VR Study: Embodiment and Action Conformity

Action Conformity in VR Milgram Obedience Scenario

Explored how embodied perspective in a virtual reality (VR) sexual harassment (SH) scenario influences action conformity in a later VR Milgram obedience experiment.
Male participants were either embodied as a male in a group harassing a woman (Group), as the woman being harassed (Woman), or experienced an empty bar (Control).
One week later, participants acted as the Teacher in a VR Milgram experiment, administering shocks to a female Learner under group encouragement.

Participants in the Woman condition administered approximately half the number of shocks compared to the Group condition.
The Control group's shock administration fell between the Group and Woman conditions.
Embodiment promotes identification with the embodied role (woman or group).
Embodiment as the woman led to delegitimization of the group, reducing action conformity.
Ethical considerations: VR studies with positive intentions can yield unexpected, non-beneficent results.

Group pressure can lead to harmful actions, as seen in Asch's conformity experiments and Milgram's obedience experiments.
Milgram distinguished between signal conformity (verbal agreement without consequence) and action conformity (behavior causing harm).
Action conformity can result in people committing harmful acts they would not normally do alone.
Alternative interpretations of Milgram's paradigm focus on shared social identity and 'followership,' where identification with the experimenter increases obedience.

Objective: To assess whether embodying a woman experiencing harassment reduces subsequent shock administration in the Milgram paradigm.
Focus: Influence of embodied perspective on aggressive behavior and breaking in-group solidarity ('followership').
Embodiment: Substituting a person's real body with a virtual body that moves in sync with their real movements.
Virtual body ownership: Can alter attitudes and behaviors related to age, race, pain sensitivity, and more.
The study examined whether experiencing sexual harassment from the victim's viewpoint reduces potential harm to victims later.

Participants: 60 male participants with a mean age of 24.6 {\pm} 4.42.
Exclusion criteria: Contra-indications for VR (e.g., epilepsy).
Ethics: Informed consent, participants told actors were involved, shocks limited to non-lethal voltage; post-experiment VR scene showing the female avatar unharmed.
Scenarios: Bar (sexual harassment) and Shocks (Milgram paradigm).

Phase 1: Participant embodied as a man in a group, witnessing harassment of a lone woman.
Phase 2: Participant relived the scenario embodied as either another man in the group or as the woman.
*Conditions: Group, woman, control.

Participants acted as Teacher, instructed by three virtual men (same as in Bar) to administer shocks to a female Learner for incorrect answers.
Shocks increased with each wrong answer, with the Learner expressing pain and the experimenters urging continuation.

Weibull distribution: Used to model the number of shocks, fitting the idea of a ‘weakest link’ in individual stopping decisions.
The distribution of the number of shocks is taken as: ns \sim (\lambda \times Weibull{i,nshocks,1}) + ((1-\lambda) \times Weibull{i,nshocks,2}).
Bayesian analysis: Employed to analyze response variables simultaneously.

Number of shocks: Bimodal distribution, with a cutoff around shock 9 or 10.
Medians and interquartile ranges: Showed different patterns of response between Group and Woman conditions
The mixture model is defined as: \lambdai = \theta{c(i)} + \varepsilon_i where c(i) is the group the i^{th} participant is in.
Weibull parameters: Two distinct groups indicated by posterior distributions
Heart rate Variability: There was no difference in the influence of the conditions on HRV in the Shocks scenario.
Heart rate increased for the control group, decreased for Woman group.
Plausibility: participants who thought plausible were more likely to give low shocks, and high shocks otherwise.

A cautionary aspect to these results is that they are double-edged. Although the number of shocks administered in the Woman condition decreased compared to the Control condition, the number of shocks given by those in the Group condition clearly increased compared to the controls.

Embodying the victim of harassment can reduce subsequent harm, but reinforcing the harasser role can increase harm.
Plausibility of the VR scenario and identification with the Learner are critical factors.
Weibull distribution suggests stopping is based on a ‘weakest link’ situation.
Further research is needed to explore causal relationships and long-term behavioral changes.