Virtual body ownership and its consequences for implicit racial bias are dependent on social context

Abstract

When people hold implicit biases against a group they typically engage in discriminatory behaviour against group members. In the context of the implicit racial bias of 'White' against 'Black' people, it has been shown several times that implicit bias is reduced after a short exposure of embodiment in a dark-skinned body in virtual reality. Embodiment usually leads to the illusion of ownership over the virtual body, irrespective of its skin colour. Previous studies have been carried out in virtual scenarios that are affectively neutral or positive. Here, we show that when the scenario is affectively negative the illusion of body ownership of White participants over a White body is lessened, and implicit bias is higher for White participants in a Black virtual body. The study was carried out with 92 White female participants, in a between-groups design with two factors: BodyType (their virtual body was White or Black) and a surrounding Crowd was Negative, Neutral or Positive towards the participant. We argue that negative affect prevents the formation of new positive associations with Black and distress leads to disownership of the virtual body. Although virtual reality is often thought of as an 'empathy machine' our results suggest caution, that this may not be universally the case.

Keywords: embodiment; empathy; implicit association test; racial bias; virtual reality.

Figure 1.

The scenario. (a) The participant embodied in a Black virtual body is in front of a mirror and carrying out embodiment exercises. (b) A participant in a White virtual body. (c) Showing the participant with the virtual reality equipment. (d) First-person view of the Negative crowd—note how two virtual women in the centre of the picture stare at the participant.

Figure 2.

Box plots for the questionnaire scores on agency and presence. Box plots showing the medians and interquartile ranges (IQR) for presence and agency. The medians are the thick horizontal bars and the IQR the boxes. The whiskers extend from max (min score, lower quartile – 1.5*IQR) to min (max score, upper quartile + 1.5*IQR). The questions are shown in table 1.

Figure 3.

Questionnaire scores on ownership (a) box plots showing the medians and interquartile ranges. (b) Bar charts showing the means and standard errors for the summative score yown.

Figure 4.

Posterior probability distributions for µ_own. The shaded areas show the 95% credible intervals.

Figure 5.

Posterior distributions of µ_atb conditional on the covariate preATB. The shaded areas show the 95% credible intervals.

Figure 6.

Implicit bias as measured by the IAT. (a) Bar chart showing means and standard errors for preIAT. (b) Bar chart for postIAT – preIAT. (c) Scatter diagram of postIAT on preIAT. (d) Bar chart of the residual errors of the regression of postIAT on preIAT.

Figure 7.

Posterior distributions of µ_iat conditional on the covariate preIAT. The shaded areas show the 95% credible intervals.