Four separate experiments were conducted to examine how embodying non-human victims influences compassion fade and broad conservation outcomes. A 2-condition (victims: one vs. many) controlled lab experiment in Study 1 explored whether compassion fade occurs when humans are exposed to (virtual) non-human casualties in VR. Study 2 then sought to determine the requirements for BT with a non-human being (sea turtle), employing a 2 (visuomotor synchrony: yes/no) x 2 (visuotactile synchrony: yes/no) between-subjects experiment. After establishing the feasibility and effectiveness of BT with a Loggerhead sea turtle body in VR, a 3-condition (victims: none/one/many) between-subjects field experiment (Study 3) examined whether animal embodiment influenced compassion fade. Lastly, Study 4 employed a 2 (victims: one/many) x 2 (victim species: same/different) between-subjects experiment to determine whether the effects of embodiment on compassion fade observed in Study 3 were contingent on victim similarity.
For all four studies, experimental protocols were approved by the Committee for the Protection of Human Subjects (CPHS) and an Insitutional Review Board (IRB). Moreover, all methods were carried out in accordance with relevant guidelines and regulations.
Statistical Analyses
All statistical analyses were performed using IBM SPSS. Assessment of mean differences in donation amount for Study 1 employed a two-tailed t-test. Subsequent studies employed one-way and two-way analyses of variance (ANOVAs). Moderation and mediation analyses were also conducted within SPSS through the use of the PROCESS path-analysis macro. Significance was set at 0.05. Lastly, across all analyses, assumptions of homogeneity were met.
Methods: Study 1
To test the assumption that compassion fade can be observed during immersive depictions of mass casualty, a two-condition (victims: one/many) between-subjects experiment was conducted. Participants (N=60; Mage=20.97) were paid $5 to experience and evaluate an immersive public service announcement (PSA) wherein threats to loggerhead sea turtles were discussed in a hospital setting alongside one or many members of the species (victims).
Stimulus Materials
To present mass versus singular wildlife casualty in a plausible context, a 2-minute, room-scale VR PSA was created for use with the Oculus Quest HMD. Upon equipping the HMD, participants were able to explore a virtual sea turtle hospital modeled after an actual rehabilitation center in Florida (“The Turtle Hospital”). Upon entering the hospital, participants were either presented with one or many (7) injured Loggerhead sea turtles (see Figure 1). Narration and animations then informed participants of various threats facing the species (e.g., ocean plastics, boat strikes), after which the PSA ended, leaving the participant free to remove their HMD.
Procedures
Upon arriving at the lab, participants were informed that they would be asked to evaluate an environmental PSA using VR in exchange for $5. After providing informed consent, participants completed an online pre-questionnaire and were then randomly assigned to one of the two PSAs. After receiving instructions on how to use the HMD, participants then completed the PSA, completed a final online survey, and received payment.
Dependent Variables
To control for individual differences in environmental concern, environmentalism was measured using a 12-item 7-point Likert scale adapted from previous work61. Measures of presence were assessed using spatial presence62 and social copresence63 scales adapted from previous work. Additionally, attitudes towards the PSA were assessed using established scales64. Pro-environmental behavioral intentions were also assessed via a 7-point Likert scale asking participants to report the likelihood (very unlikely to very likely) that they would engage in five separate pro-environmental activities (e.g., Donate money) benefitting Sea Turtle conservation efforts.
Measuring Compassion Fade
After completing the final online survey, participants were escorted into an isolated room to receive payment. Once in the room, the researcher informed the participant that they would receive five (5) $1 bills in an envelope. Furthermore, participants were informed that they had the opportunity to donate all, some, or none of the $5 towards The Turtle Hospital by placing the desired donation amount in the envelope and leaving the envelope in a basket along with other sealed envelopes. This donation amount served as the key dependent variable.
To prevent demand effects, the researcher informed the participant that all donations are anonymous, and that the researcher would not know whether participants donated or not because the decision is made after the researcher steps out of the room. After exiting the room, participants were thanked and escorted out of the lab.
Results: Study 1
A series of Bonferroni corrected independent t-tests were conducted (see Table 1). Participants in the single victim condition donated, on average, more towards sea turtle conservation (M = 2.9, SD = 1.78) than those in the many victim condition (M = 1.86, SD = 1.96), t(58) = 2.13, p = .03, d = .53. Thus, compassion fade effects were observed.
Discussion: Study 1
As expected, compassion fade was observed among our sample (Figure 2), demonstrating how victim count in immersive depictions of environmental threats can influence human desire to mitigate those threats. Individuals who viewed the PSA with many (7) victims donated considerably less towards sea turtle conservation compared to those who viewed the PSA with one (1) victim, which is in line with previous work in non-immersive messaging2. To examine how embodying non-human victims may influence compassion fade, Study 2 investigates the requirements for BT with a non-human victim’s (sea turtle) body.
Study 2: Embodying Non-Human Victims in VR (250)
An individual’s body schema is maintained by neural networks that continuously map the body and its interaction with the environment65. The malleability of the human corpus is such that one can experience illusory ownership over an artificial limb if that limb (e.g., a rubber hand) is synchronously caressed with one’s hidden real hand (visuotactile synchrony)66, or if its movements synchronize with the person’s real-world limb movements (visuomotor synchrony)67. These principles comprise the rubber hand illusion (RHI) paradigm and can be extrapolated to explain how individuals may come to inhabit a virtual body and integrate it into their self-schema (BT), which ultimately influences how one perceives threats affecting their virtual body56,68,69.
Despite growing interest in BT and its effects, the requirements for establishing BT with non-human bodies remain ambiguous. For one, the majority of studies testing the effects of embodiment have done so using humanoid avatars46,51,70,71. Second, while the style (or realism) of one’s virtual body does not necessarily detract from BT72, morphological similarity to one’s body can influence on BT72,73. In the case of animal embodiment, one can identify the incompatibility of non-humanoid avatars, and how this may impede visuomotor synchrony (e.g., How does one control a virtual tail?). Yet, despite scant literature on animal embodiment, studies suggest that visuotactile synchrony (haptic feedback) may compensate for incongruent visuomotor feedback associated with embodying an avatar with a unique morphology74,75. Thus, Study 2 sought to clarify the individual and combined effects of visuomotor and visuotactile synchrony on BT with a sea turtle body in VR (animal embodiment).
Methods: Study 2
A 2 (visuomotor synchrony: yes/no) x 2 (visuotactile synchrony: yes/no) between-subjects controlled lab experiment tested the impact of the two primary sensorimotor contingencies on BT with a sea turtle body using a custom-built VR installation about sea turtle conservation (Figure 3). Participants (N=98; Mage=20.63, SD=2.71) were provided course credit in exchange for their participation. Of the 98 participants, 22.4% identified as male (Nmale = 22), 75.5% identified as female (Nfemale = 74), and 2% identified as other (Nother = 2).
Stimulus Materials
A VR installation (Figure 3) was developed using Unity 3D software for use with Oculus HMDs (e.g., Oculus Quest, Oculus Rift), a haptic feedback backpack (SUB-PAC), and a custom-built seating apparatus to allow for a non-bipedal posture while embodying a sea turtle. The installation, titled “Project SHELL (Simulating Living Habitat Experiences of Living Loggerheads)” was created in collaboration with sea turtle experts and allows users to embody a Loggerhead sea turtle at key stages of its life: hatchling, adulthood, nesting. As the user progresses through these stages, they encounter various environmental threats (e.g., marine debris, plastics, boat strikes) and experience their immediate and long-term effects.
Story
The version of the simulation used in Studies 2-4 focused exclusively on threats encountered during adulthood, which lasted between 2-5 minutes. Upon equipping the HMD, hand controllers, headphones, and SUB-PAC, participants emerge in a sea turtle hospital with two nurses, one of whom informs the user that they have been rehabilitated after a boat strike and are set to be reintroduced into the ocean. Afterwards, users are taken to a water tank where users familiarize themselves with the locomotion. The tutorial scene teaches users how to engage in a breaststroke motion to move in the direction they are facing. After learning the swimming mechanics, participants emerge in a virtual coral reef where they are tasked with finding food scattered throughout the coral reef. Users must also resurface occasionally to refill their oxygen. During this experience, participants must avoid various threats, including oncoming boats, netting, marine debris, and microplastics. After three deaths, participants re-emerge in the sea turtle hospital and are informed that the simulation has ended.
Visuomotor Synchrony
Participants in the visuomotor synchrony conditions saw all their body movements (i.e., head rotation, arm movements) synchronously reflected onto their virtual avatar (i.e., the 3D model of the loggerhead sea turtle). Furthermore, to ensure proper induction of BT with one’s sea turtle body, at the onset of the simulation, users would see themselves in a virtual mirror embodying the turtle avatar prior to beginning the simulation, a method used in previous BT studies76. Participants in the no visuomotor synchrony conditions were in a disembodied state, with only their hand controllers visible throughout the experience.
Visuotactile Synchrony
Synchronous haptic feedback was delivered throughout the simulation via the use of a haptic backpack (SUB-PAC). For example, at the onset of the experience, the virtual nurse strokes the user’s carapace, sending synchronous haptic vibrations to the SUB-PAC. Additionally, while swimming, users feel a steady vibration representative of friction associated with moving through water. The amplitude of the vibration spikes when the user encounters various threats, as when they are caught in netting or hit by a boat. Participants in the visuotactile synchrony conditions receive these haptic signals throughout the experience. Participants in the no visuotactile synchrony conditions did not receive haptic feedback at any point. However, these participants did equip the SUBPAC during the simulation, though they were informed that the device’s purpose was to improve the tracking of the HMD and the user’s body movements, serving as another sensor to ensure accurate body positioning and movement throughout the simulation.
Dependent Variables
As in Study 1, environmentalism was included as a control variable, with, attitudes, spatial presence, and copresence also assessed. Additionally, four new variables were included: BT, inclusion of nature in self (INS), threat perceptions, and conservation behavioral intentions. BT was measured via a 15-item, 7-point Likert scale assessing participants’ level of agreement with various statements about their virtual representation 26,42. INS was used to measure the extent to which an individual includes nature into their cognitive representation of self 77. INS was measured via a 1-item, 7-point pictorial scale demonstrating a series of 7 overlapping Venn diagram circles, with one circle representing the user and the other circle representing nature. Environmental threat perceptions were measured via a 7-item, 7-point Likert scale78. Lastly, donation intentions consisted of the reported likelihood that the user would donate money towards sea turtle conservation.
Study 2 Results
A series of one-way ANOVAs revealed no significant differences in self-reported concern for environmental issues (Cronbach’s α = .84; p > .05), spatial presence (Cronbach’s α = .95; p > .05), and copresence (Cronbach’s α = .56; p > .05) across both the visuomotor synchrony groups. With regards to BT, there were no main effects of visuomotor, F(1, 94) = .54, p = ns, or visuotactile synchrony, F(1,94) = 1.06, p = ns, on BT (MBT = 4.93, SD = 1.16) (Cronbach’s α = .93). Furthermore, there was no significant interaction between visuomotor and visuotactile synchrony on BT, F(1,94) = .72, p = ns. Thus, all participants experienced a moderately strong sense that they were embodying a sea turtle body (Figure 4).
A bootstrapped multiple mediation analysis with 5,000 resamples 79 was conducted, using the PROCESS path-analysis macro (Model 6) in SPSS, to examine whether the effects of BT on donation intentions were mediated by INS and threat perceptions. BT significantly predicted INS (b = .53, p < .001), though INS did not significantly donation intentions (b = .12, p = ns). Thus, INS did not mediate the effects of BT on donation intentions. With regards to threat perceptions, BT was a significant predictor of threat perceptions (b = .23, p = .007), and threat perceptions subsequently significantly contributed towards donation intentions (b = .54, p < .001). In sum, the effects of BT on donation intentions towards sea turtle conservation were partially mediated by threat perceptions (b = .12, BootSE = .12, BootCI = .02 to .27), but not INS (b = .05, BootSE = .06), 95% BootCI [.08, .18]. BT remained a significant predictor of donation intention despite having threat perception included in the model (b = .51, SE = .14), 95% BootCI [.24, .78]. Overall, the results lend support to the notion that embodying a non-human victim increases the salience of the threats to the group, which thereby contribute to charitable giving intentions (Table 1).
Table 1 Linear regression analyses of body transfer on key dependent variables in Study 2 across experimental conditions.
BT → INS
|
R2
|
B
|
SE B
|
β
|
t-stat
|
No Embodiment/ No Haptics
|
.29
|
.17
|
1.07
|
.54
|
3.09**
|
Embodiment/ No Haptics
|
.19
|
.61
|
.25
|
.44
|
2.36*
|
No Embodiment/ Haptics
|
.13
|
.42
|
.22
|
.36
|
1.86
|
Embodiment/ Haptics
|
.11
|
.34
|
.21
|
.33
|
1.62
|
BT → Threat Perceptions
|
R2
|
B
|
SE B
|
β
|
t-stat
|
No Embodiment/ No Haptics
|
.01
|
.06
|
.17
|
.07
|
.33
|
Embodiment/ No Haptics
|
.22
|
.37
|
.14
|
.47
|
2.58*
|
No Embodiment/ Haptics
|
.11
|
.33
|
.19
|
.34
|
1.71
|
Embodiment/ Haptics
|
.05
|
.21
|
.2
|
.22
|
1.02
|
BT → Donation Intentions
|
R2
|
B
|
SE B
|
β
|
t-stat
|
No Embodiment/ No Haptics
|
.00
|
.07
|
.29
|
.06
|
.26
|
Embodiment/ No Haptics
|
.21
|
.94
|
.38
|
.46
|
2.46*
|
No Embodiment/ Haptics
|
.17
|
.51
|
.23
|
.42
|
2.19*
|
Embodiment/ Haptics
|
.49
|
.98
|
.22
|
.7
|
4.53***
|
***p < .001. **p < .01. *p < .05.
|
Discussion: Study 2
Phenomenologically, 6 factors are identified as determinants of BT: body ownership, agency, haptics, body location, body appearance, and body responses to stimuli 80. The null effects of the visuomotor and visuotactile conditions discount the theory of a compounding effect of visuotactile and visuomotor synchrony, running contrary to literature showing that visuomotor synchrony is a greater contributor to BT than visuotactile synchrony, and that the disruption of either can contribute to a break in the illusion 42,74,80. The results also run counter to previous work finding that embodiment of an avatar, regardless of the degree of realism, elicits higher BT than having no avatar 81. Slater and colleagues previously stated, “we do not know whether the illusion would break down if there were changes in the topology of the body”42. The results also disambiguate this concern by demonstrating that, regardless of visual representation, BT with a non-bipedal creature is achievable.
The null effects of visuomotor and visuotactile synchrony on BT also suggest that proprioceptive factors (e.g., non-human locomotion), may have overridden the individual effects of these components. Even those without a virtual body assumed the posture, movement patterns, and morphology (e.g., carapace proxy via the haptic backpack) of the non-human creature. Previous work argues that body representation may not be present and still elicit embodiment when individuals have “an unconscious assumption that something similar to their regular body (image) persists within a given virtual environment”82. Thus, perceptual judgments of one’s body are differentially impacted by the combination of movement, visual, and haptic factors.
Alternatively, the null differences may be rooted in properties specific to how haptic feedback was delivered. Participants experienced a sustained haptic vibration while underwater (mimicking water friction), which spiked in frequency (strength) during collisions with environmental threats. The baseline vibrations (water friction) may have habituated users to haptic sensations created a ceiling effect. Indeed, previous studies examining synchronous sporadic versus consistent haptic feedback during embodiment26 lend support to this notion, though future work is needed to support this theory.
While BT significantly predicted INS and threat perceptions, only the latter subsequently contributed towards donation intentions. Of particular interest is the direct effect of visuotactile synchrony on INS. This finding underscores the affective nature of hapticity: touch is considered “a powerful conduit for emotional connectedness”83 capable of influencing human judgment and decision-making84,85, even in cases where the source of the haptic information is separate and unrelated to the object being evaluated86. Moreover, the valance of the haptic information itself influences judgment formation, as positive (negative) haptic sensory experiences can contribute (detract) from one’s evaluation of an object/experience87. However, the haptic information presented in the simulation signaled negative (boat strikes) and neutral (ocean vibrations) information. Given the main effect of hapticity on INS, and the absence of main effects on threat perceptions, BT, and attitudes, it suggests haptic feedback informed judgment of the species, as opposed to the simulation or threat.
Study 3 Methods
Study 3 explored whether animal embodiment’s influence on users’ affective, cognitive, and behavioral responses to an environmental threat vary as a function of the number of victims (compassion fade). A 3-condition (victims: self vs. one other vs. seven others) between-subjects field experiment was conducted at a large museum during November and December 2019. Visitors (N=90; Mage=18.95, SD=16.61) participated in the study voluntarily without compensation. Of the 90 participants, 56.7% identified as male (Nmale = 51) and 43.3% identified as female (Nfemale = 39).
Stimulus Materials
A designated area in the museum was provided for the Project SHELL installation, which consisted of three chairs, each with a SUBPAC haptic backpack, an Oculus Quest, and a dedicated researcher to assist with on/offboarding (Figure 5). Each HMD was pre-loaded with one of three modified versions of the simulation. The narrative structure of Project SHELL remained consistent: users embody a rehabilitated Loggerhead in a medical facility, are reintroduced into the ocean where they encounter various threats and are returned to the hospital upon completion (i.e., 3 deaths of 5 minutes in the ocean). The victim count manipulation was made salient in all three (3) sequences; users would either be the lone sea turtle (self-as-victim), see one other sea turtle, or see seven other sea turtles. Thus, one version of the simulation projected the threats solely onto the user, whereas the other two versions place either one or many victims alongside users. Behavior of the sea turtles (victims) was programmed within Unity to mimic natural navigation and collision avoidance within the ocean scene. Lastly, due to facility restrictions and liability, the seating apparatus used in Study 2 was replaced with a swivel chair.
Procedures
Project SHELL was promoted by the museum as a temporary installation, with signage and digital communications informing visitors of the opportunity to test a VR experience about sea turtles. Visitors who approached the installation were given information about Project SHELL, the simulation, and the opportunity to participate in the unpaid user study by answering an online questionnaire after completing the simulation. Visitors were given the opportunity to test out the simulation without participating in the simulation as well. Visitors who agreed to participate were then provided with an informed consent form to sign. For children, parental consent forms were provided and signed prior to participation. After an onboarding process, participants finished the simulation and subsequently completed a post-questionnaire measuring variables of interest.
Dependent Variables
Due to the nature of the field study, namely time constraints and the need to minimize line queues, several alterations made to the post-questionnaire. Several survey items were omitted and/or modified, with the exception of BT and INS, which were retained and deployed in their original forms. Environmentalism was removed, and the threat perception questionnaire was eliminated due to length. Instead, a single item measure of perceived extinction was used as a proxy (see below for a detailed description of new measures included in Study 3). Attitudes towards the simulation, while previously measured via a multi-item Likert scale, were instead assessed via a star rating system (1 through 7). Lastly, donation intentions were measured via a hypothetical scenario88. Participants were asked to imagine winning $50 and list how much they would keep, and how much they would donate towards loggerhead turtle conservation efforts.
Study 3 Results
A one-way ANOVA revealed that the manipulation of victim amount produced significantly different estimates of sea turtles encountered in the simulation across the three groups, F(1,87) = 24.81, p < .001. One-way ANOVAs also failed to detect main effects of victim count on attitudes (F(1,87) = 1.84 p = ns), BT (F(2, 89) = .22, p = ns, Cronbach’s α = .93), INS (F(2,87) = 1.04 p = ns), or threat perceptions (F(2,87) = 2.3, p = ns). Mean scores for all measured variables across the three experimental groups are shown in Table 2.
Compassion Fade
There were no main effects of victim amount on the hypothetical donations, F(2,87) = .84, p = ns. Moreover, there were no significant differences in pledged donation amounts between the many (M = 29.85, SD = 19.16) and single victim condition (M = 34.59, SD = 16.47), p = ns. The results suggest that exposure to mass casualty while embodying a victim can inhibit psychic numbing effects (see Figure 6).
A linear regression analysis demonstrated that BT significantly predicted hypothetical donations, R2 = .05, β = 3.27, SE = 1.48, p = .03. If indeed psychic numbing were at-play during embodied exposure to mass casualty, victim amount should moderate the effects of BT on donation intentions. A bootstrapped multiple mediation analysis with 5,000 resamples 79 was conducted, using the PROCESS path-analysis macro (Model 1) in SPSS, to examine whether the effects of BT on hypothetical donations were moderated by victim amount. BT significantly predicted donation intentions (b = 9.37, p = .025). However, there was no significant interaction between BT and victim amount (b = -2.77, p = ns), R2∆ = .02, F(1,86) = 2.48, p > .05. In sum, there was no evidence of compassion fade in scenarios where audiences (a) embody the victim, and (b) encounter many other victims versus a single identifiable victim.
Table 2 Mean scores across experimental conditions in Study 3.
|
|
N
|
Mean
|
SD
|
No Victims
|
What is your age (in years)?
|
29
|
17.33
|
15.47
|
Have you used VR before?
|
29
|
1.55
|
.51
|
Time in VR
|
29
|
6.34
|
2.72
|
Fish Encountered
|
26
|
23.65
|
18.98
|
Turtles Encountered
|
29
|
.34
|
.55
|
Perceived Interactivity
|
29
|
6.10
|
1.45
|
INS
|
29
|
4.93
|
1.94
|
Threat Perceptions
|
29
|
4.07
|
1.56
|
Stars
|
29
|
6.24
|
1.05
|
Donation
|
29
|
29.00
|
16.10
|
Body Transfer
|
29
|
4.79
|
1.14
|
|
|
N
|
Mean
|
SD
|
One Victim
|
What is your age (in years)?
|
27
|
16.30
|
14.08
|
Have you used VR before?
|
27
|
1.70
|
.47
|
Time in VR
|
27
|
7.11
|
2.49
|
Fish Encountered
|
25
|
21.64
|
17.89
|
Turtles Encountered
|
27
|
3.74
|
6
|
Perceived Interactivity
|
27
|
6.67
|
1.21
|
INS
|
27
|
5.33
|
1.59
|
Threat Perceptions
|
27
|
3.85
|
1.59
|
Stars
|
27
|
6.56
|
.69
|
Donation
|
27
|
34.59
|
16.48
|
Body Transfer
|
27
|
4.78
|
1.14
|
|
|
N
|
Mean
|
SD
|
Many Victims
|
What is your age (in years)?
|
34
|
22.44
|
19.08
|
Have you used VR before?
|
34
|
1.53
|
.51
|
Time in VR
|
34
|
7.06
|
2.39
|
Fish Encountered
|
33
|
20.97
|
15.07
|
Turtles Encountered
|
34
|
10.09
|
7.33
|
Perceived Interactivity
|
34
|
6.35
|
1.30
|
INS
|
34
|
5.53
|
1.44
|
Threat Perceptions
|
34
|
4.65
|
1.41
|
Stars
|
34
|
6.12
|
.91
|
Donation
|
34
|
29.85
|
19.17
|
Body Transfer
|
34
|
4.61
|
1.36
|
To examine whether BT influenced hypothetical donation amount through mechanisms proposed in Study 2, a bootstrapped multiple mediation analysis with 5,000 resamples 79 was conducted using the PROCESS path-analysis macro (Model 6) in SPSS; The model tested whether the effects of BT on donation intentions were mediated by increased INS and threat perceptions (perceived extinction classification). As in Study 2, BT significantly predicted INS (b = .67, p < .001), though INS did not significantly donation intentions (b = -.73, p = ns). Thus, INS did not mediate the effects of BT on donation intentions. With regards to threat perceptions, BT was not a significant predictor of threat perceptions (b = -.05, p = ns), nor did threat perceptions significantly contribute towards donation intentions (b = -.56, p = ns). In sum, the effects of BT on donation intentions towards sea turtle conservation were neither mediated by INS (b = -.49, BootSE = .82) 95% BootCI [2.13, 1.17] nor perceived extinction classification (b = .02, BootSE = .25) 95% BootCI [-.35, .69], though BT remained a significant predictor of donation amount (b = 3.27, SE = 1.48) 95% BootCI [.32, 6.21]. Overall, the results lend support to the notion that embodying a victim increases the salience of the threats to the group, which can thereby increase altruistic behavioral intentions.
Due to the large variance and range in participants’ ages, and the nature of the BT illusion, it is imperative to account for developmental differences in the sample. Of particular importance is recent work suggesting that, as humans progress through childhood and develop their own-body perception, people become less sensitive to illusory BT89. A moderation analysis was implemented using the Process Macro (Hayes, 2015; Model 1) with BT as the predictor, age as the moderator, and hypothetical donation amount as the dependent variable. BT significantly predicted hypothetical donation amounts (b = 7.73, SE = 2.14, p < .001). Furthermore, there was a significant interaction between BT and age (b = -.201, SE = .07, p = .003) such that BT’s positive influence on donation amount diminished with age (Figure 7).
Discussion: Study 3
Studies examining compassion fade have established that individuals contribute more real and hypothetical resources towards ensuring the well-being of single rather than many victims90. The results suggest that compassion fade can be offset at least in part by embodied perspective-taking. Individuals who embodied a virtual loggerhead turtle and experienced threats alongside many (7) turtles (victims) did not differ in terms of hypothetical donations towards sea turtle conservation compared to those who experienced the same simulation alongside one (1) victim or no victims.
As BT served as a significant direct predictor of giving behavior regardless of victim count, the extent to which individuals feel a non-human victim’s body is their own is a viable mechanism for offsetting psychic numbing. Yet, the inability to replicate Study 2’s partial mediation lends credence to cautionary notes about the interpretation of mediators (e.g., INS) in the context of BT48. A growing body of literature argues that mediation pathways, particularly in an experimental context, may be difficult to establish in a single study91. Indeed, when coupled with the fact that BT significantly (and directly) contributed towards giving behavior in both Study 2 and Study 3, it can be argued that the self-other overlap instantiated by BT is a mechanism for altruistic behavior in and of itself.
Identifying why BT, but not connectedness with the species (INS), influences hypothetical giving requires identifying key distinctions between BT and INS. INS measures the sense of connectedness to another species, whereas BT measures the psychological sense that one has become an exemplar of that species. A “connection to nature is a global concept, affected by many factors”92, encapsulating an enduring, high-level relationship between humans and broad aspects of nature (i.e., entire species or biomes). Conversely, BT is characterized by the immediacy and salience of the sensory experience as the virtual body. The absence of granularity in the INS measure limits the ability to parse out the impact of episodic experiences on these high-level assessments of the human-species dynamic. Thus, while BT may engender closeness with the species (INS), this closeness may not dictate subsequent processing of the threats and the motivation to mitigate them in the short term through charitable giving.
The non-significant indirect effect of BT on giving behavior through threat perceptions also raises questions about the instrument used to assess perceived threat severity. Whereas the threat perception scale in Study 2 assessed future impacts of various threats on the survivability of the species, Study 3 measured participants’ perceptions of the species’ current survivability. It can be argued that it is general knowledge that loggerhead sea turtles are not extinct, which would skew estimations. Additionally, because the experimental manipulation involved the presence of many turtles (versus none or one), participants may have taken the many victims condition as a heuristic of a thriving species. The encountered threats may indeed have been perceived as more severe due to experiencing them alongside other similar victims, though the extinction classification measure may have failed to capture this heightened severity.
Another important insight relates to the moderating role of age. Previous work has demonstrated that adults are less susceptible to body ownership illusions89. Study 3 extends this work to include non-human body schema as well: the strength of the BT illusion diminished as age increased. While not the primary focus of this investigation, the moderating effects of age emphasizes is an important insight for facilities like zoos and aquariums which provide immersive experiences for children and adults. Whereas mere embodiment may suffice in inspiring youth to take action, it may be that message factors (e.g., source credibility, gamification elements) may be better suited for older adults. Future work should continue to explore the interplay between demographic, technological, and message factors in the context of immersive conservation experiences like Project SHELL.
Study 4 Methods
Study 3’s results establish support for the notion that embodying non-human victims can offset compassion fade, though whether this effect is driven by connectedness with the victimized group or heightened threat salience remains to be seen. To address this, a 2 (victims: one vs. many) x 2 (victim species: similar vs. dissimilar) between-subjects experiment was conducted in the same controlled lab setting as with studies 1 and 2.
Participants
Twenty-five university students (N = 25) were recruited using an undergraduate research pool and were paid $5 in exchange for their participation. The mean age of participants was 21.44 (SD = 3.42). Of the 25 participants, 12% identified as male (Nmale = 3) and 88% identified as female (Nfemale = 22).
Stimulus Materials
As previously mentioned, the Project SHELL simulation used in Study 3 served as the foundational basis for the simulations used in Study 4. Specifically, four (4) different versions of the experience were ported to the Oculus Quest HMD, each representing one of the four experimental conditions. As in Study 2, participants used the simulation with the custom-designed chair. Major modifications to the content itself pertained to the introduction of the new victim species (i.e., bottlenose dolphins) elaborated upon below.
To create a plausible yet salient exposure to dissimilar victims, Study 4 implemented modifications to the simulation. As in Study 3, exposure to victims were made salient at three (3) key points in the linear embodied experience: (1) upon emerging in the hospital environment, (2) in the ocean environment, and (3) upon returning to the hospital environment. The major modification of these simulations involved swapping the loggerhead sea turtle models of the single and many victims with that of another marine species: The bottlenose dolphin.
The bottlenose dolphin was chosen as the dissimilar species for several reasons. First, sea turtles may plausibly encounter other dolphins in the wild. Second, the dolphin is of a similar size as the loggerhead sea turtle, though it exhibits ample visual differences which make salient the species’ aesthetic uniqueness from the user’s embodied species (sea turtle). Third, both species face similar threats and thus may be admitted to the same marine rehabilitation centers. However, because the hospital setting used in previous versions of the simulation (Studies 1-3) were framed as taking place in The Turtle Hospital, which does not rehabilitate dolphins on-site, another organization (Mote Marine Laboratory) had to serve as the host center. The 3D dolphin models, acquired via the Unity 3D asset store, and appropriate animations were adapted for use with these models.
Procedures
The procedure of this study was identical to that of Study 2 except for the inclusion of the donation behavioral measure included at the end of the study. This donation behavioral measure was implemented using the same procedure from Study 1, with one major exception: participants were told that any donations given from their $5 would go towards the Mote Marine Laboratory and their conservation efforts, as opposed to The Turtle Hospital in Study 1.
Dependent Variables
This study employed the same dependent and control variables as used in Study 2, and the inclusion of the donation behavioral measure as used in Study 1. Additionally, given time limitations in Study 2, perceived species extinction classification served as a proxy for threat perceptions. To assess whether this served as a valid substitute, it was included in this study alongside the original threat perceptions scale used in Study 2.
Study 4 Results
A series of 2-way ANOVAs and associated pairwise comparisons found no significant main effects of victim count or victim similarity on all dependent variables with the exception of actual donations. A two-way interaction effect between victim amount and victim similarity on donation amount was predicted. To test this assumption, a two-way ANOVA was conducted with victim amount and victim similarity as two independent variables, and donation amount as the dependent variable. The results demonstrated a significant two-way interaction, F(1,21) = 14.34, p = .001, such that compassion fade was observed for dissimilar victims as opposed to similar victims (see Figure 8). Users who encountered many dissimilar victims (MManyDolphins = 1, SD = 2.06) donated less towards marine conservation compared to those who encountered a single dissimilar victim (MOneDolphin = 2.67, SD = 2.06), indicative of psychic numbing, F(1,21) = 3.46, p = .07, partial ή2 = .14, d = .99, though this effect was non-significant. This effect was reversed for those encountering victims of the same species; participants who encountered many similar victims (MManyTurtles = 3.5, SD = 1.77) donated significantly more towards marine conservation compared to those who encountered a single similar victim (MOneTurtle = .4, SD = .55), F(1,21) = 12.28, p = .002, partial ή2 = .4, d = 2.36.
A series of linear regressions were conducted to examine the relationship between BT and connectedness with each specific victim species (INS). Among participants presented with dissimilar victims, BT was not a significant predictor of connectedness with bottlenose dolphins, R2 = .09, β = -.38, SE =.38, p = ns. However, among participants presented with similar victims, BT was a significant predictor of connectedness with loggerhead sea turtles, R2 = .49, β = .7, SE =.22, p = .008. Lastly, across the entire sample, BT was a significant predictor of connectedness with loggerhead sea turtles, R2 = .38, β = .77, SE =.2, p = .001. Thus, as in Study 2 and Study 3, BT significantly contributed to connectedness with the Loggerhead turtle species.
There were no main effects of victim amount, F(1,21) = .00, p = ns, or victim similarity, F(1,21) = .03, p = ns, on the perceived extinction classification of loggerhead sea turtles. Furthermore, there were no main effects of victim amount, F(1,21) = .12, p = ns, on threat perceptions (Cronbach’s α = .78). However, a main effect of victim similarity was observed, F(1,21) = 6.76, p = .017, such that participants perceived there to be more severe threats facing different (MDolphins = 5.93, SD = .79) rather than similar (MTurtle = 5.26, SD = .51) victims. A linear regression was conducted to examine the relationship between BT and threat perceptions. The results demonstrated that BT did not significantly predict the perceived probability and severity of environmental threats facing marine life (threat perceptions), R2 = .002, β = .03, SE = .15, p = ns.
Discussion: Study 4
The human tendency to group other beings into social categories to distinguish one’s self from others is a robust phenomenon in social psychology. This propensity for in/out-group classification may explain a key finding in this study: humans donated more towards marine conservation when they were exposed to many victims of the same species as opposed to one identifiable victim of the same species, whereas the reverse was true for dissimilar species. Humans favor similarity and behave more altruistically towards those with similarity cues93. As Tian and Konrath (2019) note, “similarity is an important psychological and social construct that usually affects individual attitudes, perceptions, and behavior”94. In their systemic review of similarity effects, evidence supports the notion that human donation behavior increases when beneficiaries of giving behavior are similar to the donor across a variety of personal attributes, including those which are temporarily induced (e.g., temporary group membership). Previous work has also suggested that environmental compassion fade “may differ across individuals as a function of whether non-human animals are viewed as in-group versus out-group members”2. The results in Study 4 lend support to the notion that (a) embodying threatened wildlife (i.e., victims) in VR can reverse psychic numbing effects (compassion fade), and (b) that this effect is contingent on victim similarity. Moreover, the finding that donations were greater among those who encountered many similar victims as opposed to one similar victim runs counter to previous work showing compassion fade occurs with in-group members95. While pairwise comparisons revealed the mean differences for the dissimilar victim conditions to be non-significant (p = .07), this is likely due to the study’s sample size and lack of statistical power.