I was approached by the Office of Naval Research (ONR) in the United States about a project that would involve virtual reality (VR). The problem discussed was the risks that people acting together in a group may carry out actions that they would never carry out individually and would consider as wrong. The group solidarity would overcome the moral sense of the individuals involved. I did not want to carry out a project concerned with military matters, but instead suggested that we concentrate on the issue of male sexual harassment of women. We would consider the question of how VR might be used to change the attitudes and behaviour of men who might engage in such acts. This would be structurally similar to the problem of group solidarity overcoming individual moral actions but in this different context.

The ONR accepted the project proposal and after a very rigorous ethical procedure, the project started. Several people were recruited for the project as shown in the author list. 

The scenario designed for the proposal was an experiment with men where each participant would first be in VR as one of a group with three other (virtual) men in a bar who engaged in sexual harassment of a women. Participants would  then re-experience that scenario again either as another one of the men (Group condition) or as the woman victim (Woman condition). The idea was that experiencing the scenario from the point of view of the woman would lessen their propensity towards sexual harassment, having been in that position themselves.  

The problem was how to measure the outcome.  We did not want to rely on a questionnaire because of the problem of social acceptability (people would be likely to give answers that they thought were the appropriate ones). Instead we used the famous and controversial Obedience experimental paradigm designed by Dr Stanley Milgram in the 1960s. In the obedience experiments members of the public were asked to give electric shocks to a stranger at the behest of the experimenter. Milgram’s interest was to discover whether people would obey simply because the orders to do so came from a person of authority. In one version of the experiment a surprising 60% of people were persuaded to give even lethal electric shocks to the stranger (actually a confederate of the experiment). This has been carried out before in virtual reality in 2006 and 2018. 

Our new version of the Obedience experiment involved participants being required to give shocks to a virtual woman. There were three virtual male experimenters (wearing white lab coats) encouraging the subject to continue to give shocks over the protests of the virtual woman. Participants took part in this Obedience scenario about one week after they had experienced the Bar scenario.   

As part of their experience all participants finally saw the virtual woman again, after all data had been collected, where she explained that nothing real had happened, and that she was fine. 

We found  that the participants who had experienced the bar scene as the woman victim gave the least shocks. They tended to withdraw from the experiment much earlier than those who had experienced the bar scenario the second time as another of the men, in fact they were most likely to withdraw as soon as the virtual woman complained about the pain, and asked to be let out. This was at around shock 8 or 9 out of a possible 20 shocks.

Although we had expected that those in the Woman condition would give less shocks, we were surprised about how many more shocks those in the Group condition gave, especially compared to another Control group of participants who had experienced the Bar, but with no events taking place there.  Once we saw these results we recalled those in the Group condition to have the Bar experience again but this time from the embodied perspective of the woman. Moreover, in a later follow-up it no ill-effects were reported. However, the experiment strongly illustrated the fact that a study designed with good intentions can nevertheless have some negative outcomes, in spite of all the rigorous ethical procedures leading up to it. There needs to be even more consideration from the ethical point of view for these types of intervention.

Of course showing that behaviour was influenced in virtual reality as a function of the embodiment condition (Woman or Group) is a useful outcome, offering promise that these types of intervention may lessen the propensity towards harassment and aggression. However, it would be better, if logistically quite difficult, to carry out a long-term follow-up to see how attitudes and behaviours in the real world might have been impacted. 

This image shows the woman in the bar scenario. Two of the group of men can be seen in the reflection in the window behind.