Psychological Characteristics Associated With Cyberbullying: Focusing on the Victim-Perpetrators and Gender Differences

INTRODUCTION

Although the definition of cyberbullying varies among scholars, it is commonly defined by Patchin and Hinduja [1] as “an act of repeatedly harming others through electronic devices such as smartphones or computers in cyberspace (e.g., internet, messengers) and on social media.” The severity of cyberbullying is increasing with the development of electronic technology, leading to new forms of online harassment, such as group chat bullying and identity theft [1]. Adolescence is a critical period for establishing one’s identity through peer relationships [2]. Although mobile messengers and social media are important tools for forming these relationships, they can increase cyberbullying, negatively affecting adolescent development [3]. Moreover, remote learning during the COVID-19 pandemic increased the time adolescents spent online, thus creating a greater risk of cyberbullying. According to the 2022 Survey on Cyberbullying, 37.5% of adolescents reported victimization and 20.6% reported perpetration, and notably, 16.4% experienced both, nearly double the 8.3% reported in 2021 [4].

Cyberbullying can be perpetrated not only by acquaintances but also by anonymous individuals without temporal or spatial constraints. Bullying content can spread rapidly, making timely intervention difficult and often leading to long-lasting adverse effects. Moreover, fear of being reprimanded or causing distress to their parents or teachers often prevents adolescents from reporting their victimization, making it challenging for caregivers to identify and address the situation appropriately [5-7]. Revenge is more readily enacted in cyberspace than through traditional bullying, and the lack of social cues regarding others’ emotional responses may facilitate the transition from victim to perpetrator [8].

Cyberbullying has various negative effects on adolescents’ emotional development [9]. Adolescents who experience persistent and repetitive victimization may experience depression, negative self-concept, low self-esteem, helplessness, and guilt, which can lead to suicidality [10]. Notably, adolescents who are both victims and perpetrators often report the most severe psychological distress, frequently manifested as significant depressive symptoms and low self-esteem, leading to poorer psychosocial adjustment than that observed in those who are exclusively victims or perpetrators [11]. Therefore, it is imperative to provide therapeutic interventions and pay closer attention to adolescents involved in cyberbullying victimization and perpetration.

Several studies have suggested that the psychological impact of cyberbullying differs according to gender. Female students involved in cyberbullying experience more severe depression and report more frequent suicidal ideation and attempts, whereas male students are more likely to experience externalizing problems [12,13]. Such gender-specific findings are often attributed to the distinct developmental trajectories of how male and female adolescents navigate peer relationships and process socio-emotional information [14]. Cyberspace serves as the primary communication arena for adolescents, and its characteristics, such as anonymity, lack of face-to-face interaction, and rapid information dissemination, may amplify these developmental differences.

Recently, research in this area has expanded as the recognition of the serious impact of cyberbullying on adolescents has increased. However, public awareness of the severity of cyberbullying has not kept pace with its increasing prevalence [15]. The literature consistently identifies a range of internalizing problems as severe consequences of cyberbullying [10]. Among these, depression and low self-esteem are highlighted as two of the most central and robustly documented psychological factors associated with the dual-role experience [10,11]. Furthermore, cyberbullying is not only linked to internal distress but also to a wider spectrum of behavioral and social difficulties, including externalizing and peer relationship problems. Therefore, this study selected depression and self-esteem as core internalizing indicators, alongside broader behavioral and social dimensions, to comprehensively assess psychological distress related to cyberbullying. Individuals involved in cyberbullying constitute heterogeneous groups, differing in factors, such as the type and extent of their involvement and gender. Analyzing these subgroups can help develop tailored prevention and intervention strategies. However, little research has been conducted that simultaneously considers the characteristics of the victim-perpetrator group and gender among Korean adolescents. Therefore, this study investigated cyberbullying experiences (non-involvement, victimization, perpetration, and victim-perpetration) and the associated psychological characteristics among first- and second-year middle school students in Cheongju. Potential gender differences in psychological characteristics related to cyberbullying involvement were also explored.

METHODS

Participants

The participants were first- and second-year students from three public middle schools in Cheongju, Chungbuk. No specific inclusion or exclusion criteria were applied. Data were collected from July 14 to September 1, 2022. Considering that adolescents might be unfamiliar with the concept of cyberbullying or find it difficult to address, an educational session was provided on the day of the survey. This session covered the definition, types, severity, and legal consequences of cyberbullying. Written informed consent was obtained from the adolescents and their parents or legal guardians, who fully understood the study’s purpose and methods. The survey took approximately 30 minutes to complete. A total of 449 students completed the survey and were included in the analysis. This study was approved by the Bioethics Committee at Chungbuk University Hospital (IRB No. 2022-06-024) and conducted in accordance with the Declaration of Helsinki.

Sociodemographic characteristics

Sociodemographic data included participants’ gender, grade, academic performance, satisfaction with school life, and satisfaction with family life.

Cyberbullying experiences

In this study, cyberbullying victimization and perpetration were assessed using a scale adapted and validated by Kim [16] based on the original scale developed by Patchin and Hinduja [1]. Cyberbullying can be perpetrated by anonymous strangers or known individuals [17]. However, this study specifically focused on experiences among known peers. Therefore, our analysis of cyberbullying victimization and perpetration was limited to peers attending the same or different schools. Two items from the 13-item perpetration scale developed by Kim [16] were excluded from the analysis because they involved unknown individuals. Participants were asked to indicate the frequency with which they had bullied or been bullied by their peers over the past six months via various media, including chat apps, social media, e-mail, webpages, and mobile phone messages. The items covered a range of bullying behaviors, including intentional exclusion, disclosing or leaking others’ personal information online, and online threats or coercion. The victimization and perpetration scales each comprised 11 items. Responses were measured on a 5-point Likert scale ranging from 1 (“not at all”) to 5 (“very much”), with higher scores indicating more experience with cyberbullying. In this study, the internal consistency (Cronbach’s α) was 0.910 for the victimization scale and 0.896 for the perpetration scale.

No cutoff score has been established for cyberbullying questionnaires to define victims or perpetrators [18]. Therefore, in this study, participants were defined as having experienced cyberbullying victimization if their scores exceeded the upper quartile of the total distribution of victimization scores. The same criterion was applied to define perpetration. This classification method has been employed in previous studies, including Yang et al. [18]. Based on their cyberbullying experiences, participants were classified into four groups: Neither (N group), Victim (V group), Perpetrator (P group), and Victim-Perpetrator (VP group).

Psychological characteristics

Depression

The Beck Depression Inventory (BDI) was used to assess depressive symptoms. Participants used a 4-point Likert scale to rate their feelings and thoughts over the past week on a 21-item questionnaire. According to a study investigating the validity of the Korean version of the BDI as a screening tool for depression [19], a score of 14 or higher is considered indicative of mild depressive symptoms, warranting clinical attention and evaluation. Total scores range from 0 to 63, with higher scores indicating greater levels of depressive symptoms. The reliability and validity of the Korean version of the BDI have been well-established [20].

Self-esteem

Participants’ self-esteem was evaluated using the Self-Esteem Scale (SES) developed by Rosenberg [21] and validated by Lee [22]. This self-report questionnaire was designed to measure the positive acceptance of oneself (self-approval) grounded in self-respect. It comprises 10 items, including five positively worded items and five reverse-scored items. Responses are provided on a 4-point Likert scale ranging from 1 (“almost never”) to 4 (“almost always”). Total scores range from 10 to 40, with higher scores indicating greater self-esteem.

Psychological strengths and difficulties

The Strengths and Difficulties Questionnaire (SDQ) was used to assess participants’ psychological and social functioning. This self-report measure comprises 25 items rated on a 3-point Likert scale: 0 (“not true”), 1 (“somewhat true”), and 2 (“certainly true”). The “strengths” score is derived from the prosocial behavior subscale, which includes items on kindness and cooperation and has a maximum score of 10. The “total difficulties” score is based on four subscales: hyperactivity/inattention (e.g., attention deficit, impulsivity), emotional symptoms (e.g., depression, anxiety), conduct problems (e.g., aggressive or defiant behavior, rule-breaking), and peer relationship problems (e.g., conflict and alienation with peers). Each subscale has a maximum score of 10, resulting in a total difficulties score of up to 40. Higher scores on the strengths scale and lower scores on the difficulties scale reflect more adaptive psychosocial functioning. The clinical utility of the SDQ has been well-established [23].

Statistical analysis

Data were analyzed using SPSS version 29.0 (IBM Corp.).

The sample was divided into male and female groups to examine gender differences in cyberbullying experiences and psychological characteristics. For categorical variables, a chisquare test was performed, and residual analysis was applied to examine specific proportional differences in cyberbullying involvement between genders. Independent sample t-tests were performed for continuous variables. Chi-square tests were used for categorical variables, and analysis of variance (ANOVA) was used for continuous variables to identify any differences in sociodemographic and psychological characteristics among the four cyberbullying groups. Scheffe’s posthoc test was used to examine intergroup differences. Significant factors associated with cyberbullying victimization and perpetration were identified using multiple regression analyses, with participants’ sociodemographic information and psychological characteristics as independent variables and victimization and perpetration scores as dependent variables. These analyses were conducted separately for male and female participants to identify gender-specific differences. Additionally, a binomial multivariate logistic regression analysis was conducted with the N group as the reference group to focus on the characteristics of the VP group. The significance of the model was confirmed using a likelihood ratio test. Participants’ sociodemographic information and psychological characteristics were entered as independent variables, and odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were calculated for each variable. Separate binomial multivariate logistic regression analyses were performed for male and female participants to explore gender differences in the risk factors for belonging to the VP group.

RESULTS

Sociodemographic characteristics

Table 1 presents the participants’ sociodemographic characteristics. Of the participants, 177 (39.4%) were first-year students and 272 (60.6%) were second-year students. The sample comprised 185 male (41.2%) and 264 female (58.8%) participants. Regarding satisfaction with school life, 8 participants (1.8%) reported being “very dissatisfied,” 33 (7.4%) “generally dissatisfied,” 138 (30.7%) “neutral,” 199 (44.3%) “generally satisfied,” and 71 (15.8%) “very satisfied.” For satisfaction with family life, 3 participants (0.7%) reported being “very dissatisfied,” 22 (4.9%) “generally dissatisfied,” 95 (21.2%) “neutral,” 152 (33.9%) “generally satisfied,” and 177 (39.4%) “very satisfied.” Assuming that participants whose victimization or perpetration scores exceeded the upper quartile were involved in cyberbullying, 143 (31.8%) were involved in such activities. Based on their experiences over the past six months, the participants were categorized into four groups: 34 (7.6%) were classified into the V group, 32 (7.1%) into the P group, 77 (17.1%) into the VP group, and 306 (68.2%) into the N group.

Table 1.

Socioeconomic characteristics of participants

Characteristic	Value (n=449)
Gender
Male	185 (41.2)
Female	264 (58.8)
Grade
1st grade	177 (39.4)
2nd grade	272 (60.6)
School life satisfaction
Very dissatisfied	8 (1.8)
Generally dissatisfied	33 (7.4)
Neutral	138 (30.7)
Generally satisfied	199 (44.3)
Very satisfied	71 (15.8)
Family life satisfaction
Very dissatisfied	3 (0.7)
Generally dissatisfied	22 (4.9)
Neutral	95 (21.2)
Generally satisfied	152 (33.9)
Very satisfied	177 (39.4)
Cyberbullying group
Neither	306 (68.2)
Victim	34 (7.6)
Perpetrator	32 (7.1)
Victim-perpetrator	77 (17.1)

Values are presented as number (%).

Gender differences in sociodemographic and psychological characteristics and cyberbullying experiences

A total of 56 male (30.3%) and 87 female (33.0%) participants were involved in cyberbullying, whereas 129 male (69.7%) and 177 female (67.0%) participants were not (N group). When examining the nature of this involvement, significant gender differences were found in cyberbullying experiences. Among male students, the P group was relatively larger (n=26, 14.1%), whereas the V group was smaller (n=3, 1.6%). Conversely, among female students, the V group was relatively larger (n=31, 11.7%), whereas the P group was smaller (n=6, 2.3%). For both genders, the VP group was the largest (14.6% and 18.9%, respectively). Residual analysis indicated that the proportion of female students in the V group was significantly higher than expected (standardized residual, Z=2.5), whereas the proportion of male students was lower than expected (Z=-2.9, p<0.001). In contrast, the proportion of male students in the P group was significantly higher than expected (Z=-3.0), whereas the proportion of female students was lower than expected (Z=3.5, p<0.001). The gender difference in the proportion of the VP group was not significant. Regarding psychological characteristics, female students reported higher scores than male students on the BDI (mean [M]=11.61, standard deviation [SD]=9.83 vs. M= 8.12, SD=7.25, respectively; p<0.001) and SDQ-Emotional Symptoms subscale (M=3.88, SD=2.80 vs. M=2.99, SD=2.20, respectively; p<0.001). Female students had lower SES scores than male students (M=28.15, SD=6.52 vs. M=30.15, SD= 6.06, respectively; p<0.001).

Comparison of sociodemographic and psychological characteristics among cyberbullying groups

Table 2 presents the characteristics of the N, V, P, and VP groups. The ANOVA and post-hoc analysis results revealed several differences between groups. Compared to the N group, the VP group scored significantly higher on the BDI (M= 14.30, SD=9.88 vs. M=8.46, SD=8.11, respectively; p<0.001) and SDQ-Hyperactivity/Inattention subscale (M=4.51, SD= 2.30 vs. M=3.41, SD=2.22, respectively; p=0.002). A similar pattern was observed for the SDQ-Emotional Symptoms (M=4.84, SD=2.58 vs. M=2.97, SD=2.40, respectively; p< 0.001), SDQ-Conduct Problems (M=3.44, SD=1.71 vs. M= 2.24, SD=1.50, respectively; p<0.001), and SDQ-Peer Relationship Problems (M=2.64, SD=1.76 vs. M=2.00, SD=1.59, respectively; p=0.030) subscales. Furthermore, the VP group reported lower family life satisfaction (M=3.64, SD=0.99 vs. M=4.22, SD=0.86, respectively; p<0.001) and significantly lower SES scores (M=26.56, SD=6.46 vs. M=30.05, SD=6.17, respectively; p<0.001) than the N group. Group V also showed significant differences compared to the other groups. The BDI scores in the V group (M=16.38, SD=11.03) were significantly higher than those in the N (M=8.46, SD=8.11) and P (M=10.03, SD=6.72) groups (p<0.001). Compared to the N group, the V group had higher scores on the SDQ-Emotional Symptoms (M=5.29, SD=3.02 vs. M=2.97, SD=2.40; p<0.001), SDQ-Conduct Problems (M=3.06, SD=1.58 vs. M=2.24, SD= 1.50; p<0.001), and SDQ-Peer Relationship Problems (M= 2.85, SD=2.00 vs. M=2.00, SD=1.59; p=0.030) subscales. Additionally, the V group (M=25.59, SD=5.72) had significantly lower SES scores than the N group (M=30.05, SD=6.17) (p<0.001). Compared to the N group, the P group showed significantly higher scores on the SDQ-Hyperactivity/Inattention (M=4.94, SD=2.49 vs. M=3.41, SD=2.22, respectively; p=0.002) and SDQ-Conduct Problems (M=3.53, SD=1.78 vs. M=2.24, SD=1.50, respectively; p<0.001) subscales.

Table 2.

Comparison of participant characteristics among the four cyberbullying involvement groups

Psychological characteristics	Cyberbullying status				p
	Neithera (n=306)	Victimb (n=34)	Perpetratorc (n=32)	Victim-perpetratord (n=77)
Gender (M/F)	129/177	3/31	26/6	27/50
School life satisfaction	3.73±0.86	3.56±0.96	3.50±0.92	3.44±0.97	0.049
Family life satisfaction	4.22±0.86	3.88±0.98	3.84±0.99	3.64±0.99	<0.001; a>d
BDI	8.46±8.11	16.38±11.03	10.03±6.72	14.30±9.88	<0.001; b,d>a,b>c
SES	30.05±6.17	25.59±5.72	28.03±6.39	26.56±6.46	<0.001; a>b,d
SDQ-strength	6.69±2.15	6.85±2.31	6.16±2.08	6.23±1.93	0.424
SDQ-H/I	3.41±2.22	3.91±2.15	4.94±2.49	4.51±2.30	0.002; c,d>a
SDQ-emotion	2.97±2.40	5.29±3.02	3.63±2.20	4.84±2.58	<0.001; b,d>a
SDQ-conduct	2.24±1.50	3.06±1.58	3.53±1.78	3.44±1.71	<0.001; b,c,d>a
SDQ-peer	2.00±1.59	2.85±2.00	2.38±1.66	2.64±1.76	0.030; b,d>a

Values are presented as mean±standard deviation unless otherwise indicated. BDI, Beck Depression Inventory; F, female; H/I, hyperactivity/inattention; M, male; SDQ, Strengths and Difficulties Questionnaire; SES, Self-Esteem Scale.

Factors associated with cyberbullying victimization and perpetration

In the multiple regression analyses, various characteristics, including satisfaction with school life and family life and BDI, SES, SDQ-Strengths, SDQ-Hyperactivity/Inattention, SDQ-Emotional Symptoms, SDQ-Conduct Problems, and SDQ-Peer Problems scores, as well as gender and grade, were included as independent variables. The dependent variables were cyberbullying victimization and perpetration scores. The regression model examining the factors associated with cyberbullying victimization was statistically significant, with an overall fit of F=7.86 (p<0.001) and an adjusted R² of 0.150. An increase in emotional symptoms was significantly associated with an increase in victimization scores (B=0.55, t=2.83, p=0.003). Furthermore, an increase in SDQ-Conduct Problems score was significantly associated with victimization (B=0.83, t=3.19, p=0.001), suggesting that adolescents with more severe emotional and conduct problems experienced greater cyberbullying victimization. The model examining the factors associated with cyberbullying perpetration was also statistically significant, with an overall fit of F=4.70 (p<0.001) and adjusted R² of 0.107. SDQ-Conduct Problems score was significantly associated with perpetration. That is, adolescents with more conduct problems tended to engage in more cyberbullying perpetration (B=1.00, t=4.03, p<0.001) (Table 3). For males, the victimization model’s overall fit was F=2.98 (p=0.002), with an adjusted R² of 0.095. An increase in conduct problems was significantly associated with increased cyberbullying victimization (B=0.96, t=3.09, p=0.003). The overall fit of the perpetration model was F=4.12 (p<0.001), with an adjusted R² of 0.133. An increase in conduct problems was strongly associated with an increase in cyberbullying perpetration (B=1.15, t=4.35, p<0.001). For females, the victimization model’s overall fit was F=5.27 (p< 0.001) with an adjusted R² of 0.129. Females with more severe emotional problems experienced greater cyberbullying victimization (B=0.73, t=2.87, p=0.004). The overall fit of the perpetration model was F=3.00 (p=0.004), with an adjusted R² of 0.098. Having more conduct problems was associated with engaging in more cyberbullying perpetration (B=0.95, t=3.48, p<0.001) (Table 3).

Table 3.

Associations of cyberbullying victimization and perpetration scores with psychological characteristics in the total sample and by gender

Dependent variable	Sample	Associated factors	B	SE	Beta	t	p	Model fit
Cyberbullying victimization score	Total	SDQ-emotion	0.55	0.19	0.18	2.83	0.003	F=7.86 p<0.001 adj R2=0.150
		SDQ-conduct	0.83	0.26	0.18	3.19	0.001
	Male	SDQ-conduct	0.96	0.31	0.28	3.09	0.003	F=2.98 p=0.002 adj R2=0.095
	Female	SDQ-emotion	0.73	0.25	0.24	2.87	0.004	F=5.27 p<0.001 adj R2=0.129
Cyberbullying perpetration score	Total	SDQ-conduct	1.00	0.25	0.31	4.03	<0.001	F=4.70 p<0.001 adj R2=0.107
	Male	SDQ-conduct	1.15	0.26	0.39	4.35	<0.001	F=4.12 p<0.001 adj R2=0.133
	Female	SDQ-conduct	0.95	0.27	0.27	3.48	<0.001	F=3.00 p=0.004 adj R2=0.098

Statistics are based on the multiple regression analysis. The analysis for the total sample included independent variables such as psychological characteristics (school life satisfaction, family life satisfaction, BDI, SES, SDQ-strength, SDQ-H/I, SDQ-emotion, SDQconduct, SDQ-peer), gender, and grade. The analysis by gender included the same psychological characteristics and grade. The dependent variable is either cyberbullying victimization score or perpetration score. Only statistically significant p-values (p<0.05) are displayed in the table. adj R², adjusts for the number of predictors in the model; B, unstandardized coefficient; BDI, Beck Depression Inventory; Beta, standardized coefficient; H/I, hyperactivity/inattention; SDQ, Strengths and Difficulties Questionnaire; SE, standard error; SES, Self-Esteem Scale.

Factors associated with belonging to the VP group

In the multiple logistic regression analysis, which calculated the ORs for belonging to the VP group, with the N group as the reference, the overall model was significant after controlling for grade and gender (χ²=20.37, p=0.026). Among the independent variables, only SDQ-Conduct Problems scores significantly increased the odds of being in the VP group (OR=1.72, 95% CI: 1.20–2.45, p=0.003) (Table 4). When the analysis was conducted for male students only, controlling for grade, the overall model was not significant, and no independent variables emerged as significantly associated factors. In contrast, for female students, the overall model was significant (χ²=20.25, p=0.027). Both BDI (OR=1.20, 95% CI: 1.05–1.38, p=0.008) and SDQ-Conduct Problems (OR= 2.04, 95% CI: 1.07–3.90, p=0.030) scores significantly increased the likelihood of belonging to the VP group (Table 4).

Table 4.

Associations between psychological characteristics and the likelihood of belonging to the victim-perpetrator group (total, male, and female samples)

Psychological characteristics	Cyberbullying victim-perpetrator
	Total		Male		Female
	OR (95% CI)	p	OR (95% CI)	p	OR (95% CI)	p
School sat.	0.96 (0.56-1.67)	0.893	0.74 (0.30-1.85)	0.524	1.22 (0.49-3.03)	0.663
Family sat.	0.76 (0.50-1.17)	0.211	0.55 (0.24-1.30)	0.173	0.85 (0.47-1.54)	0.597
BDI	1.05 (0.97-1.14)	0.237	0.90 (0.76-1.06)	0.211	1.20 (1.05-1.38)	0.008*
SES	1.05 (0.93-1.18)	0.420	0.94 (0.73-1.19)	0.591	1.11 (0.95-1.29)	0.196
SDQ-strength	1.08 (0.88-1.34)	0.460	1.10 (0.75-1.63)	0.621	1.05 (0.79-1.40)	0.725
SDQ-H/I	0.98 (0.79-1.22)	0.880	0.97 (0.67-1.42)	0.887	0.90 (0.64-1.25)	0.529
SDQ-emotion	1.00 (0.79-1.27)	0.973	1.31 (0.83-2.06)	0.249	0.79 (0.56-1.13)	0.197
SDQ-conduct	1.72 (1.20-2.45)	0.003*	1.43 (0.80-2.53)	0.227	2.04 (1.07-3.90)	0.030*
SDQ-peer	0.96 (0.70-1.31)	0.794	1.58 (0.85-2.93)	0.150	0.68 (0.41-1.12)	0.132

Statistics are based on the binomial multivariable logistic regression analysis. The dependent variable is the likelihood of belonging to the victim-perpetrator group compared to neither group. Total sample model controlled for grade and gender; the model was statistically significant (likelihood ratio test: χ²=20.37, p=0.026). Male and femaled models controlled for grade. The model was statistically significant for female (χ²=20.25, p=0.027) but not for male (χ²=14.97, p=0.133). *p<0.05. BDI, Beck Depression Inventory; CI, confident interval; Family sat., family life satisfaction; H/I, hyperactivity/inattention; OR, odds ratio; School sat., school life satisfaction; SDQ, Strengths and Difficulties Questionnaire; SES, Self-Esteem Scale.

DISCUSSION

In this study, 31.8% of the participants were involved in cyberbullying, and significant gender differences were observed in the patterns of involvement. A relatively higher proportion of male students were in the P group, whereas a relatively higher proportion of female students were in the V group. These results are consistent with those of a systematic review by Chun et al. [24], who reported higher rates of male cyberbullying perpetration, but higher rates of female victimization. The most important finding was that among the 143 participants involved in cyberbullying, 77 (17.1% of the total sample) belonged to the VP group, making this dualrole group the largest, accounting for more than half of all adolescents involved. This trend aligns with the findings of Mishna et al. [8] and holds true even when the data were analyzed separately by gender. Unlike traditional bullying, in cyberspace, anonymity can reduce the influence of power imbalances, allowing for a more fluid transition between the victim and perpetrator [25,26]. Our findings suggest that once an adolescent is exposed to cyberbullying, either as a victim or perpetrator, the role transition between the two becomes highly dynamic. Thus, victim and perpetrator may not be distinct categories in cyberbullying but exist on a continuum.

In this study, the VP group experienced the most severe psychological difficulties. When analyzing the four cyberbullying groups, the differences in psychological characteristics between the VP and N groups were the most pronounced and included internalizing and externalizing problems. Compared to the N group, the VP group had significantly higher scores on the BDI, SDQ-Hyperactivity/Inattention, SDQEmotional Symptoms, SDQ-Conduct Problems, and SDQPeer Relationship Problems and lower scores on family life satisfaction and SES. This represents a significant difference from the N group for seven measures. The V group differed from the N group in five measures, whereas the P group differed in two. These findings suggest that the VP group exhibited the most severe psychopathology. According to recent research, individuals in the bully-victim group experience more emotional problems and internalizing/externalizing symptoms than those who are only victims or perpetrators [8]. This implies that any adolescent who has experienced cyberbullying, whether as a victim or perpetrator, is at risk of repeatedly experiencing both roles and accumulating negative effects, underscoring the need for early intervention. Although this study did not show statistically significant differences between the VP group and the V or P group, this may support the interpretation that the distinct characteristics between the groups are diluted by dynamic role transitions, suggesting that these three groups may exist on a continuum [27]. One unique finding of this study is that family life satisfaction differed significantly only between the VP and N groups. The significantly lower family life satisfaction in the VP group indicates that adolescents who take on dual roles are likely to experience difficulties not only with their individual psychological well-being but also in their most crucial support system: the family. This aligns with other research indicating that adolescents who experience both roles report more severe psychological distress and difficulties in the parent–child relationship than those who are only victims or perpetrators [28]. This indicates that family support plays a crucial role in preventing students who have experienced cyberbullying from falling into a vicious cycle of victimization and perpetration and experiencing greater hardship.

The multiple regression analysis including all participants revealed that SDQ-Conduct Problems scores were the only factor associated with both victimization and perpetration scores. This finding suggests that conduct problems are a key factor in intensifying cyberbullying victimization and perpetration. SDQ-Emotional Symptoms scores were significantly associated only with victimization scores and not with perpetration scores. This is consistent with the findings of longitudinal studies suggesting that adolescents experiencing emotional difficulties are vulnerable online, increasing their risk of becoming targets [29]. This study used a crosssectional design, which is a limitation because it cannot establish a causal relationship between cyberbullying involvement and psychological characteristics. However, when interpreting these results from a developmental psychopathology perspective, two complementary bidirectional pathways are likely at work [30]. The first is the vulnerability pathway, in which a youth’s preexisting psychological difficulties are a risk factor for cyberbullying involvement [31]. The findings from our multiple regression analysis suggest that these externalizing and internalizing problems are inherent vulnerabilities that lead youth to participate in bullying. The second is the impact pathway, in which cyberbullying exposure acts as a severe stressor that causes or exacerbates psychopathology. The finding that the VP group showed significant maladjustment across the broadest range of psychological indicators compared to the N group supports the notion that exposure to cyberbullying, especially the dual role of the VP group, is a powerful environmental stressor that causes cumulative mental health deterioration. These two pathways become more intricately intertwined in the cyber environment because of its high role fluidity. Youth who become involved in bullying due to vulnerability may experience more severe psychological distress as a result. This can negatively affect their coping abilities in turn, reinforcing a feedback loop that traps them in a vicious cycle of oscillations between victimization and perpetration.

The VP group was the largest among those with cyberbullying experience. Therefore, this study used multiple logistic regression to identify key factors distinguishing between the VP and N groups. The analysis including all participants revealed that only SDQ-Conduct Problems scores significantly increased the risk of belonging to the VP group. These results reaffirm that conduct problems are a key risk factor for experiencing both roles and are consistent with prior research showing that externalizing problems, such as rule-breaking, are associated with the VP group [5,32]. The mechanism by which conduct problems trigger this dual-role involvement can be interpreted in two ways. First, conduct problems are associated with aggression and antisocial behaviors that violate social norms and the rights of others. Adolescents with conduct problems may exhibit callous-unemotional traits characterized by difficulty in recognizing guilt or empathy [28]. When combined with the anonymity of cyberspace, these traits can lower the psychological barrier to perpetration by reducing empathy for others’ suffering or guilt over one’s actions. Second, these aggressive and antisocial behaviors can provoke antipathy or retaliation from peers, making them potential targets of victimization. Consequently, an adolescent who becomes a perpetrator because of conduct problems can subsequently become a victim of retaliation, and this victimization experience is then used to justify their own retaliatory behavior, leading them to engage in further perpetration [33].

The factors associated with cyberbullying involvement differed by gender. For male students, gender-specific multiple regression analysis showed that higher SDQ-Conduct Problem scores were associated with significantly higher scores for victimization experience and perpetration. This is consistent with the higher prevalence of externalizing problems in male students, suggesting that these problems manifest as aggressive behaviors that lead to a vicious cycle of retaliation or exclusion. However, the reverse causal path, in which cyberbullying exposure exacerbates these externalizing problems, cannot be excluded. These two pathways interact dynamically, reinforcing this vicious cycle. However, the logistic regression analysis exploring the risk factors for belonging to the VP group found no significant associated factors among male students. This null finding is important and requires careful interpretation from two perspectives. First, an issue of statistical power may be at play. The male VP group (n=27) was considerably smaller than the female VP group (n=50), which may have provided insufficient power to detect existing effects. Second, this result suggests that the variables measured in this study (BDI, SES, and SDQ scores) may be insufficient to explain the pathway to the VP group for male students. Thus, the male VP pathway may be driven by other key psychosocial factors not included in this model. For example, low self-control has been proposed as a key mechanism explaining the overlap between victim and perpetrator, because it relates to impulsivity leading to perpetration and vulnerability leading to victimization. In one study, this association was significant only among boys [34]. Furthermore, meta-analyses have shown a clear association between the VP group and low empathy and high callousunemotional traits. For male adolescents, these emotional deficits may be a more critical factor associated with the VP group [35]. Finally, male students tend to be more concerned with establishing status and dominance in peer relationships (social dominance orientation), which has been found to be strongly associated with male bullying behavior [14,36]. Future research should explore alternative gender-specific theoretical models that incorporate these unmeasured factors to explain the male VP pathway.

For female students, the multiple regression analysis revealed that higher SDQ-Emotional Symptoms scores were associated with increased cyberbullying victimization. This is related to the tendency of female adolescents to report more internalizing symptoms, such as depression and anxiety, than male adolescents, and this emotional vulnerability may increase their risk of becoming targets for cyberbullying. Conversely, the impact pathway, in which the experience of persistent cyberbullying victimization acts as a severe stressor that leads to increased SDQ-Emotional Symptoms scores must also be considered. Notably, in the logistic regression analysis examining the risk factors for belonging to the VP group, SDQ-Conduct Problems and BDI scores significantly increased the risk for female students. Considering that conduct problems were identified as a risk factor for the VP group in the analysis including all participants, this implies that depression is a gender-specific risk factor for female adolescents. This gender-specific finding can be explained by the relational orientations of female adolescents, who are generally more sensitive to relational loss and negative social evaluation, making them more likely to experience internalizing symptoms in response to difficulties in peer relationships [37,38]. These negative emotions may drive them to engage in excessive online self-disclosure or relationship-seeking to compensate for real-life difficulties [39,40]. Paradoxically, this behavior, under the anonymity of cyberspace, makes female adolescents more vulnerable to being targeted for perpetration and retaliation. Considering our finding that the VP group showed the most widespread psychological maladjustment, it is highly likely that exposure to the VP role itself acted as a powerful environmental stressor, leading to increased BDI and SDQ-Conduct Problems scores. This may have formed a complex and dynamic vicious cycle among female adolescents. Although this study did not find a significant direct association between peer relationship problems and cyberbullying, future research should explore the complex interplay between depression, peer relationship problems, and social media usage patterns, particularly among female adolescents, to understand how these factors lead to classification in the VP group.

This study has several limitations. First, the data were collected from first- and second-year middle school students in Cheongju, limiting the results to a specific region and age group. Considering that cyberbullying patterns may vary according to geographic location and grade level, future research should investigate cyberbullying using larger and more diverse samples that include a wider range of regions and grades. Second, this study utilized self-report questionnaires; thus, participants may have under- or overreported their cyberbullying experiences. However, most studies are based on self-reporting because cyberbullying research requires investigation into highly personal experiences. Third, this study’s cross-sectional design only allowed for the identification of associations between cyberbullying involvement and psychological characteristics and not causal relationships. Longitudinal research is essential to elucidate the close bidirectional interactions between cyberbullying experiences and psychological traits, and the overlapping nature of VP roles. Fourth, the use of the SDQ was a limitation. The SDQ is a screening tool, not a diagnostic instrument, and has somewhat reported low sensitivity and specificity. Nevertheless, the SDQ is useful for detecting emotional and behavioral disorders because of its brevity and strong psychometric properties and is internationally recognized as a widely used tools related to child and adolescent mental health [41]. Consequently, this study’s findings regarding “conduct problems” and “emotional symptoms” must be interpreted as self-reported symptom levels and not as clinical diagnoses of conduct or depressive disorder. Future studies should use structured clinical interviews to validate these results.

CONCLUSION

This study investigated the cyberbullying experiences and associated sociodemographic and psychological characteristics of first- and second-year middle school students. The results showed that among adolescents involved in cyberbullying, experiencing both victimization and perpetration was more common than experiencing only one, and the VP group exhibited the most severe psychopathology. This suggests an urgent need for interventions targeting this specific population. Furthermore, our findings indicate that genderspecific approaches are necessary for cyberbullying prevention. Specifically, to prevent cyberbullying, it may be necessary to identify and therapeutically address conduct problems in adolescents, which may be particularly crucial for male students. For female adolescents, identifying and appropriately intervening in emotional difficulties, such as depressive symptoms, may be necessary to prevent their involvement in cyberbullying.