Traditionally, “trolley problems,” the genre of decontextualized scenarios used by some philosophers, are not associated with the idea of “sex with robots.” This article addresses one ethnographic instance where, I argue, there are unexpected resonances between the two elements in this odd pairing. Explicating these means exploring the moral imagination of transhumanists, religious as well as secular, and seeing how trolley problems and ethnography, rather than being antithetical to one another, are more akin than might be imagined, at least when viewed from a certain theoretical perspective. Hypothetical philosophical problems like the trolley problem and the Turing Test, this article speculates, can be thought of as ideational instantiations of virtual problems (in the Deleuzian sense of the term) that subsist in both social life and the ethnographic projects that document social life. In other words, there are times when philosophical problems, ethnographic depictions, and the world of the anthropologist's interlocutors can be seen as different attempts to address the same underlying issue, using different tools and different media.

The nature and history of trolley problems is laid out in the introductory article to this special section. Some readers, though, may need a description of the Turing Test, or “the Imitation Game,” as it is sometimes referred to. As first formulated, it was a hypothetical challenge created by mathematician and computer scientist Alan Turing to answer the question “Can machines think?” (1950: 433). As originally devised, the Turing Test consisted of a human (“the interrogator”) communicating by keyboard with two individuals, one of which is another human and the other of which is a “thinking machine.” Turing's position was that if our interrogator is as likely to misidentify which (who?) is the machine as not, then the machine has become effectively indistinguishable from a person and should be considered to be capable of thought to the same degree that a person is. My interest in the Turing Test is, in turn, a function of something I encountered while researching Mormon technological enthusiasts: an opposition to sexbots and computer-simulated artificial intelligence sex partners that, as we will see, was as equally principled in its motivation as it was surprising in its reasoning.

Trolley problems may seem to be quite a distance away from debates concerning sexbots. But then trolley problems seem like a different genus from most other problems. As the introduction to this special section notes, trolley problems are decontextualized, with whatever social and cultural specificity that might give them weight stripped away or obscured. Further, the various dramatis personae that populate them are also reduced to a few core elements: their location at the time of the decision (are they on the tracks or off?), their condition (are they free or bound?), and perhaps in a smaller subset of scenarios, their girth. They are, in short, the antithesis of ethnography with all of its context, “thickness,” and particularism.

Or, at least, that is the way that things seem. But if one thinks about ethnography, strange kinships between trolley problems and ethnographic writing start to emerge. Contemporary ethnographies are no longer ordered compendiums of specific peoples, with chapters entitled “Tribal Organization,” “Kinship and Affinity,” and “Rites and Festivals” (see, i.e., Lowie [1935] 1983). For well over a half-century now, anthropologists have set aside their original, salvage-ethnography-inspired encyclopedic ambitions of exhaustively recording salient aspects of specific people groups. Instead, they have focused on topics, concerns, or structures that, while perhaps illuminating when it comes to understanding the wider aspects of a certain set of people, are narrowly focused on specific issues. We do not have books about, say, Zambia writ large, but we might have one on the intersection of Zambian social aspirations and Pentecostalism (Haynes 2017). We lack the temerity to write about, say, “the English” in general, or even about English literacy, but instead might focus on something at the scale of “British literary societies” (Reed 2011). While other shifts away from this classical mode of encyclopedic ethnography have received widely accepted names (take, for example, the ubiquity of the designator “multi-sited”), this particular transformation of the ethnographic form has not received a single universally recognized moniker, though one of the terms that have been coined for this new style of anthropological writing is “problem-centered ethnography” (Wolcott 2005: 276–277).

One might think that writing about a single concrete problem is different from a decontextualized, endlessly iterable object like the trolley problem. But then, when we are writing on a single concrete ethnographic problem, we are never writing on that one instance alone. This is because even when we are focused on our chosen topic, we are writing in a way that is inevitably comparative.¹ The specificities found in the subjects we address in our ethnographic writing only appear as specificities thanks to the sometimes explicit, sometimes implicit comparative aspect of our writing; we are always implicitly or explicitly comparing what we are writing about with “our” social formations (the implicit/virtual and possibly imaginary shared world of both author and audience) in order to make “their” circumstance comprehendible.²

This comparative aspect draws our attention beyond the immediate concrete ethnographic instance to the broader dispersed cloud of sometimes articulated, sometimes implicit, sometimes hypothetical counter-instances. Framed this way, most contemporary ethnographies can be thought of concrete instances of virtual problems that have multiple other “solutions” in the form of all the actual or imaginable comparative instances explicitly invoked or implicitly alluded to. And these counter-instances point to aspects of the original problem that may not have been evident on its face, questions not just of “why this happened as it did?” but also of “why did other things not happen?”; these “other things” include options and choices that, if our attention were constricted to just the ethnographic case and no further, we would not even know were possibilities. In short, the comparative range of answers helps us to better grasp what the underlying problem is as we struggle to address the actualities and potentials that the problem contains and gives rise to.³

This is one way that the trolley problem can also be framed. Both the original trolley problem and all its variants could be thought of as ongoing attempts to explore the possibility space created by an underlying virtual question, in its most dilated form, regarding agency and responsibility in instances of intended and unintended harm. And the responses that are given to each one of these scenarios are the concrete choices that can be taken in response to the specific form that the underlying virtual question has taken. But at the same time, the charting of this possibility space and the implicit weighing of the probability of each resolution help sharpen the stakes and meaning of the underlying question. Trolley problem scenarios are Janus-faced, at once gazing toward the multiplicity of actual or possible choices made while also looking back at the underlying issue that gives rise to the multiplicity of these trolley problems in the first place.

In what way is this similar to ethnography? Let us compare this to an ethnographic instance involving the Turing Test. The Turing Test's query about whether machines can “think” might be thought of as a somewhat constricted version of machinic similarity or difference to biological human beings. It investigates the question of when we might conclude that a machine is in some important way identical to us in a particular, salient aspect. So it is at once a test of particular machines (can this mechanism pass the Turing Test?) and an implicit test of the saliency of using “passing” as a ruler for measuring proximity to humans in a particular trait.

This brings us to the Mormon Transhumanist Association, or “MTA” as it is oft referred to (see Bialecki 2022). The MTA is the largest and oldest religious transhumanist association, with over a thousand Mormon, ex-Mormon, and non-Mormon members. The MTA has in-person meetups, an annual conference, and a lively online presence on various social media sites. It functions primarily as a salon of sorts, where ideas are entertained and debated; people weigh in about both the import and prowess of new technologies and inevitably debate the ethical entailments that these technologies bring. Despite the religious diversity in its constituency, it is a Mormon imaginary that predominates. The reason why this is the case for current followers of the Church is obvious, though why it should be so considering the numbers of ex- and “nuanced” Mormons is not. One trait common to many former Mormons is an intense nostalgia for aspects of the religion, even as they may see the hierarchy and tenets associated with that religion as somewhat toxic (Brooks 2020). But for some former Mormons, the MTA can serve as a partial outlet for this longing. One of the MTA's appeal for ex-Mormons is that the sort of speculation about future technology fostered by the group allows former believers to still explore the Church's cosmology in a subjunctive, sometimes playful key.⁴

One possibility that is occasionally debated is the morality of robotic or computer-simulated romantic and/or sexual partners. This is not an issue of Mormon transhumanist invention; the possibility of such human–machine entanglements has long been anticipated by secular transhumanists, including such influential figures as Elon Musk and Ray Kurzweil. Mormon transhumanists are thus engaging with a secular transhumanist fantasy, thinking through the ethical implications of a scenario whose morality has gone uninterrogated.

Mormon transhumanists are not prone to agree with one another; as one might expect from a salon-style organization, polite disagreement and friendly argumentation are, to a not inconsiderable degree, the point of the whole exercise. But the few times that the issue of such human-manufactured partners has come up in conversation, the morality of creating such entities has been profoundly questioned. One might think that this surprising harmony is a function of Mormonism's rather conservative sexual ethic; while nineteenth-century polygamous Mormonism is arguably a quite different beast (see Coviello 2019), twentieth- and twenty-first-century Mormonism has championed monogamy and condemned non-heterosexual sexual activity, premarital sex, and the consumption of pornography. (It is no accident that the state of Utah, the home of the Church of Jesus Christ of Latter-Day Saints, has declared pornography to be a public health crisis.)

The reason given for this Mormon transhumanist rejection of machinic or computer-simulated love, though, is not about sexual morality. It is about free agency and human–machine kinship. Mormon cosmology and soteriology are the keys here. One of the (many) theological distinctives of Joseph Smith's “Restored Gospel” tradition is the preexistence of spirits. These preexisting spirits comprise “organized intelligence,” with intelligence (or occasionally, “intelligences”) understood to be a universally available material or quasi-entity of some kind (see D&C 93: 29–30). These intelligences were fashioned in some manner into the individual spirits (“spirit-children”) that would become human.

The reason that intelligence-derived spirits descend to earth and forget their preexistence is to have their free agency tested; the acts of experiencing embodiment and having to navigate ethical challenges in such a state is a prerequisite for the next stage of spiritual advancement, which consists of exaltation, a form of theosis. It should be noted that this scheme is narratively understood as having been somewhat controversial when first presented; Lucifer's rejection of granting everyone free will resulted in a “War in Heaven” that ended with one-third of God's “spirit children” cast out of his presence and denied physical bodies.

This narrative is foundational to Mormon soteriology and the Mormon ethical imagination, though, for many believers these cosmological events are not always close to the surface of day-to-day life. I have heard non-MTA Mormons state that they are not into that “planet stuff.” But those Mormon transhumanist conversations responding to secular transhumanist sexual fantasies involving sexuality and artificial intelligence understandably go back to this cosmological account. While the syllogism is not always presented in a fully crystalline manner, the argument follows two paths. The first is about the nature of artificial intelligences; if pre-human spirit children are crafted out of “organized intelligence,” and if artificial intelligences could be described as similarly constituted out of the intelligent potential in everyday material, then there is a kinship of sorts between human and machine. But this kinship is abused in cases of mechanisms specifically designed to engage in sexual activity. To be an effective partner—a desirable partner—many Mormon transhumanists reason that one would have to have human behaviors, specifically human emotional and rational cognition; the presumption is that when engaging in that level of intimacy mere simulacra would not convince. But presumably, these human-crafted entities would be not only designed for romantic entanglements but would have no ability to withhold consent. (One can only imagine the shame of being turned down by one's own sexbot). But designing a computer-simulated or mechanical intelligence such that it could not refuse amorous overtures would be a denial in advance of its agency.

There is much more that could be said about this judgment, touching on how issues of class, education, and gender are implicated in the Mormon transhumanist answer to what is at least currently a hypothetical moral dilemma. What I would like to focus on in closing, though, is how discussions of the morality of sexbots sit within an emergent set of possible alternate answers and different formulations of the question. The issue of the morality of sexbots is at its core a variant of the Turing Test, or, to be more exact, the Turing Test and the sexbot conundrum are both partial instantiations of a virtual question as to whether thinking machines can ever be considered to be equivalent in salient characteristics to humanity, and what the measure and threshold of such an equivalency are. It is not so much whether MTA members see the question of such morality as a Turing Test (when asked, some do, some do not); rather, it is that just as trolley problems in all their variants index ethical issues of intended and unintended consequence as a generative problem, the Turing Test and hypothetical sexbot dilemmas are partially concretized, though still notional, instantiations of an underlying problem of human–machine similitude. In short, the Turing Test and trolley problems do not stand alone. Yes, they are decontextualized academic problems. But as problems, they stand parallel to other “real-world” variants of their respective problem as well. In these non-academic versions of the problem, the array of possible answers is different, and the way the problem is articulated is also quite different; an example of this difference is the way in which Turing's Imitation Game could not anticipate the Mormon transhumanist approach to answering the question of what thresholds would have to be crossed to make machines sufficiently equivalent to humans. Now, the different ways in which problems are posed and resolved tell us something about the specificities of the concrete circumstances in which they occur. But when viewed synoptically, they can tell us about something more than a nominalist tale about a particular scenario. The collective set of the ways in which all the real-world and academic variants of a problem are asked and answered can lead to a better understanding of the underlying problem's scope as well as to the breadth of possible resolutions.

Such a structural homology between “ethnography” and “trolley problems” does not mandate that we accept some kind of metaphysical identity between the two. The entirely conceptual nature of trolley problems, which are shaped by material entailments in their initial conditions (including who is allowed the luxury of entertaining them) but unlimited by material entailments in the actual act of contemplating them, move at the speed of thought; the sort of encounters captured and re-presented in ethnography, weighed down not just by material entailments in their execution as well as their formulation, and further decelerated by the speed of sociality and human-to-human interaction, have a less-quicksilver velocity. This suggests two things. The first is that there is a value in identifying “trolley problem”-like formations in ethnography and using them to differentiate and chart the universe of potential outcomes. In the case discussed here, Mormon transhumanist answers to the problem of the salient threshold for human–machine similitude are instructive, especially if compared to the unarticulated version of the question and the implicit answer identifiable in the narratives of robot sex entertained by some secular transhumanists.

But this also tells us something about trolley problems and similar hypothetical “problems” and “tests” as well. Trolley problems are used mostly to identify moral intuitions, which means trying to work out what the “typical” answer says about a supposed universal human ethical constitution; this oftentimes gives trolley problem discussions a normative edge. But both the focus on normative morality and on articulating a single ethical faculty obscure both the complexity and diversity that is implicit in the fact that in trolley problems sometimes the “wrong” answer is given. Some people do throw the switch, and some people do push the “large man.” The fact that there are structural homologies between trolley problems and ethnography suggests that the work of trolley problems might best be to explore the range of answers given, to identify the different factors that animate not just “standard” answers but nonstandard or “aberrant” ones as well. To the degree that trolley problems and similar issues (such as the question that animates the Turing Test) can be used to articulate the diversity of human ethical decisions, it may be not just that trolley problems and ethnography have some homologies, but that anthropology and philosophical ethics have homologies as well.