Does Embodying a Divine Avatar Influence Moral Decisions? An Immersive Virtual Reality Study

Choosing what is morally right or wrong is an essential decision that has important personal and social repercussions. Although in many circumstances, the “right” choice appears clear and straightforward, in others, it may not be so immediate, particularly when deciding between two or more options may give rise to moral conflicts. Conventionally, to investigate how these conflicts are experienced and resolved, researchers describe hypothetical situations in which people must necessarily choose between two possible alternatives, both of which are followed by undesirable and negative outcomes. These conflict situations are referred to as moral dilemmas.¹ A well-known example is the Trolley dilemma, in which people must decide between pulling a lever to deviate the course of a runaway trolley and save a number of people, traditionally five, from being killed by the collision, but sentencing one person to death, and not taking any action.² Usually, people tend to prefer the first option, defined as “utilitarian,” over the second option, defined as “deontological,” i.e., they choose the option for which the least number of lives are lost.³ This dilemma is an example of “incidental” dilemmas, in that their consequences are foreseen, but unintended.^4,5 However, in other dilemmas, defined as “instrumental,” the killing of one or more people is a direct mean to save others’ lives. An example is the Footbridge dilemma, where the decision-makers must push a person off a bridge to stop the runaway trolley through his body and thus save the five people.⁶ In this case, the utilitarian option becomes much less frequent, and the deontological one appears more often in people’s choices.³ In fact, formulating the dilemma in such a way, the omission of action is considered the less serious and the more morally acceptable option^3,7 The omission of action is often justified by appealing to the “Do not play God” principle,⁸ whereby no one has the right to make a decision that does not respect the limits of one’s own social rank or function. Inaction is perceived to interfere less with the “natural order”⁹ and people explicitly report that they are not in the position to decide who lives and who dies.⁸ Interestingly, one previous work showed that the weight of the “Do not play God” principle might be reduced when people play roles that make them responsible for protecting others.¹⁰

In the Christian-Catholic culture, the right to protect, and to decide for people’s life is attributed to the ultimate moral agent, i.e., God, who is traditionally described as the supreme moral judge who watches over humanity with the power to interfere with people’s fate.^11,12

In this vein, with this study, we explored if “being” God might change how people solve moral dilemmas. To do so, we used immersive virtual reality through which people can feel a virtual body as if it is their own¹³ and, crucially, change their attitudes and behavior according to the body physical features of the avatars they are embodying,^14–23 a phenomenon called the “Proteus effect.”¹⁴ Hitherto, the effects of embodying specific characters have been explored in domains like cognitive improvement²² or psychological well-being.²¹ Crucially, however, the effects on the moral sphere are currently unexplored (See Supplementary Data for an overview of the Proteus effect). Thus, in this study, we expand current knowledge by investigating whether embodying a moral authority affects moral conflicts. In doing so, we presented participants with incidental and instrumental moral dilemmas and recorded their choices, the time to make their decisions, their physiological reactions during the decision, and their postdecision feelings, while embodying an avatar resembling the anthropomorphic representation of the Christian God (vs. a control human avatar). This avatar has already shown to affect people’s sense of power and invulnerability, by making them overestimate their physical abilities and physiologically under-react to events threatening their physical safety.²⁴ Since the most given justification of why people make fewer utilitarian choices is because they do not want to “play God,” we hypothesized that participants virtually embodying God would first produce a higher number of utilitarian choices, exhibiting shorter decision times, and second would lead to weaker emotional responses and less negative postdecision feelings. Moreover, we expected more pronounced differences in instrumental dilemmas compared to incidental dilemmas.

Preliminary analysis on embodiment illusion score is reported in Supplementary Data. We performed data analyses using R software³³ by RStudio.³⁴ We used the package lme4³⁵ to perform multilevel mixed linear regression analyses (LMM) and multilevel log-linear regression analyses (GLMM”)^36,37 for continuous and dichotomous dependent variables, respectively.

We performed GLMM analysis on Option, with the categorical variables, Avatar (God-avatar and Control-avatar), Type (Incidental, Instrumental), and their interaction as fixed effects. We performed separated LMM analyses on VAS1, VAS2, VAS3, VAS4, VAS5, VAS6, VAS7, Decision Time, SCR in Decision making, SCR postdecision, BPM in Decision making, BPM postdecision, with the same categorical variables Avatar, Type, and Option (Deontological and Utilitarian), and their interaction as fixed effects.

As suggested by Barr,³⁸ we built the random part with the maximal structure allowed by our data without incurring in convergence issues. To select the variables to include in each model as covariates, we set multiple null models with participants’ scores on the questionnaires + Gender + Religion + Religiousness as not-interacting fixed effects. As for the random structure, we set the random intercept over participants and the random intercept over dilemmas. Statistical significance of fixed effects was determined using type III Wald F test with Kenward–Roger degrees of freedom³⁹ and the Anova function from car package. Post hoc pairwise comparisons (Holm corrected) were performed using least-squares contrasts (lsc), as employed in R’s lsmeans package.

Consistent with the Proteus effect, suggesting that people tend to modify their behavior according to the avatar they embody, we expected the God-avatar would lead people to differently experience moral conflicts, mitigating the “Do not play God” principle and leading to more utilitarian choices, shorter decision times, weaker emotional responses, and less negative postdecision feelings. In contrast to our prediction, we found no significant effect linked to the avatar for any of the dependent variables we investigated. Two main factors might explain these null findings, one linked to the intrinsic features of the task, and one relative to the main measures employed.

First, to increase statistical power, we moved from a classical one-shot presentation to the presentation of 30 different moral dilemmas. In doing so, we asked participants to imagine 30 scenarios and repeatedly answer the same set of questions. This might have made the experiment too long and repetitive, producing mental fatigue in our participants and affecting both their involvement and any possible effect of the embodiment illusion. We tried to tackle this issue by presenting the moral dilemmas with randomly interspersed motor exercises, which served as breaks and as a way to remind and strengthen the embodiment illusion, but this expedient might have been insufficient.

Linked to this, a second problem regarded the use of a two-option response in the moral dilemmas: given the frequency of their choices, it is reasonable to think that participants wanted to be consistent with themselves, avoiding postdecision dissonance,^40,41 which in turn, might have covered any embodiment effect. We were aware that blatantly choosing between two options might have been an issue; thus, we considered an alternative method, which consisted of providing only the utilitarian option and then asking participants on a VAS how likely that option would have been their choice. However, this prevented us from investigating participants’ postdecision feelings related to a decision the participants consciously made after reading the consequences of both the options presented.⁴² In fact, we were more interested in exploring moral conflict in its complexity rather than focusing only on the resolution, by observing also emotions and postevaluations.

Unfortunately, we also found no avatar-driven effect on SCR and heart rate responses, which are more unconscious and less controllable. We speculate that, again, the repetitive nature of the task might have led participants to experience less distress after repeatedly facing similar dilemmas, paying less attention, and putting less effort into trying to imagine the various situations described. Results could differ greatly if participants were immersed in a virtual scene in which they could “act” the moral dilemmas and see the consequences of their action. In doing so, the task would require fewer mental resources, avoiding imaginative effort, and we could observe automatic responses rather than deliberate judgments.

Although the results did not confirm our hypotheses, this work is one of the few studies that aimed at investigating the effect of the embodiment illusion on the moral dimension.^43–45 Results did not offer definitive evidence of the nonexistence of the Proteus effect in this domain, as there are multiple factors that may contribute to explain these null findings (e.g., the task nature, duration, selected measures, little interaction with the avatar and environment). This is precisely why further investigation is required to explore the experimental conditions under which the Proteus effect works most effectively (e.g., lower cognitive load, greater realism of scenarios, less task repetitiveness) and to determine whether the use of avatars strongly related to the moral dimension—as the God-avatar—might influence one’s moral standards.

Importantly, we found compelling effects concerning dilemma type, chosen option, personality traits, and religion affiliation that we discussed in Supplementary Data, thus supporting and extending literature on decision making in moral dilemmas.

This work was supported by the BIAL Foundation Grant (258/20) “In God’s shoes: embodying the avatar of the supreme moral authority modulates psychophysiological indices of one’s own morality”. Moreover, the support of the Institut d’ Etudes Avancées, Paris, is gratefully acknowledged.