پیش بینی رفتار مزاحمت سایبری نوجوانان: تحلیل ریسک طولی

The current study used the risk factor approach to test the unique and combined influence of several possible risk factors for cyberbullying attitudes and behavior using a four-wave longitudinal design with an adolescent US sample. Participants (N = 96; average age = 15.50 years) completed measures of cyberbullying attitudes, perceptions of anonymity, cyberbullying behavior, and demographics four times throughout the academic school year. Several logistic regression equations were used to test the contribution of these possible risk factors. Results showed that (a) cyberbullying attitudes and previous cyberbullying behavior were important unique risk factors for later cyberbullying behavior, (b) anonymity and previous cyberbullying behavior were valid risk factors for later cyberbullying attitudes, and (c) the likelihood of engaging in later cyberbullying behavior increased with the addition of risk factors. Overall, results show the unique and combined influence of such risk factors for predicting later cyberbullying behavior. Results are discussed in terms of theory.

Cyberbullying is defined as, “… any behavior performed through electronic or digital media by individuals or groups that repeatedly communicates hostile or aggressive messages intended to inflict harm or discomfort on others” (Tokunaga, 2010, pg. 278) and is an emerging societal issue. Indeed, some studies indicate that over 30% of their sample has been cyber-victimized1 (e.g., Hinduja and Patchin, 2008, Huang and Chou, 2010, Li, 2007 and Ybarra et al., 2007). Barlett and Gentile (2012) argued that the best way to inform interventions aimed at reducing cyberbullying is to study the predictors of this “new” form of antisocial behavior. If certain psychological predictors of cyberbullying can be identified and replicated using a variety of research methods on different samples, then interventions can be developed to, hopefully, decrease the influence of these predictors to reduce the likelihood of later cyberbullying. The current study used the risk factor approach (Gentile & Bushman, 2012) to test the unique and combined effects of several possible cyberbullying risk factors using a four-wave longitudinal design on an adolescent sample in an attempt to further empirically elucidate the predictors of cyberbullying behavior.

Results3 Zero-order correlations Table 1 displays the zero-order correlations between the continuous variables as well as additional descriptive information about each questionnaire. Inspection of the correlation matrix shows several plausible cyberbullying risk factors. Of note, early cyberbullying attitudes significantly correlated with later cyberbullying behavior. In addition, anonymity and cyberbullying attitudes tended to be significantly correlated with each other within each wave of data collection. Furthermore, age did not significantly correlate with any of the variables. Table 1. Correlations between variables. Variable 1 2 3 4 5 6 7 8 9 10 11 12 1 – 2 .50** – 3 .18 .58** – 4 .45** .42** .68** – 5 .34** .58** .55** .42** – 6 .25* .65** .47** .42** .64** – 7 .24 .45** .59** .61** .63** .58** – 8 .21 .31** .59** .52** .50** .46** .60** – 9 .27** .34** .06 .18 .37** .27* .27* .09 – 10 .03 .14 .12 .23 .26* .29** .19 .12 .39** – 11 .03 .06 .06 .30** .23 .32* .36** .14 .36** .43** – 12 −.01 .11 .09 .35** .15 .24 .06 .02 .30** .57** .43** – Mean 3.93 3.99 3.83 3.60 16.21 16.82 15.43 15.34 8.94 8.26 8.45 7.38 StDev 1.75 1.82 1.93 1.18 5.14 5.97 5.09 4.94 2.97 3.73 3.68 2.78 Min 3.00 3.00 3.00 3.00 9.00 9.00 9.00 9.00 5.00 5.00 5.00 5.00 Max 13.00 14.00 16.00 9.00 30.00 40.00 29.00 36.00 18.00 25.00 21.00 17.00 N items 3 3 3 3 9 9 9 9 5 5 5 5 Range 3–18 3–18 3–18 3–18 9–45 9–45 9–45 9–45 5–25 5–25 5–25 5–25 Alpha .67 .56 .86 .70 .63 .73 .69 .59 .44 .75 .74 .60 5th Percentile 3.00 3.00 3.00 3.00 9.00 9.00 9.00 9.00 5.00 5.00 5.00 5.00 Median 3.00 3.00 3.00 3.00 16.50 17.00 15.00 16.00 9.00 7.00 7.00 6.00 95th Percentile 8.00 8.00 7.65 6.00 26.30 27.00 27.00 25.00 14.25 15.50 15.00 13.50 Note: 1–4 is cyberbullying frequency at Waves 1–4, respectfully; 5–8 is cyberbullying attitudes Waves 1–4, respectfully; 9–12 is anonymity attitudes Waves 1–4, respectfully. **p < .01, *p < .05. Table options Sex differences Table 2 displays the results of several independent samples t-tests that were conducted to determine whether sex differences emerged across the variables. Results showed no significant sex differences on any variable. Table 2. Mean sex differences (SD in parentheses) across various outcomes. Outcome Wave Males Females Nmales Nfemales t Cyberbullying frequency 1 3.95 (1.80) 3.91 (1.72) 41 55 0.12 2 4.17 (1.78) 4.02 (1.95) 30 44 0.32 3 4.17 (2.75) 3.72 (1.32) 24 36 0.84 4 3.88 (1.62) 3.43 (0.77) 24 37 1.44 Cyberbullying attitudes 1 17.17 (7.31) 15.49 (5.04) 41 55 1.60 2 17.18 (7.31) 17.40 (5.57) 28 43 −0.14 3 15.83 (5.21) 15.24 (5.29) 24 37 0.43 4 16.12 (5.90) 14.60 (4.29) 25 35 1.16 Anonymity attitudes 1 8.66 (2.74) 9.15 (3.15) 41 53 −0.80 2 7.67 (4.08) 8.72 (3.35) 30 43 −1.21 3 8.67 (4.42) 8.41 (3.20) 24 37 0.27 4 7.64 (3.51) 7.22 (2.41) 25 37 0.57 Table options Risk factor analysis Using the methods previously described, we first conducted a risk analysis testing the risk of scoring high on cyberbullying attitudes or behavior for each risk factor individually while holding the other possible risk factors constant. To do this analysis, one risk factor was identified (e.g., cyberbullying attitudes) and we first inserted the low risk (5th percentile) starting value into our prediction equation while holding the other possible risk factors at their median starting value. Then, we inserted the high risk (95th percentile) starting value into the same equation and compared the two output percentages (see Table 1 for all starting values). This procedure was repeated for each risk factor at each wave when cyberbullying attitudes and then cyberbullying behavior was the dependent variable. Results are presented in Table 3, which convey the likelihood of scoring low or high on the dependent variable as a function of each individual risk factor. Table 3. Percent likelihood of scoring high on the outcome variable at low (5th percentile) or high (95th percentile) for each risk factor individually. Wave 1 Risk Factor CB CB Attitudes Anonymity Sex Low High Low High Low High Male Female Outcome Wave 2 CB 33.46 48.56 9.43 79.75 31.76 37.17 37.26 29.86 Wave 3 CB 31.03 72.14 9.31 75.62 35.01 26.22 31.44 30.63 Wave 4 CB 33.57 93.35 25.07 46.42 15.62 65.37 30.86 36.39 Wave 2 CB Attitudes 18.85 31.54 NA NA 12.88 29.59 19.22 18.48 Wave 3 CB Attitudes 15.49 57.9 NA NA 8.01 32.74 15.7 15.29 Wave 4 CB Attitudes 10.56 65 NA NA 6.11 20.53 9.42 11.83 Wave 2 Risk Factor CB CB Attitudes Anonymity Sex Low High Low High Low High Male Female Outcome Wave 3 CB 27.05 54.19 13.15 53.2 29.4 18.47 28.07 26.06 Wave 4 CB 46.74 44.63 26.07 73.29 17.85 73.39 22.24 30.3 Wave 3 CB Attitudes 5.92 20.9 NA NA 4.46 18.25 8.42 4.12 Wave 4 CB Attitudes 20.2 49.78 NA NA 19.94 20.54 15.72 25.56 Wave 3 Risk Factor CB CB Attitudes Anonymity Sex Low High Low High Low High Male Female Outcome Wave 4 CB 49.31 94.95 32.14 80.42 43.5 71.27 51.12 47.5 Wave 4 CB Attitudes 7.6 42.12 NA NA 6.1 17.44 7.74 7.45 Note: CB = Cyberbullying. Table options It is clear that cyberbullying attitudes are a very important risk predictor of later cyberbullying behavior. For instance, the likelihood of being high on cyberbullying behavior at Wave 2 increased from 9.43% when cyberbullying attitudes were low to 79.75% when these attitudes were high. This pattern was fairly consistent across all waves (although the degree of difference between the likelihood of being a cyberbully at low vs. high risk of cyberbullying attitudes changed at different waves). Also, early cyberbullying behavior was an important predictor of later cyberbullying behavior, as expected. For cyberbullying attitudes, results showed that early cyberbullying behavior was an important risk factor for the development of later cyberbullying behavior. For instance, the likelihood of scoring high on cyberbullying attitudes at Wave 4 increased from 10.56% (low Wave 1 cyberbullying) to 65% when cyberbullying behavior at Wave 1 was high. These results were similar across other waves of data collection. Also, anonymity was an important predictor of the development of cyberbullying attitudes. Although the difference in low vs. high anonymity was not as dramatic as the difference when cyberbullying behavior was the predictor, results still shows a substantial increase in the likelihood of forming cyberbullying attitudes at low vs. high levels of anonymity. For instance, the likelihood of having high cyberbullying attitudes at Wave 3 increased from 4.46% (at low anonymity) to 18.25% when anonymity at Wave 2 was high. These results were fairly consistent across other waves of data analysis. Second, we used the risk analysis to determine the risk of scoring high on cyberbullying behavior cumulatively. To do this analysis, we determined the risk of scoring high on cyberbullying behavior at low risk of each variable (using the 5th percentile starting values). Then, we systematically changed one possible risk factor's starting value from the 5th percentile to the 95th percentile (high risk) until the starting value for all risk factors represented high risk. We repeated this procedure for each wave of data collection. These results are conveyed in Table 4 that shows the likelihood of being a cyberbully with the addition of one risk factor. Table 4. Percent likelihood of scoring high on the outcome variable for the cumulative risk factor analysis. Number of Risk Factors at Wave 1 0 1 2 3 4 Outcome Wave 2 CB 7.22 9.80 16.94 21.36 91.13 Wave 3 CB 10.76 11.14 41.90 32.23 93.49 Wave 4 CB 12.19 9.77 75.06 96.84 98.76 Wave 2 CB Attitudes 11.17 11.66 23.09 46.06 NA Wave 3 CB Attitudes 7.90 8.13 39.89 78.76 NA Wave 4 CB Attitudes 6.88 5.42 47.40 78.15 NA Note: CB = Cyberbullying 0 = Female, Low Cyberbullying, Low Anonymity, Low Attitudes. 1 = Male, Low Cyberbullying, Low Anonymity, Low Attitudes. 2 = Male, High Cyberbullying, Low Anonymity, Low Attitudes. 3 = Male, High Cyberbullying, High Anonymity, Low Attitudes. 4 = Male, High Cyberbullying, High Anonymity, High Attitudes. High cyberbullying attitudes were not analyzed for later cyberbullying attitudes outcomes due to the high percentages this analysis yielded. Table options Results from this analysis present a very clear picture of the impact of multiple risk factors in predicting later cyberbullying attitudes and behavior longitudinally from only Wave 1 predictors. Indeed, inspection of Table 4 shows that when all four risk factors were high at Wave 1, the likelihood of being a cyberbully at Waves 2, 3, and 4 were 91.13%, 93.49%, and 98.76%, respectfully. This suggests that if we have information about only four risk factors of cyberbullying behavior (sex, early cyberbullying behavior, cyberbullying attitudes, and perception of anonymity), we can predict one's future cyberbullying behavior with over 90% accuracy months later. Table 4 also shows the difference in the likelihood of later cyberbullying behavior with the addition (or subtraction) of a risk factor, which also sheds light on the relative importance of that single risk factor. For instance, the likelihood of being a cyberbully at Wave 2 increased from 21.36% (when participants were male, had previously cyberbullied, and felt anonymous) to 91.13% when early cyberbullying attitudes were high and added to the equation. Finally, results depicted in Table 4 also show the importance of early perceived anonymity in predicting the formation of later cyberbullying attitudes. For instance, the likelihood of scoring high on Wave 3 cyberbullying attitudes increases from 39.89% (for males who had cyberbullied before) to 78.76% with the addition of high anonymity perceptions to the equation.