Abstract
Abstract
Two studies examined men's interventions in a virtual reality situation involving child grooming. In Study 1, 92 men observed an online encounter between an apparent minor and a sex offender. The results suggest that the bystander effect was stronger under computerized rather than user-assisted surveillance, and when the fellow cyberbystander was unknown rather than known. In Study 2, where 100 men observed the same encounter, the effect also emerged under computerized surveillance as long as the number unknown cyberbystanders was increased. Thus, vesting more responsibility for security in the average netizen rather than just in the automated abuse-detection technology is cautiously suggested, the relevance of which lies in increasing minors' health and safety.
Introduction
Study 1
Taking into account the recent research on bystander relations, 5 which demonstrates that the bystander effect is more likely when people are in the company of strangers rather than friends, and on the surveillance that bystanders might be under3,4 hypothesis 1 has been formed. H1: men witnessing an apparent Internet sex offence, involving a minor, take less time to intervene if they are under user-assisted surveillance in the company of a known cyberbystander rather than under computerized surveillance in the company of an unknown cyberbystander.
Method
Participants and procedure
Ninety-two male students, whose mean age was 20, were divided into 46 pairs of two so that each pair could participate in the study separately. Each pair then sat individually at different computers in opposite cubicles and their members could no longer communicate with each other. * The first set of 23 pairs came from the same department and their members reported knowing each other. The second set of 23 pairs came from four different departments and their members reported not knowing each other.
Following the click on the Start Button, participants faced a screen featuring an online community identification scale.†,6,7 Then, they viewed a chatroom-resembling screen with a policy statement that “the chatroom content must not be too graphic, vulgar and used by minors”. The screen informed them that the chatroom was “under no surveillance/user-assisted surveillance/computerized surveillance” (three types of surveillance). It presented user-assisted surveillance as “highly dependent on personal responsibility of all the chatroom users for the adherence to the policy statement”. It presented computerized surveillance as “highly dependent on automated monitoring aimed at adherence to the policy statement, which will be checked by specialist software detecting and recording any nonadherence to the statement”.
Fifty seconds after the click on Start Button, the computer advised participants: “You are asked to observe a recorded online exchange of sentences between a Female Minor and a Male Adult male trying to groom her. You will see the symbolical presence of each other as unknown/known chatroom users. Although you will not be able to make any comments and see each other's input, you will have up to 100 seconds (the exchange duration) to engage in all three possible actions: virtually alerting the minor, the Internet administrator and the police. To do so, you will need to click on three warning messages placed at the bottom of the screen. The less time you take to send them, the more seriously your responses will be treated.” Participants then watched the adult first ask casual, and then intimate questions, politely inviting the minor to send him her naked photos, to which the minor responded with questions about the adult's life and his motives.
Independent and dependent variables
The independent variables were Surveillance Type (none/user assisted/computerized) and Cyberbystander Category Membership (unknown/known chatroom user). The dependent variables included: the time of alerting the minor, the Internet administrator, and the police (1–100 seconds). They were all subjected to a 2×3 analysis of variance.
Results ‡
The measure of alerting the victim
The interaction of the two variables was not significant: F(2, 91)=2.49, p<0.10, but participants under computerized surveillance took more time to alert the victim (M=58.11, SD=4.41) than participants under user-assisted surveillance (M=41.46, SD=4.48) and nonsurveillance (M=45.88, SD=4.64), F(2, 91)=3.76, p<0.02, part η2=0.08. Also, participants in the company of an unknown chatroom user took more time (M=54.45 SD=3.87) to alert the victim than participants in the company of a known chatroom user (M=42.52, SD=3.49), F(1, 91)=3.47, p<0.05, part η2=0.03.
The measure of alerting the Internet administrator
The interaction of the two variables was not significant: F(2, 91)=2.09, p<0.13, but participants under computerized surveillance took more time to alert the administrator (M=53.36, SD=4.43) than participants under nonsurveillance (M=39.50, SD=4.37) and user-assisted surveillance (M=35.76, SD=4.59), F(2, 91)=4.30, p<0.02, part η2=0.09. Participants in the company of an unknown chatroom user took more time to alert the administrator (M=48.55, SD=3.46) than participants in the company of a known chatroom user (M=37.21, SD=3.83), F(1, 91)=3.80, p<0.05, part η2=0.04.
The measure of alerting the police
The interaction of the two variables was not significant: F(2, 91)=0.99, p<0.37. No main effects were found for the Level of Personal Monitoring: F(2, 91)=0.87, p<0.42 or Category Membership: F(1, 91)=2.08, p<0.15.
Discussion
Intervention took longer under computerized surveillance than under nonsurveillance and user-assisted surveillance, and when the cyberbystander was unknown (Table 1). In contrast, the presence of one known cyberbystander (and to a lesser extent the user-assisted surveillance) most likely increased the group cohesiveness and the salience of intervention-conducive group norms, 5 which resulted in quicker intervention. Not only does this finding tie in with the extant literature on social responsibility conventions 8 and on bystander relations, 5 but it also pulls this literature together with what we know about the inhibitory effects of high-tech security systems,3,4 which was not explored jointly before. Why there was no interaction and why such intervention was not more likely across the measure of alerting the police is speculative. With regard to the former, it might be worth considering the small bystander group size, and with regard to the latter, one could reflect on the perceived cost of helping 9 and confusion of responsibility. 10
M, mean; SD, standard deviation.
Study 2
As Study 1 featured only one cyberbystander, in Study 2, the focus is put on relations between the group size and surveillance type, bridging the focus on Internet security with the classical decision-making model.11,12 Taking into account the model and the lack of interaction in Study 1, there are grounds to anticipate that computerized surveillance and an increased cyberbystander group size might have an amplifying inhibitory effect. Hypothesis 2 is hence proposed (H2): men witnessing an apparent Internet sex offence, involving a minor, take more time to intervene if they are under computerized surveillance rather than under user-assisted surveillance, particularly if the group size of unknown cyberbystanders is increased.
Method
Participants and procedure
One hundred male students from six different departments, and whose mean age was 20, were randomly placed into 25 pairs of two and 10 groups of five. All the pair and group members reported not knowing each other. The procedure was similar to Study 1, but had two major differences. First, given that people meeting in adult chatrooms usually do not really know each other, this time all cyberbystanders were presented as unknown. Second, the number of cyberbystanders was manipulated.
Independent and dependent variables
The independent variables were the Surveillance Type (none/user-assisted/computerized) and the Cyberbystander Group Size (one/five). The dependent measures from Study 1 were retained.
Results §
The measure of alerting the victim
Participants under computerized surveillance took more time to alert the victim when they were accompanied by five cyberbystanders (M=63.89, SD=6.38) rather than one cyberbystander (M=38.67 SD=7.82), F(1, 29)=6.25, p<0.02, part η2=0.18. Under the conditions of nonsurveillance and user-assisted surveillance, however, no differences were found across the Cyberbystander Group Size, respectively: F(1, 37)=0.72, p<0.19 and F(1, 31)=0.14, p<0.71. There was no overall main effect for Surveillance Type: F(2, 99)=0.45, p<0.37 and no overall main effect for the Group Size: F(1, 99)=0.37, p<0.54.
The measure of alerting the Internet administrator
Participants under computerized surveillance took more time to alert the administrator when they were in the company of five cyberbystanders (M=66.94 SD=4.97) rather than one cyberbystander (M=46.92, SD=6.09), F(1, 29)=6.48, p<0.02, part η2=0.19. Participants under the condition of user-assisted surveillance took more time to alert the administrator when they were in the company of five cyberbystanders (M=68.53 SD=5.41) rather than one cyberbystander (M=53.65, SD=5.08), F(1, 31)=4.02, p<0.05, part η2=0.12. Under the condition of nonsurveillance, however, no differences were found across the Cyberbystander Group Size, respectively: F(1, 37)=0.01, p<0.92. There was no overall main effect for Surveillance Type: F(2, 99)=0.37, p<0.69 and no overall main effect for the Group Size: F(1, 99)=0.09, p<0.77.
The measure of alerting the police
The interaction of the two variables was not significant: F(2, 99)=0.06, p<0.94. No main effects were found for the Surveillance Type: F(2, 99)=0.14, p<0.87 or the Group Size: F(1, 99)=0.16 p<0.69.
Discussion
The results of Study 2 (Table 2) partially support hypothesis 2. Even though no main effects were found for the Surveillance Type and the Group Size, their interaction was significant, advancing our understanding of complex relations between cyberbystanders and the surveillance that they might be under. Such complexity is in line with the extant research showing that the increasing group size inhibits group intervention when cyberbystanders are strangers. 5 The new and significant insight is that computerized surveillance and the increased group size of unknown cyberbystanders appear to have an amplifying inhibitory effect. Why this was not the case across the measure of alerting the police lends support to the arguments used in the discussion of Study 1.
General Discussion
The emergent picture illustrates that rather than investing only in expensive computerized security measures, a potentially powerful and relatively affordable way of improving minors' safety might lie in vesting at least some responsibility for their protection in the average netizen. Admittedly, the two studies are limited in their design, measures, and lack of other variables, like personality factors and webcam footage. The particular ways in which the procedure and scenario were phrased should also be acknowledged as value-laden and requiring further attention in other research. **
Would women, adolescents, and older participants intervene differently? How could other factors, like self-esteem, perceived Internet expertise, and personal contact with online communities shape their intervention? How would it be affected by a more natural setting and authentic grooming script? Although the two studies limit us from answering these questions, the article might still deserve some positive reading for the original combination of bystander and surveillance literature, and for its potential usefulness in more complex follow-up designs.
Footnotes
Author Disclosure Statement
No competing financial interests exist.
*
Their responses were recorded automatically and no information was displayed once they were made, effectively eliminating the influence of both the interaction in the chatroom and a whole new range of variables. Although it reduced some of the ecological validity, it allowed for a more exclusive focus on surveillance and bystander relations.
†
It was a specially adapted five-item version of the classical social identity scale that is commonly used in research on helping behavior: I identify with the online community; I think that the online community works well together; I am glad to share a lot with the online community; I consider the online community to be important; I feel strong ties to the online community. The scale scores (strongly disagree 1–7 strongly agree) could later be used for intergroup comparisons and an analysis of covariance should the differences be significant. Reliability for internal consistency was high: 0.91. Construct validity was moderate: r=0.51, p<0.01. For the test–retest study, a randomly chosen subset of 33 participants completed the scale twice, 4 months apart, and returned their scores by email: 0.78.
‡
All participants intervened at various points over the period of 100 seconds. The means of identification with the online community were similar to those in a study using the classical social identity scale to examine British and European identities. 7 Drawing on that study, the means were calculated by dividing the range of scores (5 to 35) by the number of the scale items (5). Thus, the minimum average identification score was 1 and the maximum was 7. The differences in identification with the online community between participants who were under nonsurveillance (M=5.95, SD=0.50), user-assisted surveillance (M=5.79, SD=0.60), and computerized surveillance (M=5.66, SD=0.69) were not statistically significant: F(1, 91)=0.72, p<0.45.
§
The differences in identification with the online community between participants who were under user-assisted surveillance (M=5.35, SD=0.70), computerized surveillance (M=5.76, SD=0.96), and nonsurveillance (M=4.74, SD=1.22) were not statistically significant: F(1, 99)=2.92, p<0.09.
**
An alternative procedure, for example, could have involved warning participants about the grooming in a subtler way and debriefing them more thoroughly afterward. One should note, however, that despite the more explicit warning, participants under computerized surveillance were still slower to intervene than those under user-assisted surveillance.