What it’s like to be a _____: why it’s (often) unethical to use VR as an empathy nudging tool

Our claim can also be construed in the following way: empathy-based VR nudgers achieve their nudging effects in virtue of a form of deception that we argue are manipulations of (non-conscious) System 1 and (conscious) System 2 processes. The perceived goods of empathy-based VR nudges almost never outweigh the unethical deception, loss of autonomy, and manipulation associated with VR EEEs especially in the light of available, and less problematic, alternatives to VR empathy-based nudging. Some authors have occasionally framed the ethics of nudges in terms of System 1 and System 2 manipulation (see Sunstein, 2015).

Augmented reality (AR) devices also exist and can be used as nudge vehicles. Instead of simulating an entirely new three dimensional world for users to inhabit, AR devices (like the new defunct Google Glass) overlay information onto our normal experiences (i.e., by providing a floating directional arrow to guide you through an unfamiliar city as you explore it). While empathy enhancing nudges have currently been developed for VR HMDs, what we say in this paper can extend to AR attempts to generate empathy as well (i.e., by changing how you perceive your own, or other people’s, racial or gender identity).

Citations omitted for review.

Citation omitted for review.

Citations omitted for review.

What we quickly learn from these results is that a simulation’s ability to generate virtually real experiences in its users is at least partially a result of the user’s psychology and background beliefs. Users who believe in time travel, ghosts, and dragons are more likely to experience simulations containing these elements as more context-real than users without these beliefs (citation omitted for review).

That a user treats an experience as virtually real need not indicate that they will continue to act as if that experience were real after-the-fact. What matters, though, is that subjects who are in the middle of the experience will react to it as if it were real. As a result, context-real and perspectival faithful simulations also share some features with context-real and perspectivally faithful dream experiences (Chalmers, 2017).

According to Sunstein, a person’s welfare is benefited when it makes them “better off, as judged by themselves” (Sunstein, 2015, p. 429). We wish to remain agnostic about how best to make sense of welfare for the sake of nudging as it seems sensible to claim that public welfare may reasonably render a nudge permissible even if those nudged may not be materially benefitted (as with opt-out organ donation nudgers).

Some might argue that transparency only matters instrumentally, insofar as it enhances autonomy and its absence undermines it. However, even if this is true, understanding whether and how a nudge is transparent or obscure is still helpful in assessing its ethical status.

It’s unlikely that transparency fully removes the nudging power of many nudges. For example, hyper-targeted political ads, intended to emotionally nudge users, may retain some of their nudging power even if the nudge’s nature is made transparent to subjects. In this sense, that a nudge is transparent (or not) does not fully determine the permissibility of the nudge. A nudge’s transparency is an important piece of the puzzle when assessing the ethics of a particular nudge but it may be outweighed by all things considered judgments of the nudge’s effects on autonomy and individual/social welfare.

That we can be nudged in this way is one reason why nudge ethicists have called for the development of conditions of competence and trust among nudge developers (Selinger & Whyte, 2010).

For example, urinals will sometimes have a fly printed on their surface as a way to nudge users toward aiming at the ideal place on the urinal to reduce splashing. Such a nudge may, arguably, be ethically deceptive in the all things considered sense. Why would this be? First, such a nudge isn’t obviously an educative nudge. The printed fly is deceptive in the sense that it might fool a user into believing, at least temporarily, that a real fly is on the urinal but the force of the nudge is derived from its value in providing users somewhere to aim on an otherwise featureless surface. An astute person may realize that the fly is merely printed on the urinal and nonetheless be successfully nudged by it. Second, such a deception (minor and short-lived though it might be) also is likely to align with the user’s own goal of avoiding splash. By contrast, “lying calories'' goes beyond deception and manipulates users by subverting their autonomy. We say more about deception and manipulation below when discussing the ethics of VR EEEs.

Emphasis added. We argue that no simulation can show human users what it is like to be a cow or to feel cow feelings or have cow thoughts. In part, we locate part of the problem with VR empathy simulations as resting on a fundamental mistake about the nature of experience that we see repeated throughout the psychological literature on VR embodiment. Though we say more about this later, we think it confusing to parse out how these researchers think human and cow experiences are related and individuated such that we can say that a person’s experience, in VR, of walking on all fours in a virtual pasture, is relevantly like an actual cow’s experience of walking naturalistically in a field. See (citation omitted for review) and (Nagel, 1974) for more on this general problem in the philosophy of mind.

Emphasis added.

Emphasis added.

Some nudges might avoid the problem of semantic variance altogether because they nudge their users on the basis of physiological or non-cognitive psychological capacities widely shared regardless of a user’s context. To the degree that such nudges are cognitively impenetrable (to the degree that a person’s thoughts do not affect the power of the nudge), then they would avoid the problem of semantic variance. Images of diseased organs on the cover of cigarette packs, for example, may trigger evolved disgust responses that nudge users to smoke less frequently regardless of their cultural context. Whether or not any nudges, including the one just mentioned, actually fit these criteria is a matter of significant controversy.

Emphasis added. Altough she does not use the term “structural intersectionality,” consider also Elena Ruíz’ (2017) framing of intersectionality in terms of experience: “As a descriptive term, [intersectionality] refers to the ways human identity is shaped by multiple social vectors and overlapping identity categories (such as sex, race, class) that may not be readily visible in single-axies formulations of identity, but which are taken to be integral to robustly capture the multifaceted nature of human experience” (335).

By primitive, we mean only that mirror-neuron empathic structures appear in many mammals, including rats, and thus are likely to have evolved early relative to other more complex cerebral structures found in humans and other apes (Carillo et al., 2019).

Emphasis added.

We pause here to note that the problem we’re raising for virtual reality simulations is focused specifically on those wishing to use VR as an empathy enhancing nudge. To use VR in such a way requires creating a simulation depicting someone’s experience with the intention of having that experience be shared with those who are intersectionality different from the person being simulated. VR simulations designed to provide a generic perspective (i.e., a VR simulation of sitting in a stadium during the Olympics) wouldn’t run into any of the ethical issues we’re raising here for empathy enhancing nudges. Similarly, it’s entirely possible for someone to design a VR simulation of their own experiences and intended for their own use. Insofar as the designer and user of a VR simulation like this are similar to one another, then self-empathy of this sort would also be relatively problem free. We say “relevantly” because there is some evidence that we can fail to empathize with our past selves, especially after undergoing transformative experiences (see Levine, 1997). For more on how structural intersectionality can impact empathy even across those who identify as members of the same race or gender see Táíwò (2020).

This is, we want to emphasize, an empirical matter. It may turn out that some intersectional features (race, gender, sex, etc.) may be more influenced by internalized bias than others (class, nationality). Our point in producing this example is to note that those who will be subject to (or whom subject themselves to) empathy based VR nudge simulations are unlikely, for these sorts of reasons, to successfully mirror the experience of those whose bodies they virtually inhabit. This is, in one way, to reiterate the concerns we raised earlier about empathic perspective-taking and the importance of subdoxastic features. See also Ruckmann et al. (2015) for a similar study using fMRI to assess empathic contagion.

Users can thus be manipulated in several ways. For example, a user can be manipulated toward forming (and/or acting) on inappropriate beliefs (i.e., that a food has substantially more calories than it really does) but they can also be manipulated if they come to form morally benign beliefs through inappropriate/ deceptive means (e.g., convincing a child that Santa Claus does not exist because he was murdered in the nineteenth century). In the case of VR empathy-enhancing simulations, we believe both forms of manipulation are present.

Insofar as designers of VR empathy-enhancing simulations accept even the weakest forms of structural intersectionality (i.e., that internalized concepts can structure the content of experience) then they ought to know that their simulations cannot succeed at the task they’ve designed them to do.

Another question, outside the scope of our argument here, is whether “lying calories” would be morally acceptable to use if we lacked a non-manipulative alternative. Any response to this issue would require a complex analysis of the value of the social goods that “lying calories'' would help realize. Once we account for all of the institutions that would be involved in maintaining the opacity of “lying calories” and the effects of creating and maintaining institutions designed to paternalistically manipulate public beliefs about health, we’re skeptical that “lying calories” could be justified, but it isn’t beyond the realm of possibility. We thank an anonymous reviewer for pushing us to clarify this aspect of our argument.

A corollary issue here is that manipulations of this sort will fail to make their subjects morally better people even if they succeed in changing their behavior. Writing about behavioral moral enhancements of this sort, John Harris (2013) writes that “…moral enhancement, properly so called, must not only make the doing of good or right actions more probable and the doing of bad one’s less likely, but must also include the understanding of what constitutes right and wrong action” (172). Because subjects of VR EEEs are using emotionally-laden false beliefs to change their behavior, it’s unlikely that they’re behaving, as Harris suggests, on the right kinds of reasons when they act.

It might be argued, however, that such nudgers are ultimately permissible. Why would that be? Because becoming more sympathetic with the plight of those experiencing homelessness or with members of other marginalized communities (1000 Cut Journey, Carne y Arena, etc.) would align with the values of the person using the simulation. That is, such nudgers may be used to manipulate oneself into a position that one ultimately desires and endorses upon reflection. The literature on self-deception is interesting, and vast, (Bermúdez, 2000; Bortolotti & Mameli, 2012; Lynch, 2016; Mele, 1997) but it seems that such responses miss the point of the critique. In order to properly consent to the use of such simulations, and in order for such nudges to sidestep the ethical issues we raise here, their function must be transparent to the user. Our claim here is that such transparency would break the simulation’s power as a nudge. If a subject knows, in advance, that the experiences they will have are not the experience of “what it’s like” to be the person represented by the simulation, then the simulation would cease to function as an empathy-enhancing nudge and thus, by one’s own lights, one ought not desire to use the simulation for empathy-enhancing reasons if one cares about the ethics of nudging.