudice against out-groups.

The answer of H2 is going to be linked to the research question of this paper (RQ) and will demonstrate which groups are perceived as competing in South Korean society in the 2020s, thereby illustrating the negative relationship between contact and prejudice. This will allow for a specific examination of the empirical status of the relationship between contact and prejudice in South Korea.

Impact of Online Hate Speech on Prejudice

Finally, this study examines the impact of being targeted by online hate speech on prejudice levels. Online hate speech can have a profound psychological impact on its victims, leading to increased levels of stress, anxiety, and fear. These psychological effects can, in turn, influence attitudes toward out-groups. When individuals are subjected to hate speech, they may develop heightened sensitivities and defensive attitudes, which can exacerbate existing prejudices.

The experience of being targeted by online hate speech can also drive individuals to seek support from their in-group, further entrenching ingroup biases and increasing prejudice against outgroups. This phenomenon can be understood through the lens of social identity theory (Tajfel, 1982; Tajfel & Turner, 2000), which suggests that individuals derive a sense of identity and selfworth from their group memberships. When their group is under attack, individuals may rally around their in-group, leading to stronger in-group cohesion but also increased hostility towards out-groups.

Moreover, online hate speech often involves dehumanizing and derogatory language, which can strip the target of their individuality and reduce them to negative stereotypes (Citron, 2014; Wachs et al., 2022). This dehumanization process can make it easier for individuals to justify their prejudiced attitudes and discriminatory behaviors (Citron, 2014; Kopytowska & Baider, 2017).

Therefore, this study posits that being a target of online hate speech increases prejudice against outgroups. The harmful effects of online hate speech extend beyond the immediate psychological impact on the victims; they also contribute to the broader societal problem of prejudice and discrimination. Understanding the mechanisms through which online hate speech influences prejudice is crucial for developing effective interventions to combat it.

• H3. Experiencing online hate speech will be positively associated with increased levels of prejudice against out-groups.

In summary, the theoretical framework of this study combines contact theory and intergroup competition theory to provide a comprehensive understanding of the factors that influence prejudice. By examining the role of personal contact, perceptions of competition, and the impact of online hate speech, this study aims to shed light on the complex dynamics of prejudice and offer insights into potential strategies for reducing it.

Data

This study collected data through a survey titled “A Survey of Korean Citizens’ Perception on Online Hate Speech,” which was conducted in February 2023 with the assistance of a polling company, M-Lab.^1,2 The survey aimed to gather detailed information on the perceptions and experiences of online hate speech among Korean adult internet users. This data was compiled from responses of volunteers participating in a survey conducted by M-Lab, using a respondent pool representative of the adult population in Korean society. Quota sampling was conducted to ensure that the sample reflects the overall population distribution. The dataset included responses from 1,000 Korean adult online users. The distribution of the sample data by gender and age, along with the distribution of the population, is presented in Table 1.

Table 1.

2023 Population Distribution of Data (population in 20’s-60’s)

Measures

Social Network Size

The study's main independent variables include (1) social network size and (2) the experience of being the target of online hate speech, referred to as online attack experience. The social network size variable was assessed by asking respondents how many people they personally acquainted with each other in each of the twelve mentioned groups. The responses were measured on a scale from 0 (none) to 4 (more than 6 people). This variable aims to capture the extent of personal contact and exposure respondents have with members of the groups in question. Many empirical studies on contact theory suggests that greater personal contact can reduce prejudice by fostering understanding and empathy (Ahmed, 2017; Pettigrew et al., 2007). However, these effects are not always universal. Researchers such as Allport (1954) and Pettigrew and Tropp (2006) have argued that prejudice reduction can be expected when contact occurs on an equal footing and in an interdependent manner. In addition, intergroup competition theory suggests that intergroup contact can lead to conflict in situations of intergroup resource competition (Kim et al., 2011; Sherif et al., 1988). In order to fully explore the complexity of these contact effects, this study has conducted the analysis in two steps. First, individual networks were included in the model. Second, based on the variable effects and prior research on South Korean society, those networks were classified into three types and reintroduced into the second model. The specific details are presented in Table 2 below.

Table 2.

Social Network Components of Contact Group Types

The network variables for each group were standardized by adding up the number of “acquaintances” in that group and then normalizing the values.

Model of Analysis

The analytical approach employed in this study involved the use of a negative binomial regression model. This model was chosen because the dependent variable, prejudice, exhibits characteristics of a count variable and is overdispersed—meaning that the variance exceeds the mean. Count variables are those that represent the number of times an event occurs, and in this study, it refers to the number of ‘yes’ responses indicating prejudice. Overdispersion is a common issue in count data and can lead to inefficient estimates if not properly addressed.

The dependent variable, ‘level of prejudice’, has a skewed distribution as shown in Figure 1.

Figure 1.

Distribution of the Prejudice LevelNote. Unit of the analysis: Person

To ensure proper model estimation and validation, we validated the presence of overdispersion as follows. In addition, the null hypothesis of “mean and variance are equal” assuming a Poisson distribution was tested, and the null hypothesis was rejected at a significance level of.001 (z = 8.61), and a dispersion value of 3.03 was obtained (based on the final model). Since the null hypothesis of a Poisson distribution was rejected, this study adopted a negative binomial model to analyze the data. This is because the negative binomial model is a popular alternative when a Poisson model cannot be used to estimate a regression model for additive data (Cox, 1983). The model helps in understanding how the independent variables, particularly social network size and online attack experience, influence the level of prejudice among the respondents.

Descriptive Statistics

The basic descriptive statistics of each variable are presented in Table 3.

Table 3.

Sociodemographic Characteristics of the Survey Respondents (N = 1,000)

The data includes a total of 1,000 respondents, comprised of adults aged 18-69. 51% of the respondents are male and 49% are female. Looking at the distribution of education, 68% of respondents were enrolled in college or had graduated from college, and 12.2% were enrolled in graduate school or had graduated from graduate school. 19.8% have a high school diploma or less. The distribution of economic class consciousness averaged 2.65 (SD = 0.79) on a scale of 1 for the lowest class and 5 for the highest class. Political affiliation averages 2.98 (SD = 0.75) on a scale with 1 being very conservative, 5 being very liberal, and 3 being moderate. Respondents spent an average of 287 minutes per day on online media. Of the 10 types of online hate speech, the average number of times they experienced an attack was 1.83 (SD = 2.45).

The extent to which the respondent knows people from a particular group in their social network was measured on a scale of 1 to 4, ranging from “none” to “six or more.” Looking at the scale scores for each of these networks, the lowest scoring group is the LGBTQ+ category (M = 1.19, SD = 0.51), which can be interpreted as the least visible group in the respondent's social network. Other groups with low scores include feminists, foreigners, the economically disadvantaged, and people with disabilities. The highest scoring groups were people from other regions (M = 2.88, SD = 1.05), people of different religions, and young adults.

The correlations between each variable are presented in Table 4.

Table 4.

Intercorrelations of Indicators

Af ter model estimation, we checked multicollinearity with the VIF index and found that there was no problem with the criterion of VIF < 5.0.

Impact of Social Network Size on Prejudice

The first-level results are summarized in Table 5. This table presents the coefficients showing the influence of each social network size on prejudice level. Each estimated value is derived from the individual analysis when the respective social network variables were included alongside the control variables.

Table 5.

The Coefficient Effects of the Network Variables

The results show that the direction of coefficients varies depending on the network with which group members are connected. Among the statistically significant variables, contact with foreigners, the impoverished, feminists, and the LGBTQ+ community shows positive (+) values, meaning that as contact increases, levels of prejudice also rise. On the other hand, contact with people from other regions, those of different religions, and young people in their 20s and 30s yields negative (-) values, indicating that as contact increases, levels of prejudice decrease.

It indicates that, also in South Korean society in the 2020s, contact with outgroups does not solely function to reduce prejudice. Previous researches on outgroups suggest that when people are perceived as competing groups or when members experience status anxiety, contact tends to reinforce discrimination and hatred. Based on this discussion, this study categorized the nine social networks into three groups. The results shown in Table 5, along with the context of contemporary South Korean society, allowed for the classification of social networks into the following three categories: (1) traditional welfare recipient groups, the so-called beneficiary groups, (2) emerging competing groups, and (3) diversity groups (see Table 2).

First, the beneficiary groups are categorized as individuals with disabilities, seniors aged 70 and older, and those living in poverty. These groups have traditionally been considered welfare recipients. In terms of coefficient effects, these groups show either no significant impact (individuals with disabilities, seniors aged 70+ or a positive value.

Second, the emerging competing groups are classified as foreigners and feminists. South Korea had been in low levels of foreign migration; however, the number of foreign workers and immigrants is steadily increasing. This influx is often perceived as bringing diverse vitality to society. However, South Korea, like other societies, faces issues of prejudice and discrimination against foreigners. In this discourse, foreigners are frequently associated with “job scarcity” and “increased crime” (Bang et al., 2010).

The feminist network is categorized as competing group based on the emerging gender issue in South Korea. Studies on hate speech indicate that since the 2010s, “feminism” has emerged as a new keyword in gender conflicts (Jung & Cho, 2021; Park et el., 2020). Notably, the sentiment of “anti-feminism” is often linked to the perception that men are “suffering losses” (Seok, 2023). Research by Lee and Kim (2021), which performed topic modeling analysis on hate discourse in post-COVID-19 South Korea, reveals that feminism and immigrants are uniquely associated with job-related keywords, unlike other social minor groups.

Third, other groups, including people from different regions, different religious backgrounds, young adults, and LGBTQ+, are classified as ‘diversity group’. These groups are not traditional welfare recipients and are not currently perceived as competitors for jobs in Korean society. Furthermore, except for the LGBTQ+ group, as shown in <Table 5>, these groups tend to exhibit reduced prejudice level with increased contact.

This classification is based on individual analysis results and social context; therefore, the following section will validate whether these three network types provide statistically adequate explanatory power.

Impact of Social Network Type on Prejudice

The final analysis of the research model is presented in <Table 6>. ‘Model 1’ is the base model with only control variables. ‘Model 2’ is the research model with explanatory variables. According to the LR test of the fit of the two models (-2LL), the fit of Model 2 is significantly improved compared to that of Model 1. Also, when comparing both AIC and BIC indicators, the value of Model 2 is smaller, so it can be evaluated that the introduction of explanatory variables has improved the research model appropriately.

Table 6.

Results of Base Model(1) and Final Research Model(2)

Table 6 also presents the regression coefficient values converted to exp(β), or ‘incidence rate ratio’ (IRR), for easy interpretation of the analysis results. If the IRR value is greater than 1, the number of occurrences of the dependent variable increases when the independent variable increases.

In this part, we first discuss the results for Hypotheses 1 and 2 of the study’s three hypotheses. Hypothesis 1 predicted that when outgroup members are present in an individual's social network of direct contacts, prejudice will decrease as those contacts increase. The study categorized the types of network contacts into rival groups, beneficiary groups, and diversity groups and tested the effect of each network variable. The variables that proved to be statistically significant were competitor group and diversity group, but their effect coefficients were in opposite directions.

For the diversity group, an increase in network contacts has a decreasing effect on bias. Interpreting the IRR values, a one-unit increase in network contacts for the diversity group reduces bias by 20%. In contrast, for the competing group, an increase in network contacts has an increasing effect on bias. A one-unit increase in network contacts to the rival group is associated with a 17% increase in bias. The beneficiary group effect fails to reject the null hypothesis. In other words, the effect of social network size on prejudice varied significantly depending on the type of network. These results provided robust support for both contact theory and intergroup competition theory, indicating that these frameworks are valuable for understanding the dynamics of prejudice.

First, the effect of the diversity group network variable is consistent with contact theory. This finding aligns with contact theory, which posits that increased personal contact with diverse groups can reduce prejudice by promoting more accurate and empathetic perceptions of outgroups. In this context, knowing more individuals from different regions or age groups likely facilitates positive interactions and reduces social distance, thereby mitigating prejudicial attitudes.

However, the study found that knowing more groups that respondents perceive to be in competition with them was linked to a significant increase in prejudice against this group. This outcome supports intergroup competition theory, which suggests that when different groups are seen as competing for limited resources, prejudice and conflict are likely to arise.

Foreigners and feminists, who have been categorized as competing groups, have recently been called the new conflict groups in Korean society. As the influx of foreigners continues to increase and the number of migrant workers has grown significantly, the perception that foreigners are taking jobs is spreading rapidly in Korean society. At the same time, there has been an increase in the number of demeaning and hateful remarks against foreigners. Although they may seem like very different groups, hostility toward feminism and women who advocate for it is also rooted in competition for social resources. While there is no direct discourse of “women taking jobs” as there is with foreigners, the perception that feminism and gender equality discourse threaten men’s socioeconomic status is rapidly spreading in South Korea. In other words, these groups are seen as threats to the existing social system's resource allocation order and as competitors in the resource allocation arena. They are positioned differently from the disabled, the elderly, and the poor, who have traditionally been the beneficiaries of the social welfare system. Social welfare recipient groups can also be considered competitors as groups that take social resources. Therefore, the coefficient effect of the beneficiary group variable is also positive. However, this coefficient fails to reject the null hypothesis at the p<0.05 level.

Taken together, these results highlight that an individual's prejudice against an outgroup is influenced by whether that group is perceived as a competitor for social resources. If a group is viewed as competitors for limited resources, this perception can fuel negative attitudes and discriminatory behaviors towards them. This finding supports the group competition theory. On the other hand, contact theory was more applicable to groups that are unrelated to resource allocation and not seen as beneficiaries of the social welfare system, such as individuals from different regions and younger age groups. Increased interpersonal contact with these groups was associated with reduced prejudice, supporting the idea that personal interactions can break down stereotypes and foster more positive perceptions.

The above analysis shows that the context of a society plays an important role in shaping prejudicial attitudes, and therefore contact theory and intergroup competition theory should be applied differently depending on the specifics.

Effects of Online Hate Speech on Prejudice

The study also examined the impact of experiencing online hate speech on prejudice levels. In the analysis, the variable of experience of being targeted by online hate speech was found to be significant (p<0.001). Translating this into an IRR value, we can expect a 13% increase in the level of prejudice against outgroups for a onetime increase in the experience of being targeted by hate speech. This indicates another harmful consequence of online hate speech, which not only affects the immediate psychological wellbeing of the victims but also influences their attitudes towards out-groups.

When individuals are subjected to hate speech, it can lead to increased stress, anxiety, and defensive attitudes, which may exacerbate prejudicial views. Individuals exposed to hate speech, or those who perceive themselves as targets of such speech, may experience feelings of degradation, as well as negative emotions such as stress, anger, and anxiety. These emotions can undermine their selfesteem and lead to a perception that their in-group is under threat. This perception can, in turn, result in defensive mechanisms towards the in-group and an exclusionary attitude towards the outgroup. Additionally, victims of online hate speech may seek support from their in-group, leading to stronger in-group biases and increased hostility towards out-groups. Because the fundamental structure of hate speech inherently involves the process of othering, which involves distinguishing oneself from a different group. Online hate speech, inherently negative and damaging, has the potential to reinforce existing prejudices and create new ones.

The study's findings suggest that online hate speech acts as a catalyst for prejudice, amplifying negative attitudes and contributing to a more polarized and divided society. This underscores the urgent need for effective interventions to combat online hate speech and mitigate its harmful effects on both individual attitudes and broader social cohesion.

Disclosure Statement

No potential conflict of interest was reported by the author.