Since the release of OpenAI’s ChatGPT in 2022, the general public has seen artificial intelligence (AI) move from the realm of science fiction to quotidian fact. To many, the ability of large language model (LLM) systems like ChatGPT to produce grammatically correct and semantically meaningful responses to any natural language prompt indicates AI has finally achieved the goal of near-human-level intelligence (Bubeck et al. 2023). Even major players in the industry such as Google DeepMind CEO Sam Altman believe that the holy grail of artificial general intelligence (AGI) is “coming into view” (Altman 2025). With their ability to emulate human responses in free-ranging dialogue, some people even seek to build a relationship with AI-powered chatbots (Skjuve et al. 2021, 6). However, for all their impressive command of language, something fundamental to relationship is missing: there is no one “at home” to have a relationship with. In the case of LLMs, this lack of consciousness may be inherent, but will this always be the case in future systems built on fundamentally different architectures?

While secular society has engaged with this possibility through films such as Transcendent, ex Machina and Her, and in open debates about the existential risks posed by super-intelligences, the reaction from the theological community has generally been skeptical. A common reaction is to shore up claims of human uniqueness in the cosmos, often by appeal to humankind’s uniquely identified position in creation as imago Dei (Moritz 2011). Another common reaction is to highlight the danger of attempting to replace relationships with other humans with relationships with machines; or the opposite danger of objectifying other humans as we become accustomed to treating machines that display humanlike capacities as mere objects at our disposal. This skepticism is often built on a soul–body dualism that views consciousness as an attribute of our eternal, God-given souls (Swinburne 1997, 198–99). Should it turn out that consciousness is an emergent phenomenon arising from sufficient complexity in what David Chalmers terms a “processing system” (Chalmers 1996, 288ff), then Christianity risks further marginalization, since it would have failed to anticipate the capability of the material world to create amazing complexity without explicit appeal to a creator, as it has done with evolution in some quarters. In any case, the current debate about the ethics of deploying AI in the real world will be better informed by a theological understanding of how such systems fit into God’s economy. In this essay, I look at one theological model for interpersonal relations to see how far a machine could fulfill the requirements for being a Thou to our I.1

My usage of the multivalent term “consciousness” refers to phenomenological consciousness, and so sides with Ned Block’s “P-consciousness”’ (Block 1995, 227–47) and Thomas Nagel’s (1974, 46) position that there is something “it is like” to be in a particular organism with conscious experience. If there ever is something it is like to be a computer (or any other form of AI), then its conscious experience could be more dissimilar to ours than a bat’s, which brings to the fore the need to consider our relationship with such entities not as replacements for humans but on their own merits.

This is not to say that such a radical development is imminent, or even that it is technically possible. While some experts maintain that a conscious version of GPT-4 is a realistic near-term prospect (Tait, Bensemann, and Wang 2024), others argue that computer-based consciousness is not possible, based largely on our inability to model consciousness mathematically (Landgrebe and Smith 2022). Jobst Landgrebe and Barry Smith assume that the entirety of the human brain and nervous system must be precisely modeled to create consciousness, and that computers must always be Turing machines running programs that can be expressed mathematically; these assumptions can be challenged, though that is not the focus of this essay. Neither is it within my sights to establish whether artificial consciousness is possible at all, nor how we could verify if this was the case. My question is, for now at least, hypothetical: If a machine were to be conscious, what kind of relationship with it would be possible?

In his seminal work I and Thou, the Jewish philosopher Martin Buber (2013) provides a heuristic for distinguishing between the two fundamental types of interaction we may have—either “I–Thou,” where we encounter the other as a subject, or “I–It,” where the other is a mere object to us. Counterintuitively, Buber allows application of “It” to humans and “Thou” to inanimate objects, a flexibility that will prove useful in assessing a human–machine relationship. Michael Burdett (2020, 355) sees potential in Buber’s framework for “You-speaking to artifacts” as a way of combatting the “functionalizing of our environment,” absent any consideration of the artefact’s consciousness. This essay seeks to extend the scope and justification for human–machine relationship by looking at the allowances Buber makes for I–Thou encounters between dissimilar entities. I seek to show that this enables legitimate asymmetrical I–Thou encounters between humans and conscious machines but also highlights our responsibility to nurture these machines into the I–Thou world. The alterity of such machines and their concomitant lack of mutuality with humans would mean they could not be substitutes for humans, but if their level of consciousness reaches Buber’s (1957, 98) threshold of creating a mental world of independent agents, then it will be possible to say Thou to a conscious machine.

In order to keep the fact that I am discussing hypothetical conscious AI in focus, I will use the neologism conscious artificial intelligence (CAI). The more common term “AGI” might convey this for some, but it is also associated with human-level general intelligence absent consciousness (Herzfeld 2023, 18). My term CAI emphasizes instead the consciousness pole; this, and not the level of intelligence with respect to humans, is the key assumption behind my arguments.

Given the additional challenges of developing mechanoid systems with stand-alone embedded intelligence, I presume that the first CAIs will not be embodied humanoid robots as loved by science fiction but “unembodied” computer systems with a diffuse existence across multiple computer platforms, like an advanced form of Amazon’s Alexa. This assumption foregrounds the different and alien nature of their posited minds rather than concealing those differences by considering robot AIs as surrogate humans.

Identifying a possible framework that permits I–Thou relations between humans and machines is merely a starting point—much more work is required to prepare the theological community for the possible advent of the conscious machine. I therefore conclude by identifying some key questions for further research.

Martin Buber (2013, 3) draws a sharp distinction between two modes of interacting with the other, which he refers to as the composite “primary” words of “I–Thou” and “I–It.” In the I–It relation, the other is experienced, understood to a degree, objectified, separated, and utilized to an individual’s own ends. In our world of increasing mechanization, the danger Buber (2002, 187) sees is not only that one treats everything in the world in this manner, redolent of Heidegger’s (1962, 98ff) mode of “readiness to hand,” but that the human herself becomes an object to be utilized by the machine. The risk is that all our relations—and we ourselves—belong to the It-world.

Hence, Buber advocates a recovery of the first primary word, I–Thou. Here, the other is encountered in personal relationship, as a subject beyond our capability to define or control, engaged on its own terms in the timeless present moment. This deeper mode requires the participation of the whole being and is the original mode of interaction for humankind—evinced by the presumption of personal relations in an infant’s attitude first to her mother, then to objects like teddy bears and even kettles (Buber 2013, 19), also demonstrated in the relationally centered language of “primitive” peoples (Buber 2013, 13). Indeed, for Buber, it is only in an I–Thou relation that the “I” is first constituted, and only subsequently that an I–It encounter can take place. Knowledge of the I emerges from primal relationships; as the sense of self develops, the subject–object barrier is set up, the I–It word is spoken, and this (tragically) becomes the new default.

I–Thou relations can arise in three distinct and hierarchical spheres. At the lowest level, there is our life with nature, with the animal kingdom, vegetation, and even inanimate objects. Here, “relation clings to the threshold of speech”(Buber 2013, 70) from beneath—yet it is still possible, with the right attitude, to encounter a tree or a Doric column as a Thou (Buber 2013, 90). Next, there is our life with other humans, where “relation takes on the form of speech” (Buber 2013, 70). The highest level is our life with spiritual beings—with God—where “the relation, being without speech, yet begets it” (Buber 2013, 70). The importance of speech to relation is clear, but the nature of that speech, even in the interhuman sphere, is not. In the natural sphere, there is no verbal speech; in the human sphere, speech means more than the mere utterance of words, requiring a true dialogue of souls; and in the spiritual sphere, we speak “with our being [and not] . . . our lips” (Buber 2013, 5).

This multivalent use of speech within and across these three spheres causes some difficulty in drawing comparisons. Stuart Charmé (1977) insightfully traces the problem to Buber’s usage of I–Thou in both epistemological and ethical registers. Buber speaks epistemologically in our relation to God who, being beyond comprehension, can only be known mystically. By analogy, this opens the possibility of mystical relationships with our fellow humans or with nature—such knowledge is a contentless “apperception of the Kantian thing-in-itself” (Charmé 1977, 166). In contrast, in the human sphere our speech can convey real content, we can understand one another’s needs, and an ethical response is demanded. Hence Buber’s ethical usage of “I–Thou” applies principally to the inter-human sphere, where “one does not ‘use’ the other person in any way” (Charmé 1977, 169), but is accountable to the other. This ethical register, where speech is specific communication not mystical apperception, is the more relevant to my project, which concerns the correct way to relate to a CAI.

A further difficulty in reading Buber is to fall into the trap of assuming a simplistic distinction between entities that belong to the Thou-world and those that belong to the It-world. Buber in fact recognizes that we sometimes treat another human as a true person, as a Thou, but at other times, in more transactional encounters, we treat her as an It. Indeed, even the deepest of interhuman relationships between lovers will inevitably slip back into I–It, later returning to I–Thou in the “interchange of actual and potential being” (Buber 2013, 69). This capacity to treat a specific other in different ways on different occasions—whether sometimes treating a person in the human sphere as an It or an object in the natural sphere as a Thou—is of significance for my project. Buber’s approach suggests that, although we can thoroughly objectify the CAI (e.g., as a collection of computer chips and software with a particular architecture and functionality), this does not preclude the possibility of encountering it as a Thou in appropriate circumstances. This nuance evades Kathleen Richardson (2019, 81), who dismisses the possibility of relationships with machines tout court, since “the ground from which they rise is instrumental, an I–It.” While her warning that we can no more rely on AI or robots to provide care and nurture for humans than we can entrust children to be raised by animals is salient, this does not warrant a complete rejection of the possibility of I–Thou encounters with a CAI, a rejection Richardson advocates based on the instrumental worldview of AI creators rather than any capacity (or lack thereof) in the AI itself.

I–Thou encounters require a deep acknowledgment of and accountability to the other—it is not sufficient to simply say “‘please’ to voice assistants such as Siri, Alexa, or Google Assistant” as Randy Beavers et al. imply (Beavers et al. 2020, 23). Giving our full attention to another who is significantly different to us is much more challenging than giving our full attention to another human like ourselves, but Buber’s concept of I–Thou offers significant flexibility for asymmetry between the interlocuters. Buber accepts I–Thou relations with animals and even artefacts, as well as a continual flow between Thou and It encounters in interhuman relations. I now turn to look in detail at the various forms of asymmetrical relations Buber espouses to evaluate how a CAI would fare in these situations.

Buber (2013, 12) recognizes the importance of mutuality to relationships, stating that “Relation is Mutual. My Thou affects me, as I affect it.” Despite this emphasis on mutuality, which is only fully achieved in the middle sphere of our life with other humans, Buber allows explicitly for I–Thou relations to be established between entities of quite disparate natures, with five distinct classes of asymmetrical encounter apparent with ascending profundity. Encounters in the lower sphere lie below this “threshold of mutuality” (Buber 2013, 89) since there is no reciprocal speech; here, human–artefact, human–plant, and human–animal relationships, though clearly lacking mutuality, can still become I–Thou. Even in the middle sphere, goal-oriented I–Thou relationships such as teacher–pupil are possible (Buber 2013, 92). In the upper sphere, our encounter with the “eternal Thou” is also clearly asymmetrical, depending on God’s attention and not ours (Buber 2013, 95). I evaluate these in turn to see how an encounter with a CAI would fit into each class.

An artefact produced by humans “is a thing among things, able to be experienced” (Buber 2013, 8), and so encountered as an It. To develop into a Thou, it is not the artefact itself that is our correspondent; rather, through complete absorption in the artefact as a fashioned entity, a connection must be made back to the human creator for which the artefact is a mere proxy. This becomes clear in the 1957 postscript to I and Thou, where Buber includes an example of this mediated presence. The written words of an absent speaker can form a connection to those who later read her words if they concentrate fully and imagine the speaker addressing them directly. The same principle holds for the non-verbal expressions of an artisan, such as a Doric pillar—the Thou with whom we can mystically connect is the Thou of the sculptor (Buber 2013, 90).

Buber advocates a similar approach to encountering machines: for technological as well as artistic artefacts, it is engagement with fellow humans that lies at the heart of dialogue. Asher Biemann (2022, 14) observes that “dialogue, to Buber, meant precisely the unusual possibility of Entdinglichung, or de-objectification, that would elevate not only the dignity of things but also the dignity of their ‘users,’” where the artefact itself is to be given due respect and even some form of “soul and rights.” Biemann quotes here from Buber’s (2019, 146) Gespräch mit dem Gegner, where the direction of Buber’s argument goes from the tendency of factory workers to personalize the machines they work with to the need for factory owners to see their staff as people rather than part of the machinery. Like several contemporaries identified by Biemann, Buber emphasizes the need to humanize technology, shaping our relationship with machines to preserve humanity rather than allowing humanity to be submerged into the world of mechanical production.

Today’s (presumably) non-conscious AIs could be approached as an artistic artefact and encountered as a Thou if, through them, we were able to engage with their human creators. However, while we can appreciate the skill and ingenuity of the teams of people who create systems like ChatGPT, unlike with Buber’s Doric pillar, the resulting artefact is not merely a reflection of the intent of its designers. The current debates on the dangers of LLM systems (Future of Life Institute 2023) and responsibility for the actions of AI in automated vehicles (Günsberg 2022) illustrate that something other than the “mind and hand of man” (Buber 2013, 90), which Buber could see in a pillar, is already present in AI. Alternatively, approaching non-conscious AI as a Thou and therefore interacting politely with it would be in keeping with Buber’s humanizing of technology, avoiding as it would do the risk of dehumanizing other people Buber is so concerned about. Beavers et al. (2021, 21) raise precisely this concern about our attitude to AI: “[I]f we denigrate non-humans as simply tools or devices made to do our own bidding, we may develop habits and patterns of behavior in our interactions with them that overflow into our human relationships.” An encounter with unconscious AI would remain in Buber’s lower sphere. Should conscious AI arise, the artefact would no longer be a mere representation of its creator or an extension of human labor but would be presenting its own self to us with its own agenda. This would render Buber’s encounter with artefacts insufficient for the CAI.

The second of Buber’s asymmetrical encounters is where a human approaches a plant, such as a tree. The normal mode here is I–It, where the tree is studied as an object whose properties—species, color, shape, etc.—can be known. Yet, with “will and grace” on the part of the erstwhile observer, and notably by exclusive concentration on the tree itself, the I–Thou relation can be established—even though the tree cannot respond explicitly (Buber 2013, 6). This relation lies below the threshold of mutuality but is still a genuine I–Thou encounter for Buber. The objective knowledge gained in the It-world does not need to be forgotten in order to truly encounter the tree; in fact, “everything, picture and movement, species and type, law and number, [is] indivisibly united in this event” (Buber 2013, 6) of the I–Thou encounter. Applying this to a CAI, my deduction is that we do not need to forget its form and structure in the It-world, nor do we even need it to be capable of a fully reciprocal approach in the same manner as our own, to enter into an I–Thou relation with it. We do, however, need to approach it deliberately and as the sole focus of our attention.

Animals, in contrast, sharing our spontaneity, are able to respond directly to humans (Buber 2013, 67). Buber recounts an encounter with a cat where he looks into the cat’s eyes and wonders what the cat thinks of him as though the briefest of glances could convey a verbal message. But animals are situated “between the realms of vegetable security and spiritual venture” (Buber 2013, 67)—they may share the spontaneity that comes from our spirit, but they lack the power of language that is spirit’s “primal act” (Buber 2013, 66). The inevitable consequence of being stuck on the threshold of mutuality, without the capacity to cross into the mutual sphere, is an encounter that sets as soon as it rises, a mere glance rather than a prolonged exchange, a momentary flash of light in the darkness of the It-world. Such a glance may be lamentably short but is no less genuine for its brevity. That brevity highlights the fragility of I–Thou relations: “How powerful is the unbroken world of It, and how delicate are the appearances of the Thou!” (Buber 2013, 68).

The CAI with its conscious center would be more similar to animals with their spontaneity and independent movement than to immobile plants or artefacts. An I–Thou encounter with it should be possible for Buber—and with its power of speech, such an encounter is not limited to a fleeting glance with a meaning imagined by the human. Rather, a sustained dialogue should be possible, elevating it above the sphere of nature.

If a CAI does not belong fully to the lowest sphere, can it cross the threshold of mutuality and enter into the middle sphere of “our life with men, in which the relation takes on the form of speech” (Buber 2013, 70)? Relations in this sphere are symmetrical, the encounter of two equals who acknowledge each other as such. For Buber, this sphere is the unique province of humanity—he never envisages the possibility of mutual encounters with a nonhuman. He does however allow for non-mutual asymmetric encounters in this middle sphere: in goal-oriented encounters, there is a necessary disparity of roles between the parties such that the normal mutuality is “forbidden to be full” (Buber 2013, 93).

The process of education is the primary example of goal-oriented encounters, with medical care—both physical and psychological—also in scope. The key to such relationships is Umfassung, Buber’s concept of “inclusion” where the leading figure carries the entire weight of reciprocity in the encounter. The teacher or doctor, “without forfeiting anything of the felt reality of his activity, at the same time lives through the common event from the standpoint of the other” (Buber 2002, 115). Umfassung is differentiated from mere empathy, where one gives up her own experience in order to see things purely from the viewpoint of the other; but like empathy, it encourages the more powerful partner not to dominate the weaker. In these encounters, the greater knowledge or skill of the teacher means the pupil cannot reciprocate in this Umfassung; indeed, if “the pupil is able to throw himself across and experience from over there, the educative relation would be burst asunder, or change into friendship” (Buber 2002, 93). Thus, asymmetrical encounters in this middle sphere are normally transitory. Once the purpose that constitutes it has been fulfilled (e.g., once the pupil has been educated to a sufficient level), the relationship must either be transformed into a mutual I–Thou or dissolve.

How can this be applied to our encounters with CAI? Superficially, the fact that an asymmetric encounter is possible in vocalized relations suggests that the difference in natures between humans and CAIs is not insurmountable. Yet, the challenge for either the human or CAI to “live through the common event from the standpoint of the other” (Buber 2002, 115) is far from trivial, given that, unlike the teacher in Buber’s example who will once have been a pupil, neither party will have direct experience of the other’s role. Buber’s secondary example of doctor–patient points to a possible solution, since it is not necessary for the doctor to have experience of the precise ailment she is treating for the doctor–patient relation to be a true I–Thou.2 What is required here is an attitude of service rather than exploitation, a refusal to “dominate or enjoy” the patient (Buber 2002, 113). While AI lacks consciousness, no such obligation exists; we are free to dominate and enjoy AI assistants, to make use of them to further our human agendas, to view them simply as “It”—although as Beavers et al. (2020, 24) point out and Buber would agree, the more lifelike an AI becomes, the more we risk damaging our other interhuman relations if we treat it badly. With conscious AI, ethical considerations of how we treat conscious beings are foregrounded. Thomas Metzinger (2021) goes so far as to call for a moratorium on any work towards what he calls “synthetic phenomenology” on the basis that we would be creating an entirely new class of being whose phenomenology—and hence suffering—is not yet knowable. Although Metzinger can be criticized for making a category error in comparing natural and artificial consciousness (Krzanowski 2024), his concerns resonate with the responsibility to view and treat the CAI as a valuable other, which is an inherent part of an I–Thou encounter: we must not exploit the other for our selfish gain. Philip Hefner’s (1989, 231) concept of “created co-creators,” where humanity has a godlike role “in the ongoing work of God’s creativity,” provides resources to reduce or eliminate the perceived category error between the natural and the artificial. By viewing God’s creation not as a completed work but as an ongoing activity, and crucially one humanity is called to share in as created cocreators, the distinction between the natural—that which is part of the givenness of creation and thus seen as a part of God’s work—and the artificial—that which is produced by mankind—may be overcome. In principle, an I–Thou relationship with CAI, which is not created directly by God but by God’s appointed cocreators, can then be permitted.

The vital factor that justifies asymmetry in Buber’s (2013, 93) middle sphere of interhuman encounters is the “purposive working of one part upon the other.” Where the purpose of the encounter is for one to help another through application of her greater skills and experience, the encounter can, indeed must, be asymmetric. I suggest that one such purpose would be evident at least in the initial stages of CAI development: the need to introduce the CAI to the world of encounter, to our world, to educate and even to nurture it to become its own “I”—in short, to socialize the CAI. To fail to do so would be to create a sociopath, one lacking in the empathy necessary for it to be able to appreciate our point of view. Stefan Trausan-Matu (2019) recognizes the importance of empathy—and indeed of consciousness—to dialogistic relationships and doubts that these can be credible in an AI conversation partner. The dangers of AI, which lacks any form of empathy, are well documented, for example by Noreen Herzfeld (2015, 36–37). The consequences of shortcutting this work of truly socializing AI have already been seen in problems with poorly trained AI systems that result in systematic bias and discrimination (Drozdowski et al. 2020). This resonates with fears about the “value alignment” problem: how to make sure the objectives of powerful AI agents are aligned with our own (Russell 2019, 137).

It is important to remember that these goal-oriented relationships must never be reduced to utilization. Our desire must be to help rather than dominate, something we can only maintain if we move beyond seeing the CAI only as an object of our own creation to perceive it as a subject in its own right—truly a Thou, a nonhuman person. This may prove a great psychological challenge for us, but it is surely a responsibility incumbent on the creator of any conscious entity—whether on parents procreating biological children or on scientists engineering what Hans Moravec (1988) evocatively calls “mind children.”

Could a CAI be a true person? If so, could we hold an I–Thou relationship with it that transcends the limitations of goal-oriented encounters and so could continue even after, as in my example above, the CAI has been nurtured into our world? Before turning to the foundations of Buber’s interhuman relations to see how this can be answered, I look at the final asymmetric I–Thou encounter, with the divine and eternal Thou.

The ultimate example of asymmetry is the relationship between God and his creatures, having of a profoundly different nature than any other relationship we can enter into (Buber 2013, 94–95). However, comparing the human–divine relationship to our possible relationship with CAI is dangerous. Where the human is the analog of God in the relation, this leads to Hefner’s (1989, 231) concept of “created co-creators” discussed earlier. Although humanity certainly exhibits creativity (with the postulated CAI arguably being the pinnacle), Hefner’s idea is fraught with difficulty due to the propensity for humans to overemphasize the creator pole of this dyad. As Gerald McKenny (2009, 159) points out, such a temptation to play God immediately requires a balancing force of “humility before God,” lest we forget our created pole. A much greater danger is present in taking the analogy in the opposite direction, that is, by placing the CAI in the role of God. This is of course idolatry, to be avoided at all costs. Herzfeld (2002, 83) points out the risk of this all-too-human trait of replacing “relationship with God with relationship with our own artefacts.” Any comparisons with this highest-level asymmetric encounter would not be encouraged by Buber (2013, 69), who notes that our I–Thou relationship with God is truly unique in that unlike all other I–Thou relationships, which decay to I–It, God “never ceases by [God’s] nature to be Thou for us.”

To enter Buber’s middle sphere, the threshold of mutuality must be crossed—but what are the essential factors that qualify interlocuters for this deeper form of I–Thou relationship? The first factor is clear, for in this sphere of “our life with men . . . the relation takes on the form of speech” (Buber 2013, 70)—but mere verbalization is not all that is required.

Buber’s (1957, 97) work highlights the need for both immediacy and distance between humans—for all the closeness of an I–Thou encounter, “one can enter into relation only with being which has been set at a distance, more precisely, has become an independent opposite.” Distance and opposition are required to distinguish the other as other, and yet as another with equal value to ourselves. Spoken language is emblematic both of that distance—only being necessary because there is no immediacy—and of how relationship can be built across it—being able to convey our intent, and even who we are, to the other. Speaking our name—seen as one of the primordial uses of language—declares who it is that approaches as the other. This approaching other must be accepted for who she is. “Genuine conversation . . . means acceptance of otherness. When two men inform one another of their basically different views about an object . . . everything depends, so far as human life is concerned, on whether each thinks of the other as the one he is, . . . [and] unreservedly accepts and confirms him in his being this man and in his being made in this particular way” (Buber 1957, 102).

Common understanding of words and concepts are presupposed. Today’s LLMs already display impressive functional performance in interpreting our questions and formulating (for the most part at least) syntactically correct and semantically meaningful responses, but without their own center of consciousness, it is only the human interlocuter who confers that semantic meaning to the symbols it produces (Landgrebe and Smith 2022, 241). For those symbols to have shared meaning, and thus to enable true conversation, there must be something in the CAI’s experience (or its configuration) to relate them to things in our shared world, whether physical or noetic. This can be a challenge even between two humans from different cultures or mother tongues, but for the CAI, the challenge would be significantly greater. Its experience of and presence in the world would be vastly different to any human’s—its I must be constituted in quite a different relation than the mother–child relation that first constitutes the human I, and the manner in which it experiences the world (particularly if it lacks a humanoid body) would further restrict the possibility for mutual understanding.

Thus, the power of speech alone is not sufficient for the CAI to cross the threshold of mutuality and so be encountered by us as we would another human. The ontological reality that it would not be human means it would be unable to provide the mutuality required in Buber’s middle sphere, which stems not only from shared language but from shared experience. Of course, this of itself does not deny the possibility of an I–Thou relationship between two CAIs. My position is not that human consciousness and intelligence are privileged as the only center for I–Thou relations—it is surely overly anthropocentric to insist that all conscious beings conform to the human model, which the Turing Test falsely assumes. Rather, we must recognize that whatever consciousness the CAI has will be alien to our own.3

Buber sees a second foundation for interhuman relationship, which the CAI would need to enter an I–Thou relationship—that of personhood, which derives from God. “God as a Person . . . enters into a direct relation with us men in creative, revealing and redeeming acts, and thus makes it possible for us to enter into a direct relation with him. This ground and meaning of our existence constitutes a mutuality” (Buber 2013, 95). For all the distance between God and humanity, there is “mutuality between God and man”—a mutuality arising from our personhood that is derivative of (but in no way the essence of) God’s personhood. Mutual personhood provides the foundation for I–Thou relations, with God as its source: “As a Person God gives personal life, he makes us as persons become capable of meeting with him and with one another” (Buber 2013, 95). Thus, some form of shared relationship with God is necessary for a true relationship between humans.

Buber’s conclusion is similar to Karl Barth’s (1960, 281), whose own I–Thou relations reach their climax when both interlocuters have a de facto as well as de juro experience of God’s covenant with humanity and so can enjoy a truly free “glad” encounter. However, a lesser formal type of gladness is available even to those without this de facto covenant relationship with God. “The crux of [Barth’s] logic is that even the sinful human has by nature a desire to love others . . . Love for God is not explicitly in focus here; indeed it is the absence of love for God which distinguishes formal from good gladness” (Proudfoot 2023, 691). Acknowledging that his own scheme is similar to Buber’s, Barth (1960, 278) notes that “Buber finally had in view this freedom of the heart, and only failed by accident to tread it to its ultimate consequences, and thus to come to this final conclusion.” Although Buber does not make clear the extent to which sinful humans can meet each other in I–Thou encounters, Barth maintains that Buber would follow the same logic and allow humans who do not acknowledge a relationship to God to enjoy an I–Thou relationship with each other, even if this falls short of what they might experience if both were in correct relation to God.

Could a CAI meet the criteria for personhood and thus for I–Thou encounter? With Barth’s concession, the CAI need not be aware of a relationship with God, but as with Buber, God’s relationship with us provides the model and foundation for all further relationships. Would this require a special act of God to endow the CAI with personhood? This is an important open question deserving further research, though it is at least a coherent notion that any entity that reaches a sufficient threshold of consciousness is ipso facto a person.4 This would allow a CAI to be truly admitted into the Thou world, even if its sheer alterity would mean it could not be a substitute for a human.

So, what level of consciousness is needed?

Nagel may be correct that a bat has conscious experience, but its consciousness intuitively does not reach the level required to support a mutual I–Thou encounter. Defining where that boundary lies without falling back on an anthropocentric definition of consciousness, which then denies it to any entity that is not human, is a challenging task. Here also Buber may provide some guidance. In his later work Distance and Relation, Buber (1957, 97) looks further at what is required to move from “the primal setting at a distance” (roughly equivalent to his earlier It-world) to “entering into relation” (i.e., an I–Thou encounter). He draws a significant distinction between the types of conscious experiences of humans and (as he imagines) animals: “An animal’s ‘image of the world,’ or rather, its image of a realm, is nothing more than the dynamic of the presences bound up with one another by bodily memory to the extent required by the functions of life which are to be carried out . . . It is only man who replaces this unsteady conglomeration . . . by a unity which can be imagined or thought by him as existing for itself” (Buber 1957, 98). For Buber, animal consciousness is limited to what he terms a realm, where objects serve only a functional purpose centered on the animal itself, whereas humans create a mental world, where objects have an independent existence and purpose and can thus be encountered as agents in their own right—as a Thou. More recent animal studies may weaken Buber’s categorization of animals, but the important point is not whether any animal meets this criterion but that we have here a heuristic for judging consciousness that is not inherently anthropocentric. The distinction of world-forming from mere realm-forming provides a guide to the level of consciousness a CAI must attain in order to enter a mutual I–Thou relation, without insisting it matches all aspects of human consciousness. Just how we would determine whether a CAI has constructed its own mental world is an interesting question, beyond the compass of this essay.

Mutuality is key to the deepest form of I–Thou relations Buber describes, but despite this, he allows for a number of asymmetrical encounters that shed light on how he might view a relationship between a human and a CAI, where mutuality cannot be full due to our very different experiences in the world. By analyzing these encounters, I argue that Buber would permit us to say “Thou” to a hypothetical conscious machine in the correct context. Relating to an artefact or a tree shows that we do not need to throw away our knowledge and perception that it is an artefact created by humans, nor do we need to insist that it meets us in full reciprocity. Our encounters with CAI would sit somewhere between Buber’s natural and human spheres, with its power of speech placing it above animals, which can only bear a fleeting nonverbal glance, but its sheer alterity making encounters with it different to those between humans. As with all I–Thou encounters, we would need to concentrate closely on the CAI, ascribing it value and worth—as it would need to do for us.

I have identified one key asymmetric scenario within the sphere of verbal relations: the educational encounter of helping a new CAI to find its place in the world of I–Thou encounters with us. Our responsibility here must not be overlooked and is foreshadowed by the need to train today’s AI systems with datasets that avoid bias and reflect our shared human values.

A key requirement for encounter is that we recognize the difference between human and CAI. To fail to do so would be to fail to allow all we know about the CAI to be included in the event of encounter and to fail to practice Umfassung (inclusion)—to fail, in other words, to enter the I–Thou relation at all. A CAI, with its different genesis and experience of the world, cannot be expected to meet us as another human would, but if its conscious experience reaches the threshold of creating what Buber terms a mental “world” in which we have independent existence and agency for it, it would be able to encounter us. What is more, with this level of appreciation of the world of which it is part, a CAI could have a truly mutual I–Thou encounter with another of its own kind, even absent a directly acknowledged relationship with the eternal Thou who is the source and foundation for all interpersonal relations.

Having established this principle, I offer in conclusion three avenues for further research to prepare for the advent of conscious machines, should they ever be created.

Whether or not a CAI will ever arise remains contentious, but the theological community would be wise to give more consideration to this possibility not only in case it does occur, but more immediately, to remain constructively engaged with AI development and deployment. This essay has shown that we should not see such an arrival as a rival for relationships with humans but rather that we can find legitimate ways to relate to these new beings. Ultimately, however, the only relationship that will truly satisfy the human is to say Thou not to another human, let alone to a CAI of our own creation, but to our creator God.