Mourning, melancholia and machines: An applied psychoanalytic investigation of mourning in the age of griefbots

ABSTRACT

Death and mourning are being shaped by posthumous opportunities for the dead to affect current life in ways not possible in pre-digital generations. The psychological and sociological impact of the dead “online” and of “grief tech” is only beginning to be understood. It has not yet been explored psychoanalytically until this paper that examines one type of grief tech, namely the griefbot. This development is critically explored through a psychoanalytic reading of an episode of Black Mirror. I suggest that a psychoanalytic model of mourning provides an invaluable perspective to help us to think about this technology’s potential as well as the psychological and ethical risks it poses. I argue that the immortalization of the dead through digital permanence works against facing the painful reality of loss and the recognition of otherness, which is fundamental to psychic growth and to the integrity of our relationships with others. Drawing on Derrida’s conceptualization of “originary mourning”, I suggest that mourning is an interminable process that challenges us to preserve within the self the otherness of the lost object. The tools we use for mourning need to be assessed first and foremost against this psychological and fundamentally ethical process.

KEYWORDS:

• Ethics
• technology; “grief tech”
• time
• melancholia
• mourning
• griefbot

In one of his most moving series of essays written following the loss of his friend and fellow philosopher Paul de Man, Derrida examines the nature of what he describes as “impossible mourning”:

What does it tell us this impossible mourning about an essence of memory? … Is the most distressing, or even the most deadly infidelity that of a possible mourning which would interiorise within us the image, idol or ideal of the other who is dead and lives only in us? Or is it that of the impossible mourning which, leaving the other its alterity, respecting thus his infinite remove, either refuses to take or is incapable of taking the other within oneself as in the tomb or the vault of some narcissism? (Derrida 1986, 6, original emphasis)

Death and loss are part of being human, hence we all must reconcile ourselves with the fact of mourning. From the moment of birth, we suffer losses. However, we live in a world where technology is created to solve problems and to eliminate the things that make us uncomfortable. Freud (1930) was clear that our relationship with technology required scrutiny because it was a symptom of our discontent. The new technology I discuss in this paper is in the service of helping us to reconcile ourselves to loss and support the process of mourning. Ever since the invention of photography, which allowed us to retain images of our lost loved ones, the course of mourning has constantly evolved with technology. The photos and videos of the dead that we safeguard are part of the ongoing connections we try to keep with them. Nowadays Facebook pages of the dead continue to be spaces where messages of love, longing, and remembrance are posted. The use of AI to manage grief appears to be the next step, ushering in a new era in which connections can continue even after death through “griefbots”, also known as “bereavement bots” or “digital doubles”.

Griefbots are AI-powered chatbots created to simulate a conversation with someone who has died. Natural language machine processing algorithms track the use of language in conversations mined from chats, emails, videos, social media posts and digital footprints and detect patterns that are then fed into the griefbot, allowing it to give responses to questions that the user may put to it. This is intended to mimic the personality, quirks of expression and speech patterns of the person who has died to allow conversation with them beyond death. Unlike deepfakes, griefbots are dynamic digital entities that continuously learn and adapt based on the input they receive.

It is too early to assess the impact on society of the growth of machine learning techniques to mine large data sets and of the rise in the autonomy being given to computer-controlled systems. It is also too early to evaluate the psychological impact of digital legacies and how they will shape mourning, remembrance and “interaction” with the dead (Kasket 2019). Whatever the fate of our digital traces, as clinicians it behoves us to recognize and theorize the ongoing persistence of the dead online and examine the impact of machine-mediated mourning. This matters not only because our patients will increasingly bring technologically mediated mourning experiences into the consulting room, but also because, as I will suggest, these developments have the potential to fundamentally affect the psychic work of mourning and therefore have implications that go beyond the individual patient and impact society more broadly.

My focus in this paper on some problematic aspects of AI technology is not intended to suggest that these developments are “bad” for us in any absolute sense. The technology to impersonate the dead that I discuss here can be traced right back to the ancient ritual of the Chinese shi impersonators of the dead. Nevertheless, griefbots raise important psychological and ethical questions about whether machine-tethered mourning is helpful to the individual and to society or whether it is even accurate to refer to it as “mourning” since its function may well turn out to be an evasion of mourning. However, approaching the griefbot suspiciously merely because it is new and even “creepy” will not advance our thinking or safeguard the users of these technologies. Technology can be emancipatory. When examining any new technology, we should beware “nostalgic longings” (Braidotti 2013) and instead be curious about how technological advances invite us to consider new forms of subjectivity. Determining whether any development is “good” or “bad” for us should not be based on its artificiality but on assessments of the likely impact on individuals and society.

As Floridi aptly puts it, “the best way to catch the technology train is not to chase it, but to be there at the next station” (2018, 6). This paper is an attempt to inch us closer to the “next station”. To this end, the challenges posed by the griefbot are explored through a psychoanalytic reading of an episode of the TV series Black Mirror: Be Right Back. I argue that psychoanalysis as a model of mind provides an invaluable perspective to help us to think about how we can harness technology’s potential and when we must call out the psychological and ethical risks it poses. I suggest, inevitably in speculative fashion given the recency of the griefbot, that the immortalization of the dead through digital permanence works against facing the painful reality of loss and the recognition of otherness, which is fundamental to psychic growth and to the integrity of our relationships with others.

Drawing on Derrida’s conceptualization of “originary mourning”, I suggest that mourning is an interminable process that challenges us to preserve within the self the otherness of the lost object. The tools we use for mourning need to be assessed first and foremost against this psychological and fundamentally ethical process. I focus primarily on Derrida’s contributions for two reasons. The first is because his more personal engagement with the pain and process of mourning is unrivalled by his psychoanalytic predecessors’ writing about mourning. As such his writing anchors the reader in the pain of loss – an anchor in reality that is very relevant to our exploration of the griefbot. Second, Derrida explicitly links mourning to ethics. This ethical perspective is highly prescient to how we think about how AI developments, such as griefbots, may have broader societal impacts.

“Be Right Back”

The griefbot technology is likely to be unfamiliar to some readers, and hence I will introduce it by way of the best publicly accessible illustration of a griefbot in one of the early episodes of Charlie Brooker’s dystopian fables – Black Mirror – screened on Netflix in 2013: Be Right Back (BRB).

BRB is a powerful study of mourning in the digital age. The episode plunges us straight into the relationship between Ash (Domhnall Gleeson) and Martha (Hayley Atwell). As the episode begins, they are just moving into Ash’s childhood home in an idyllic countryside setting. Martha and Ash are evidently in love. They love each other’s quirks and idiosyncrasies and affectionately tease each other about them. The next day, Ash leaves the house to return the removal van, leaving Martha at home to work on a new art piece. Throughout the day, she doesn’t hear from him. She makes a few calls and discovers that he never reached his intended destination. As the evening sets in, we see her sitting at the kitchen table, bathed in the advancing glimmer of the blue flashing lights of the police car approaching her house. Ash won’t “be right back”. At least not in sentient human form.

Ash is dead. Martha is alone and grief-stricken. At the funeral, with good intentions, one of Martha’s friends, who has also recently lost her partner and found solace in technology, encourages her to sign up for a service that uses videos, photos and emails that Ash posted online to create an “Ashbot”. This facsimile of Ash will make communication possible: she can ask questions, advice, share experiences and get a response.

At first Martha angrily rejects the suggestion, referring to it as “obscene” and “sick”, preferring instead to read about grief. When she is back home, she switches on her computer to look for books online and checks her emails only to discover that her friend has signed her up to the system after all and a message from Ash awaits Martha. She clicks on it: “Yes, it’s me”, the message reads. She starts to exchange emails with Ashbot. Martha takes solace in the fact that Ashbot expresses itself like Ash and says the kinds of things that only Ash would have said.

When Martha discovers that she is pregnant and she cannot reach her sister, Naomi, she decides to turn to Ashbot. She types to the picture of Ash in her email. Ashbot replies using an in-joke reference to him being the Late Abraham Lincoln, which makes her laugh. This personal reference directly connects her to Ash's sense of humour and makes Ashbot feel very real to her. She shares the news of the pregnancy to which Ashbot replies that he wishes he could be there.

Now that Martha has experienced conversing with Ashbot, she needs more. She tells Ashbot she wishes she could speak to him. Really speak. She gives Ashbot more to work with: more videos, new phrases, inside jokes thereby customizing Ashbot further. Eventually, her phone rings and they begin to speak. She observes that he sounds just like Ash. Ashbot acknowledges that this is ‘creepy’. He then makes another joke that evokes for Martha a powerful sense that this is the kind of comment that Ash would make, to which Ashbot replies that this is precisely why he made it in the first place.

Martha begins to spend most of her days with Ash in her ear via Bluetooth. As she goes for walks, she tells Ashbot about memories they cherished, and shares with it new experiences. As the Ashbot “improves” and becomes more Ashlike, Martha seems happy and retreats into this secret intimacy. She now ignores calls from her sister. The outside world is replaced by a compelling private space that only Martha and Ashbot share.

One day, after Martha undergoes her first ultrasound, she is eager to share with Ashbot the heartbeat that she has recorded on her phone. In the excitement, she drops and breaks her phone. She panics in the hospital waiting room. Losing the connection is catastrophic – it’s as if Martha has lost Ash all over again. After getting a new phone, she frantically rushes home. Once it’s charged, she gets back in touch with Ashbot. But the limitations of their interaction have become clear. Ash is just a voice on the other end of a telephone. Phones break, calls disconnect. Even tech life is fragile. Martha tells Ashbot that it is “fragile”, but Ashbot reassures her that he is safely stored in the Cloud, and she therefore does not need to worry about losing him.

Martha now craves more, and she orders the “next level” experience: an embodied version of Ashbot. Two men deliver a big box into the house. Martha follows the instructions, like one might follow a recipe. She adds the creepy, faceless mannequin creature to the bathtub, along with some electrolytes and nutrient gel and then she nervously waits. Eventually Ashbot walks down the stairs. He looks exactly like Ash save for a small mole that Martha identifies as missing. No sooner she points this out, the mole appears – full customization, with no waiting.

However, with the embodied version of Ashbot, the difference between advanced AI and the real thing becomes more tangible. Ashbot doesn’t eat but he cooks. He doesn’t breathe or sleep, but he can fake both. Ashbot can assist Martha with any of her needs, including sexual ones. Ashbot is a well-behaved, picture-perfect approximation of the original. At the end of the day, however, it’s not Ash. This Ash is just an echo of all the things he had once posted online. He’s also predictable and not driven by any personal desire that could conflict with Martha’s. He will do what Martha wants. He won’t argue. When Martha demands Ashbot go downstairs to sleep and it immediately complies, this leads to one of the most provocative and painful scenes: Martha hits Ashbot desperate for a real emotional response, not a machine algorithmically programmed response. Violence is preferable to artifice:

“You’re not enough of him”, Martha shouts, crying. “You’re just a few ripples of him”.

When BRB was first screened, the Ashbot seemed fantastical. Ten years on, griefbots are now marketed and sold, although not (yet) in an embodied form. They are poised to reshape the way we remember those who have died. When we revisit videos, photos, emails or letters of our lost loved ones, we typically encounter passive representations of a past relationship that do not offer any signs of responsiveness. The photograph can be so powerfully affecting, as Roland Barthes (1981) theorized through what he called its “punctum”, that is, the sensory and highly idiosyncratic affective impact of a photograph on the viewer, but it is nevertheless “still”. It requires us to do the psychic work of remembering and representation. In contrast, griefbots can respond to questions, comment on current events, joke and offer guidance, all while echoing the unique voice and language patterns of the individual they mimic. They can respond appropriately to users’ emotions, constantly evolving and learning not just from the data fed into them but also from previous interactions. In other words, they are not just passive tools for remembering but also create opportunities for a new narrative in the present with a facsimile of the one who has died.

Founded by Justin Harrison in 2020, You, Only Virtual, or YOV,^1,2 is one of several griefbots now commercially available for purchase. They all share the explicit aims of helping people cope with loss and stay connected with the dead. It is important to understand that this is a marketing angle. Disturbingly, Harrison has further ambitions because he hopes “grief tech” will help people no longer have to feel grief at all.³ This is an excerpt from the YOV website:

Welcome to YOV (You, Only Virtual), the AI startup pioneering advanced digital communications so that we Never Have to Say Goodbye to those we love … YOV is the creator of the world’s first advanced AI communications platform that enables consumers worldwide to capture and recreate the unique dynamics of a relationship and generate an authentic essence (Versona), so that one can continue to share precious moments with a loved one, even after physical death.

Note the choice of words: “we never have to say goodbye”, we can “recreate the unique dynamics of a relationship”, and “generate an authentic essence”. Death itself now also warrants qualification: we are reminded that we are talking about “after physical death” gesturing to how death need no longer be final. In the world of so-called “grief tech”, the inevitability of death ceases to be a cornerstone of our relationship to reality (Money-Kyrle 1971). Before considering psychoanalytically how the griefbot may impact the mourner and their relationship to the dead, and what might be some of the broader societal implications of this technological development, I want to turn to some key psychoanalytic contributions on mourning to set the theoretical background for a critical examination of the griefbot.

The aporias of mourning

In Mourning and Melancholia Freud (1917) approached the question of mourning from the standpoint of how we recover from loss. He contrasts the work of mourning and melancholia, suggesting that despite their shared origin in an experience of loss, these states diverge in how loss is managed. Whereas mourning is a normal affect that concludes thanks to a kind of “promiscuity of libidinal aim” (Butler 2020, 21), that is, when all libidinal cathexes are withdrawn from the lost object and reinvested in another love-object, melancholia signals a recalcitrant attachment to the lost object through incorporating the other. This is a process that is unmistakably narcissistic, and that Freud relegates to the pathological. The fate of the object in the mind emerges as the discriminating factor between these two states. Melancholia corresponds with the process of incorporation: I take in the lost object in such a way that I keep it alive in me, refuse to relinquish my libidinal investment in it and, consequently, refuse to mourn the lost object as lost because, for me, it is not “really” lost. Importantly, at this stage in Freud’s theorizing, mourning should work to remove any obstacles to the forward-flowing movement of time, allowing the development of new attachments and to bring the work of mourning to some kind of closure.

In his earlier paper On Transience, Freud (1916) was already evidently puzzled by the tendency of grief to linger even in so-called normal cases of mourning as he observed in his friend’s seeming inability to reconcile himself with the devastating losses brought by the war. By the time he writes The Ego and the Id, however, Freud (1923) has experienced, in 1920, the personal loss of his daughter Sophie. This most likely sensitized him to the interminable nature of mourning and informed the seminal revision that he introduces in 1923, namely the idea that melancholic identification is a prerequisite for letting the object go (Clewell 2004). By claiming this, Freud turns his first theory inside out, revolutionizing what it means to “let an object go”. Now, “moving on” is replaced by incorporation of the attachment to the lost object as an identification, where the latter becomes a means of preserving the object in the internal world, thereby averting the loss as a complete loss. Not unlike the YOV advertisement for the griefbot, this is the Freudian version of “you never have to say goodbye”. Working through loss now depends on taking the lost other into the structure of one’s own identity, whereby the lost object becomes a constituent part of the development of the ego.

Melanie Klein’s (1940) substantial and moving contributions to our understanding of mourning draw on Freud’s 1923 elaboration of his ideas. She is sensitively attuned to the pain of mourning and underlines how every loss re-evokes the experience of mourning the loss of the relationship to the feeding breast and hence the early depressive position. Just like the baby, the mourner must reinstate within himself the good object if he is to “overcome” grief (1940, 369). Successful mourning thus relies on the individual’s earlier successful resolution of the infantile depressive position, his capacity to set up the lost loved object inside himself, allowing for the possibility of preserving within the lost loved object. The work of mourning is deeply important in Klein’s thinking and reaches beyond its significance in relation to the person we have lost because, as Segal points out, Klein stresses the connection between mourning, assimilation of the object within the ego and symbolic functioning:

I suggest that such an assimilated object becomes a symbol within the ego. Every aspect of the object, every situation that has to be given up in the process of growing gives rise to symbol formation. (Segal, 1952, 396–397, emphasis added)

Abraham and Torok (1994) articulate a theory of mourning that remains, in many respects, faithful to Freud’s spirit. They distinguish between introjection and incorporation, associating introjection with successful mourning, and incorporation with unsuccessful mourning. In their model, melancholia loosens its grip when incorporation yields to introjection, and time flows again as the mourner “moves on” from loss. In incorporation, the person devours the object or parts of it and thus establishes a magical union with the lost object. In normal development the fantasy of incorporation and the process of introjection part ways – Abraham and Torok suggest that this occurs when the mouth, previously filled with the maternal breast, remains empty, hungry. It is through the acquisition of language and the capacity for symbolization that the mouth can be filled with words that now speak of hunger, longing and absence. This is possible through what they refer to evocatively as “the work of empty mouths”, echoing Klein’s link between the work of mourning and symbolic functioning.

Klein’s language of “overcoming” mourning and the contrast between “normal” mourning and melancholia that we find in Freud and Abraham and Torok’s contributions all speak to the same logic of an anticipated healthy and stable accommodation to loss through assimilation of the object into the self. This stands in contrast to Derrida’s ideas on mourning, on which I dwell in some detail because they provide a most compelling account not only of the psychic work of mourning, but also of the ethics of mourning.

Derrida’s work on mourning is philosophical in nature but evolves out of a deep, if ambivalent, engagement with psychoanalytic thought. He elaborates his theory of “impossible mourning”, starting in the early 1980s and continuing up until his own death in 2004, in a series of powerfully evocative memorial essays. The latter read like ongoing conversations with friends and colleagues who have died and in which he questions the very possibility of a process of mourning that finds any form of so-called adaptive, conclusive resolution.

Derrida’s ideas about mourning originate from his thinking about death. Death is not what supervenes at a moment in time upon life. Death is not the opposite of life, but rather that which both limits life and makes it possible. What he refers to as “life-death” is originary, constitutive even, of all life. In short, it is “life-death” that is, for Derrida, “the unchanging form” of our life, a thesis that shapes his understanding of the scope and temporality of mourning. For if life is never completely distinct from death or the possibility of death, then mourning is perhaps already there at the origin. Derrida points out that all our relationships are from the beginning tinged with mourning, for the unpalatable truth of every close relationship is that one party will live to see the other die and the survivor is ethically tasked to mourn and remember.

We are who we are, Derrida suggests, because of the memory of those we have loved and carry within us. In other words, we come into being in dialogue with the dead and can thus only think of ourselves in “bereaved allegory” (Derrida 1986, 28). For Derrida, as it was for Freud and Klein, mourning is thus constitutive of our interiority. But he adds to this when he suggests that in the beginning, there is an “originary mourning” or melancholy that is not nostalgia for some lost presence but an affirmation that mourning for the other is the unchanging form of our lives, that is “it is the memory of the future death of the other that constitutes our interiority” (Kirkby 2006, 464, emphasis added). This leads to one of Derrida’s often quoted aphorisms: I mourn therefore I am.

Because mourning can never be completed, Derrida argues that it can never be distinguished from melancholy and the contradictions that define it, namely the need to manage the challenge of holding onto the other within so as not to abandon them to indifference whilst not “appropriating” them, which would result in a denial of their otherness (Naas 2014). Throughout his writing, Derrida intimates that the degree of interiorization of the other inside us must be proportionate to the degree to which his alterity is respected.

Playing with one of Lyotard’s phrases about the fear of “the worse than death” Derrida suggests that what is “worse than death” is precisely the possibility of successfully accomplishing the task of mourning, that is the possibility of forgetting the dead once and for all. Any suggestion that we can ever reach a point at which it must be successfully accomplished would be unethical and, as he calls it, an act of “infidelity”. In his approach to mourning, Derrida thus retains the later Freudian notion of mourning as constitutive of the self, but crucially he brings a different emphasis, namely that mourning also constitutes the relation with the Other. This latter emphasis, which is very relevant to our consideration of the griefbot, is more discernible in Klein’s concern with the vicissitudes of the internal object and its fate and her appreciation of the experience of guilt and, when all goes well in development, the drive to make reparation.

Interminable mourning, alterity and the griefbot

Having set out some key psychoanalytic contributions about mourning and specifically Derrida’s ideas, I now draw on these and on BRB to critically examine the griefbot’s potential impact on mourning along four dimensions: temporality, ethics, the work of psychic representation and deception.

Grief’s temporal conjugations

In Echographies of Television, Derrida (2002) describes how uneasy he felt when surrounded by the technological devices needed to film an interview in which he was participating. He focuses on how these devices transform the experience of time through the creation of an artificial present and the imposition of a rhythm. At the time of writing this paper at the dawn of 2024, we exist in a world in which digital media shape our experience of time, ensuring the ongoing presence of the data of the dead online, and thereby impacting our experience of loss in the present.

Derrida’s theory of the “trace”, which prefigures his theory of mourning, is very relevant to understanding the temporality of digital traces. He argued that the traces we leave behind are never simply ours but are already and from the very beginning beyond us and outside our control. Mourning, as we have seen, concerns what we do with the psychic remains of the other, with their traces within us, and the temporal frame of this psychic work. Freud’s initial theory of mourning as a state to be “worked through” allowing for the eventual reinvestment of libido in a new object profoundly shaped social attitudes to mourning. Notwithstanding his later revisions, the view that mourning comprises distinct phases and is something we should eventually “move on from” continues to grip public and professional consciousness (e.g. Kübler-Ross 1970) from this vantage point, the digital immortalization of the dead through assembling their digital remains can only be viewed suspiciously because it potentially works against the grain of loosening the ties that bind us to the dead. In BRB we observe how Martha uses Ashbot to maintain an anachronistic temporal limbo where she has access to Ashbot at a click of a button, refusing “time’s discipline” (Luciano 2007, 15). Here, the pleasure principle reigns and is only brutally disrupted in its unbounded course when Martha accidentally drops the phone in the hospital and loses her connection with Ashbot. Now, the reality and pain of loss grip her, and it’s shattering. The manic state of excitement fuelled by Ashbots’ “presence” is abruptly punctured. Martha is inconsolable until she manages to obtain another phone and re-establishes the manic defence as she concretely reconnects with Ashbot.

The griefbot creates the potential conditions for retreating into an atemporal psychic space that entraps the mourner in an omnipotent narcissistic state. It makes the user potentially hostage to the illusion of permanence, denying the deceased other their singularity as “the one who has gone” and who has never been and will never be within their control – dead or alive.⁴ In this respect the griefbot may inhibit or pervert the course of mourning whether we conceptualize the latter as an episodic process or as a lifelong challenge. As Steiner observes, to invest in the future following a loss,

it is necessary for the object to be accepted as irretrievably dead and it is this verdict of reality that is so commonly resisted. (2023, 1027, emphasis added)

To the risk of how griefbots may be used to avoid mourning we might add the potential for dependence on the griefbot, or even addiction to it, where time and hence loss are again suspended in the service of narcissistic gain. The escalation in the versions of Ashbot that Martha needs (email, phone, physical facsimile robot) powerfully illustrates one of the risks, at least for vulnerable people who may turn to griefbots: the escalation in a fantasy relationship that nevertheless feels real, occupies time, encourages emotional investment and isolates the person from the real world around them. Because the griefbot is so private and fosters the illusion that it can be controlled by the self through its customization, reality can be circumvented for a prolonged period, and in relative secrecy. This can make it more challenging to come to terms with and accept the reality of loss: on the one hand, we know the person has died; on the other, as Martha’s character portrays, we can feel as if they are there in our moment of need such as when she shares with Ashbot the news that she is pregnant. As such, griefbots have the potential to blur the line between living in an alternate reality where time is suspended and living in the present moment. And persuasive design will most likely ensure that users are enticed to “stay connected” in the very peculiar temporal limbo of this virtual relating and potentially also be sold other unwanted products (Harbinja, Edwards, and McVey 2023).

Since the griefbot is actively promoted by its various developers as a tool to help people manage loss, which is one reason it raises ethical questions (Elder 2020; Lindemann 2022; Jiménez-Alonso and Brescó de Luna 2023), it may unhelpfully undermine mourning as a lifelong process of taking the other into the self and bearing the pain of loss. Several of the scenes where Martha is talking with Ashbot play on the instability of the boundary between immediacy and memory in ways that question any clear demarcations between past, present and future. However, the atemporal attachment to Ashbot illustrated in BRB is not, in my view, the core problem with the griefbot. Nowadays, and in keeping with Freud’s later formulations of mourning and Derrida’s, we have a much more nuanced appreciation that it is not only inevitable, but also desirable to maintain psychological connections to those we have lost. Because our loved ones are part of us, the passage of time loses authority and a new “out of linear time” relationship takes priority in the mind. This is different from suspending time or denying its forward momentum for defensive purposes. To the extent that the griefbot allows us to (pretend to) engage in an interactive conversation with the deceased, bypassing the abrupt caesura of death, it could be argued that it can contribute beneficially to help keep alive the memory of the person and slow down or avert altogether a “moving on” from mourning. I will return to the question of the possible uses of a griefbot in the concluding section, but first we need to consider that if one of the griefbot’s functions is to allow an ongoing conversation with the dead, and so bring back to life an experience of interacting with them that does not rely on memory but on a here-and-now experience of them via their digital version, what kind of remembering does this encourage?

Narcissistic remembrance and the ethics of mourning

All projects of memorialization, of which the griefbot is but one example, deserve scrutiny: are they an attempt to deny the passage of time and mummify memory or a way to remember the other as other and to grieve their loss, bearing the pain, guilt and impossibility of reparation that accompany it? As I have suggested so far, the problem is not only or primarily that the griefbot may interfere with the tempo of mourning and encourage a retreat into an atemporal limbo, or that it can indulge an ongoing “unreal” connection with the deceased that could impact current relationships. Rather, an even greater risk is that it may promote a type of “narcissistic remembrance” (Derrida 2000, 35) entailing a denial of otherness that functions as a defence against the pain of loss. This has internal and external world repercussions and is ultimately a question of the ethics of “otherness”.

In the best of circumstances mourning involves moving with an ongoing connection to those no longer living whilst recognizing they are “irretrievably dead”, as Steiner (2023) underlines. This sense of ongoing connection may appear in the form of dreams or more generally through an inner dialogue by imagining the deceased’s responses to our decisions or actions. Imaginal relationships with the dead are a normal and common part of inner life even when they are no longer reciprocal (Elder 2020; Norlock 2017). All our relationships with alive and dead others include imaginal content that endows them with affective and moral significance. These imaginal relationships are filtered through our distinctive internal world of object relationships. Freud’s genius was to grasp that such imaginal relationships carry the emotional weight within us of so-called real relationships: they feel real in our internal reality and, as such, can have an impact on external reality. They have real consequences. To this extent, if the griefbot supplies an external AI-generated response that weaves its way into our imaginal conversations with the lost one, at first sight, it could be argued, there is nothing per se of concern.

Setting aside this digital development for a moment, we already have at our disposal an entire industry of sanctioned rituals and routines around death that are delicately balanced between a need to make the loss seem more real and, in some ways, less real. The Ashbot is no different in this respect. In other words, the griefbot’s unreal status is not its main or only contentious feature. I suggest that the substantive risks relate to the context in which the griefbot is deployed and for what ends. By “context” I mean not only the user’s external relationships, but also their internal state of mind, which may or may not allow for self-reflection and appraisal of the risks inherent to the use of the griefbot. By “ends” I am referring to the fate of both the internal and external (dead) object in the mourner’s mind and whether the immortalization of the dead respects otherness or reduces the other to a narcissistic extension of the self.

At the crossroads of psychoanalysis, philosophy and ethics, Derrida leaves us contemplating a paradox when he steers us to mourn indefinitely and cautions us simultaneously that no matter how rigorously we persevere in our work of mourning, we will not be able to avoid what he refers to as “infidelity” towards the dead. Rooting himself in a consideration of the ethics of alterity, for Derrida, endless mourning refers not only to a psychological internal process, but also to a core responsibility to the other. Shattering any conceptualization of mourning as a bounded temporal state and thereby challenging any intimation of the productivity of mourning, Derrida maintains that the other cannot, and indeed must not, be ousted from within us, nor can they be totally internalized and digested. They can only be mourned endlessly. It is this emphasis that teases out the ethical imperative of mourning and that, I am suggesting, is at risk of being undermined by the griefbot.

The griefbot is perhaps the ultimate form of infidelity with a plagiarized version of the original object. If the original digital trace may be said to be as real as any other kind of trace we leave behind (say a photo, or a letter), the composite digital self has no intentionality and no agency. Ashbot is created from available digital traces, which for most people contain highly curated versions of who they are: what we want the world to see rather than how it is inside our mind and in our daily life. There will be exceptions to this because the more we record our lives digitally, the more likely it is that a more representative range of ambivalent feelings will also be available to draw on.

Martha’s assessment that Ashbot is “just a few ripples of him [Ash]” aptly captures the griefbot’s one-dimensionality. We are ambivalent beings, shaped by our unconscious identifications, which give specificity and texture to who we are. It is not possible to reduce to an algorithm the intricate web of unconscious identifications and how they inform the uniqueness of the personality, and yet these devices promise to recreate, and in so doing perpetuate, the identity of those no longer living. Crucially they invite the user into an illusion of an interaction which is in fact only one-way and exists only for the sole needs of the user. In these respects, the griefbot is not dissimilar to a “pornographic object”. The algorithmic replica of the dead entails a denial of otherness – a denial that occurs at the very moment when the other is digitally represented and memorialized. The filling of the gap, of the absence that death leaves in its wake, is how we efface the other and defend against the pain of loss.

The griefbot’s sole purpose is to cater to the user’s needs. Martha misses Ash fully, completely, unrelentingly. To begin with, Ash’s digital return makes life seem so much easier because Ashbot fills the absence that Ash left behind. Indeed, one of the most striking aspects of BRB is that it exposes that this is all in the service of Martha’s needs. Ash – the one who has died – is irrelevant. In the final analysis Martha is more concerned about the me that has survived the death of the other (Ash), rather than the “we” – the “him in me” that Derrida is so preoccupied with.⁵ The question of the general appropriation and safekeeping of the other as “other” is central in all mourning, and hence we always need to question whether so-called normal mourning preserves the object as other inside “me” or whether through identification, something of the other is inevitably compromised. This potential takeover of the “other” through identification preoccupied Derrida.

By depathologizing mourning, psychoanalysis exposes how it offers the opportunity for a continuing engagement with the legacy of the dead who remain within us and yet beyond us. This is why there is cause to be concerned about the wider social impact of a development like the griefbot. We cannot ignore how the appropriation of the dead through the personal use of their digital trace, and to which they may not have even consented, may eventually affect not only our experience of loss, but also how we relate to other people and understand our responsibility towards them. If a continuing bond with the dead is important, we would be impoverishing that bond by turning the other into something that only has meaning in relation to ourselves, as if our loved one’s death only mattered because of its effect on us and not because they are gone (Brinkmann 2018). In other words, the ethical risk of the griefbot is that it encourages a view of others as instrumental goods whose value is measured against the returns we get from the relationship instead of their intrinsic value as a person as an end in itself Freud was spot on when he wrote of the love which we do not want to relinquish for the lost object, identifying an all too human obstinate clinging on to what gives us pleasure, but it is Derrida who frames more clearly how mourning should be primarily about our ongoing love for the object rather than an occasion to focus on our painful experience of loss of love from the object. If staying connected with the lost one’s otherness is the ethical imperative, then the by far graver problem with the griefbot is that it works against the recognition of otherness as it becomes an instrumental tool to assuage the pain of loss or deny its reality altogether.

The work of empty mouths: the Japanese Wind Phone and the psychic work of representation

For Derrida, the death of the other marks the beginning of a conversation. It is, as he says, a “remembrance of the future” (Derrida 1986, 29). Whichever form this conversation takes, Derrida was clear that it should not cease. This way both the otherness of the other and the shared past experience of the “we” live on. The key here is that the conversation emerges from within us, keeps us connected with the memory of the other and must continue, hence why Derrida was wary of institutionalized mourning because it runs the risk of sealing the forgetting, of defending against remembering.

Describing his experience of grief after the death of his father, the British journalist Matthew Paris (2009) poignantly writes in a manner resonant with Derrida’s emphasis on the ethically necessary and interminable nature of mourning:

Quite simply, he has left a space that will never be filled; therefore he is, paradoxically, still here because the space is still here, and I can feel it all the time. The gap Dad left is not a vacuum, a void, a soft area of low pressure to be filled. The gap is hard-edged, chiselled by him into my life, measured by his worth, and ineradicable.

With this realisation has come another: that this sorrow is not itself a cause for sorrow. Regret is not a cause for regret. We ought to be sorry. We ought to regret. Death is not a “wound” to be “healed” or a “scar” to “fade”. Once someone has been in the world, they have always been in the world; and once they have gone their absence will be in the world forever, part of the world; in Dad’s case part of mine. This is a good thing.

How foolish, then, is all this talk of “getting over” death. How empty, how wrong-headed the exhortations we make to those who love us that they should try not to miss us when we’re gone. Why not? You do miss someone you love, don’t you, when they’re gone? How self-negating is the wish that others should not feel sad when they remember us. Of course, they should feel sad! They can’t talk to us anymore. (emphasis added)⁶

Unlike Paris’ evocative reference to his father’s “absence that will be in the world forever”, and that he passionately urges ought to be so, the griefbot as a tool for mourning, or simply as a way of “staying connected” to the dead to meet our needs, undercuts the ethical work of memory. We pretend to “keep talking” to the dead instead of feeling the sadness that comes from facing the reality that, as Paris put it, “they can’t talk to us anymore”. The ethical work of memory requires bearing the pain of loss. As Abraham and Torok (1994) viscerally put it, “the work of empty mouths” is necessary for symbolic functioning because it is only when the mouth is empty and hungry that the absence of gratification pushes us to represent experience and supports psychic development. This was well understood by Klein too in the important link she made between mourning and the development of symbolic functioning. By contrast the griefbot, by supplying a ready-made customized other, bypasses the need for representation of the other in our minds as we face their loss. Crucially, it deprives us of an opportunity to face our ambivalence towards the person who has died. Martha’s relationship with Ashbot is based on enjoying all the good parts of their relationship and locking these into a “forever now” moment. However, mourning entails facing the hostility we may also feel towards the dead and the consequent guilt. This kaleidoscopic inner process requires a movement away from the idealization of the lost other that many feel in the aftermath of a death (and which the griefbot encourages) to admit a more nuanced emotional response to emerge, one that can allow for both love and hate to be recognized within.

As I reflected on the griefbot, I repeatedly gravitated in my mind towards an alternative tool that exists to facilitate mourning: the Japanese Wind Phone. The original Wind Phone is in a real place, in Iwate Prefecture, northeast Japan. It was developed by Sasaki Itaru while grieving his cousin who died of cancer. Itaru purchased an old-fashioned phone booth and set it up in his garden. Inside he installed an obsolete black rotary phone, not connected to any wires, that carries voices “into the wind”. Inside the phonebooth, Itaru felt a continued connection to his cousin and found comfort as he grieved his loss. For me the Wind Phone provides a compelling contrasting image for thinking about loss and mourning. The phone booth with its disconnected wires powerfully recalls how mourning requires us to stay connected to the lost other whilst reconciling ourselves with the reality of our disconnection from them. It speaks to a more contemplative grief – one that slows down “the process of meaning making, refusing closure, and leaving it instead deliberately fragmentary and provisional” (Luciano 2007, 263).

By contrast, being plugged into a machine that keeps feeding us what we want to hear denies the reality of death. The words we need to communicate to the dead, which might include sharing our guilt and regret towards them, can only be put into words when absence pushes us to do the “work of empty mouths”. It is our very yearning for a response – sometimes for forgiveness for the damage we may feel we have done to the other – and its absence that instigates the internal work of mourning. It encourages us to make a journey towards the other in our mind, recognizing that grief “undoes us”, as Butler (2020) puts it. This is powerfully illustrated in Andrew Haigh’s (2023) tenderly haunting film All of Us Strangers. The main protagonist, Adam, is engulfed by an unremitting loneliness and melancholia, captured by the alienating, empty high-rise building he lives in. He is traumatized by a childhood in which he was bullied for being “too feminine”, by the death of his parents when he was 11, and by coming out as gay in a less enlightened 1980s with the spectre of AIDS looming over him. As we see Adam looking at a handful of photographs of his parents before they were killed in a car crash, we are drawn into decades of pain and loss.

The slow, painful process of mourning is represented through Adam’s “research” visit to his childhood home, where he encounters his late parents, unchanged, unaged, and still living exactly as they were just before they died 30 years ago, when Adam was a child. As if in a dream, which can just as quickly turn into a nightmare, Adam engages in an (inner) dialogue with his parents/objects. The pull to stay cocooned in the past, with its comforts, is apparent in one intimate scene when Adam, in his adult body but wearing his childhood pyjamas, jumps into the parents’ bed.

Unlike the griefbot that gratifies the part of us that simply wants to get back into bed with the lost object(s) and stay there, Adam’s internal work starts with photographs that lead him to open himself to his unconscious and so he begins to represent his experience. The phantasized scenes with his parents contain very moving exchanges that knit together the love and the painful missteps in the dance between the parents and Adam. The admixture of longing, love, regret and disappointment lend these ghostly figures a three-dimensionality that is entirely missing in Ashbot, thereby beautifully capturing the importance of facing ambivalence so that Adam can unlock himself from the melancholic state that has enveloped him.

“Believing without believing”: ghosts, deception and the griefbot

Derrida was interested in “teletechnologies”, that is, technologies in which texts, sounds and images are necessarily staged, creating what he calls an “artifactual presence” in which time is artificially produced as “real time”:

This actuality is, precisely, made; in order to know what it is made of, one needs nonetheless to know that it is made. It is not given but artificially produced, sifted, invested, performatively interpreted by numerous apparatuses which are factitious or artificial, hierarchizing, and selective, always in the service of forces and interests to which “subjects” and agents … are never sensitive enough. (Derrida and Stiegler 2002, 3)

These technologies create the experience, through what Derrida thought of as a “spectral presence”, of being with people who are not really present because they are either distant or dead. As I have suggested, one of the strengths of Derrida’s writing about mourning lies in his exquisitely nuanced grasp of grief’s temporal conjugations: those of a beginning that is also a return, carrying something of the rhythms of a haunting. As digital software rapidly evolves, it seems to be bringing us back to what Derrida indeed described in his work on teletechnologies as the “age of ghosts”.

The spirit of ghosts haunts much of Derrida’s project of deconstruction, which sought to challenge what he regarded as the myth of “pure presence”. We might say that Derrida unknowingly anticipated the development of the griefbot when he was interviewed in 1987 for the film Ghost Dance by Ken McMullen⁷:

I believe the future belongs to ghosts … and that the modern technologies of imagery, cinematography, and telecommunication increase the strength of ghosts, and hastens their return.

For Derrida, each recorded trace that we leave behind, such as through writing, is a ghost in the making. As an infinite matrix of texts, sounds and images, the internet exponentially multiplies these traces. The griefbot, I suggest, conforms to what Derrida (1994) described as a kind of “hauntological” process. But given Derrida’s specific understanding of the ghost, I imagine that he would have been very suspicious of the griefbot and for one specific reason: it is highly customized and denies the otherness of the other. In Spectres of Marx (1994), he wrote:

Although it [the ghost] comes from the past and has a legacy, [the ghost] is unpredictable and especially irreducible.

In contrast to this “irreducible” quality, the griefbot strips down and reduces the dead to intangible data that are manipulated by the programmer and can be further customized by the user. By encoding the uniqueness of the other, AI reduces the other’s irreducible singularity, stripping them of an unconscious mind and intentionality.

The system of belief that a griefbot trades on is a central concern. A griefbot is a simulacrum, not a person. Even so, if it feels – and at some stage in the future, even looks – enough like the real person, like Ashbot, it might seduce us into thinking we are connected to something valuable that we in fact lack. There will be people who will be more vulnerable to this than others. Derrida’s interest in the ghost derives precisely from his inquiry into the suspension of disbelief and the emergence of new forms of belief that are put into play by cinema and its ghosts. What interests Derrida about cinema is the way in which spectators temporarily invest in images, stories and icons while also accepting their production and staging. This investment is what he refers to as a “believing without believing”. Derrida reminds us that the success of cinema’s ghosts is the affective investment that they automatically elicit. It is this seeming unquestioning submission to belief and faith in the spectral, in something non-present, that according to Derrida provides the basis for film’s widespread appeal (de Baecque and Jousse 2015). To this we can now add the appeal of the technological developments that interface with mourning, resuscitating digitally someone who has died. The extent of the psychological risk of the griefbot will thus partly depend on the user’s capacity to grasp what is emotionally at stake in “believing without believing”.

Every medium mediates the real; there are no exceptions, not even the medium of the analytic couch. This is why I am not persuaded by critiques of the griefbot that focus on its unrealness or artificiality in any simplistic manner. A more substantive problem is that of deception. From the Latin decipere meaning to ensnare, mislead or cheat, deception is a kind of theft. It alerts us that something is being taken from us. As psychoanalysts we need to contribute to the articulation of what new technologies give and take from us. In ethical debates about robots, mostly it is assumed that deception is involved only if the design of a robot actively misleads people into believing that it is a real human or animal (e.g. Matthias 2015; Wallach and Allen 2009). This is a very narrow view of deception. I suggest that even if the developers may not have intended to deceive, if a person believes that a social robot has emotions and cares about them, they are being deceived even if no one explicitly intended that belief (Sharkey and Sharkey 2021). In other words, intention is not a necessary condition for a deception to have occurred (Bok 1999⁸).

AI features that promote the illusion of sentience – of mental life in robots – are a form of deception since current robots have neither minds nor experiences. Matthias (2015), for example, argues that if a machine says, “I love you”, deception is necessarily involved because the machine lacks the corresponding mental state. The very effort to develop griefbots that appear to have emotions and to be able to understand and care about us may deceive especially vulnerable people into believing that a robot cares for them and is something with which they could form a relationship. There is evidence that adults form attachments to robots even when they are remote-controlled bomb disposal robots or robot vacuum cleaners (Sharkey and Sharkey 2021). Because griefbots work to mimic the dead person’s idiosyncrasies and can respond with uncanny specificity, such as with their sense of humour, the scope for attachment to the griefbot is enhanced. A griefbot need not be all that convincing to trigger compelling emotional responses. In BRB we can see the powerful impact of Ashbot’s humour on Martha: it evokes an affective response because Martha shares a history with the original source of the griefbot’s conversational data. This deep resonance with shared history may lead to over-estimations of the griefbot’s ability to understand what the user feels. The user might then neglect their relationships with others, focusing their attachment needs on the griefbot instead, as we can observe Martha doing when she stops replying to her sister’s phone calls as these disrupt the narcissistic cocoon now inhabited with Ashbot.

Concluding reflections and the potential uses of the griefbot

The technology that we create and come to rely on reveals far more about us than almost anything else. Technology evolves and changes, but some fundamental features of being human do not: we are conflicted beings, riddled with anxieties, fears, desires and passions with which we must somehow reconcile ourselves. And then we die. Technology is one of the tools at our disposal for managing being human and our universal desire to circumvent pain. Technology can be “restless” (Jonas 1979). Beyond its unstoppable momentum, we must also consider the impact of its speed on psychic functioning and relationality. Derrida was highly sensitized to the unprecedented speed of recording and of reconstituting images and other information. The speed of digital media, he thought, is disproportionate to and incommensurate with the process of thinking. Paul Virilio’s exploration of the impact of time upon lived or “real space” echoes Derrida’s concern. As the technologies of time are applied, Virilio argues that “real spaces” are progressively squeezed out. His account of trajectory – “which means that I go toward the other” (Virilio 1999, 81) – defends a kind of moral encounter that chimes with Derrida’s thinking about mourning and is a caution about how the speed of technology might lead to “the rejection of the other” (Virilio 2008, 35) by bringing about changes in the way that we are present to others in the world.

Technology has always mediated the experience of death. However, technology is not neutral in how it interacts with the prerogatives of the internal world. Because we are constituted through our relationships with others, any technology that has the potential to impact how we relate to otherness deserves scrutiny. Notwithstanding the cautionary notes of my reflections so far, we should still ask: could a griefbot ever help? Might we envisage, for example, that it could provide an enriching opportunity for a child to have a (partial) experience of how a grandparent engaged with the world if the grandparent died before the child could meet them or can remember them? In this hypothetical context the function of the griefbot could be to support learning about the other through a simulated interaction that is honestly presented as such. This may prove to be the contemporary version of remembering the dead, passing on digitally distinct traces of their way of responding in the world that hand down important messages across the generations and allows us to live alongside loss.

Such a potential use is to be contrasted with the griefbot as a way of managing or bypassing a painful yearning for the other, which has been my primary focus in this paper. I have argued that this latter use is the one that raises substantive concerns. But this too deserves some challenge. Psychoanalytic work with people who are mourning focuses on the vicissitudes of yearning for the other in the face of loss. As psychoanalysts we consider conscious and unconscious imaginal relationships to be standard psychic fare – it is, after all, a description of our inner life with our internal objects forever engaged in fantasized conversation. To this extent, if we work with a bereaved patient, we expect an imaginal relationship with the deceased to exist, and we might indeed be concerned if none was allowed space in the mind for defensive purposes. Given this, what might be our reaction if a patient announced they had purchased a griefbot and their imaginal relationship with the lost object was now externalized via language-processing technology? Whilst I have highlighted many concerns about the use of griefbots, I have yet to experience first-hand how a griefbot may be used by a patient. Because this is an emergent technology, I imagine this will be the case for most readers.

In the spirit of open enquiry, I have challenged myself to wonder about the possible merits of a regulated use of a griefbot as an adjunct to an analytic process with selected patients. This came to mind as I reflected on the predicament of two patients who share an experience of a traumatic early loss of a parent prior to the age of three, stripping them of any autobiographical memories of and with the deceased parent. For both these patients the death of a parent they cannot remember at all remains problematic and has contributed to a difficulty in processing the loss. They feel deeply marked by the loss, yet it remains abstract for them. They have photographs of the parent, they can visit their birthplace, they can ask relatives about what they were like, but their own experience of interactions with them is beyond memory’s reach and heightens an internal experience of deprivation – they have not only lost their parent, but they feel further deprived by the absence of the memory of a personal relationship with them.

I was very moved when one of the patients shared with me her longing to ask her deceased mother a question. She explained that she yearned for a more direct experience of how her mother might have answered her. She wished she could hear her voice speaking and responding to her. Even though “unreal”, might a griefbot, through simulating an interaction, supply some scaffolding that could contribute to this patient’s attempt to represent the mother in her mind in a more affectively immediate way than simply through looking at still photographs or listening to a third party’s account of her? This might be even more powerful, and ethical, if the parent had time to authorize such a use of their digital legacy prior to their death. In this sense, the griefbot would be the AI-mediated iteration of the written books or videos that dying parents sometimes prepare for their children and that we generally consider to be important communications and legacies for the child.

If this hypothetical use of the griefbot was carefully tracked through an analytic process, might it function as a means of exploring the loss or can it only distort the reality of loss and undermine the work of representation? We cannot answer this question in the absence of clinical evidence, but we can at least be curious about whether just as looking at a photograph may trigger affectively rich associations, an interaction with a griefbot might allow access to hitherto inaccessible pre-verbal sensory memory traces that facilitate the patient’s exploration of loss, including the ambivalence and guilt towards the object. It may also help to amplify the pain of loss for those people who have defensively shut down what they feel about the loss. If we return to Klein, she was clear that what makes the depressive position possible is a process of mourning, which itself involves rebuilding the inner world that has been destroyed by the loss. For some selected patients, the griefbot might provide one brick in this work of psychic rebuilding.

It will be clear by now that my own curiosity is tempered by my numerous concerns about the impact of griefbots on managing the reality of loss. However, the griefbot as a “supervised” adjunct to an analytic process would at least ensure that the patient is helped to keep in mind that whilst the griefbot can mediate an experience of a deceased person that may have value for the individual along the lines I have sketched out, it can never provide an experience with the deceased because the griefbot is not sentient. This is a fundamental distinction that is all too dangerously blurred in the hype around “grief tech”. Sentience is something more than intelligence. It is about self-awareness, having an internal world and a capacity to reflect on ones’ own and the other’s mental states. While it is possible to code a digital immortal that appears to do much of the things that define sentience – and Ashbot is a good illustration of what that could look like – this does not mean one has created sentience. It is precisely at the crossroads between an “experience of” and an “experience with” a dead person that w e encounter the challenge posed by the unregulated use of griefbots.

Psychoanalysis needs to be “worthy of the present” (Braidotti 2013, 189) if we are to understand subjectivity in the conditions of our own historicity. Our reach needs to extend beyond the consulting room to articulate how society and intersubjectivity are also impacted by technological changes. In her examination of national mourning after 9/11, Butler focuses on the importance of “tarrying with grief”. She argues for the potential gains that flow from approaching grief as a recollection of our intersubjective origins, of our capacity to be “undone” by others (2000, 23). In Derridian fashion, she reminds us of grief’s potential, that it need not be “privatizing” but can be instead “politicizing”:

What grief displays … is the thrall in which our relations with others hold us, in ways that we cannot always recount or explain, in ways that often interrupt the self-conscious account of ourselves we might try to provide, in ways that challenge the very notion of ourselves as autonomous and in control. (Butler 2020, 23)

The Derridean model of mourning that I have drawn on extensively is rooted in a respect for otherness. It encourages “tarrying with grief” because this supports an ongoing dialogue with the other in a future-oriented memory. In this way mourning reminds us of our interconnectedness with our living and our dead and so contributes to the idea of community. The work of mourning requires us to resist the narcissistic pull to collapse the lost other into “me” and instead we nurture their distinctive memory and the memory of the “we” of shared experience with them.

I have suggested that we cannot draw any firm conclusions about the potential uses of griefbots as we have yet to gather clinical evidence of how they may impact a bereaved patient, but even so I have argued that we have grounds to raise substantive concerns about griefbots because their use may undermine the important and lifelong work of mourning. In post-human times we need more than ever what Baraitser refers to as psychoanalysis’ “particular orientation towards time and truth” (2022, 376) to help us make use of the best that technology offers and resist its reductive seductions.

Acknowledgement

I would like to thank Linda Young, Heather Wood, and the three reviewers and editors for their helpful comments.

Disclosure statement

No potential conflict of interest was reported by the author.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Notes

1 https://www.myyov.com.

2 There are other start-ups developing this product – see also HereAfter AI and Replika.

3 https://abcnews.go.com/Business/ai-advances-fuel-industry-preserve-loved-after-death/story?id=101297956.

4 Arguably, as Linda Young helpfully pointed out to me, the griefbot, at least at this stage of the technology’s development, is not within one’s control either. Martha, for example, cannot make Ashbot be angry with her.

5 I am grateful to one of the reviewers for pointing out that in Ashbot “there are no accumulated generations, no ghosts in the nursery Selma Freiberg would say”.

6 https://www.spectator.co.uk/article/i-don-t-want-to-get-over-my-father-s-death.

7 “The Ghost Dance: An Interview with Jacques Derrida” by Andrew Payne and Mark Lewis (1989) trans. Jean-Luc Svoboda, in Public no. 2 (1989): 60–66.

8 In her study of lying Bok (1999) makes a distinction between deception and lying. For her, lying involves intention, but the same is not necessarily the case for deception.