Location
Wolff Auditorium, Jepson Center
Start Date
4-4-2025 9:15 AM
End Date
4-4-2025 10:00 AM
Description
Chaired by Tim Clancy, Ph.D. (Gonzaga University)
The advent of Large Language Models has made it possible for people to produce generative AI chatbots that imitate the conversational style of the dead, fine-tuning the system’s outputs to mimic speech patterns by drawing on their text messages, social media posts, emails, and other digital remains. What should we think about this practice? And how should we think about it?
Much of the literature on digital ghosts is either focused on their psychological impacts to individuals, or on what we might think of as straightforwardly ethical concerns, focusing on the specific obligations and concerns we may have in interacting with the dead, without much regard for social practices. But this elides an important distinction that emerges in historical literature on care of the dead: the difference between grieving and mourning. Grief is relatively immediate, unintentional, emotional, and experienced by individuals. Mourning takes longer, is more proactive, and is ritualized, with a public, social face.
Shifting our focus from grieving to mourning motivates a turn to Confucian philosophy, notable for its detailed defense of ritual in the service of supporting both social networks and individuals’ emotional capacities. It advocates embedding personal, individualized grief in formal mourning practices, contingent but emotionally potent social rituals.
To the Confucian, much of the current discussion suffers from an excessive focus on digital ghosts’ impact on individuals, overlooking the social coordination issues that may arise in mourning our beloved dead, ones that may both exacerbate and blunt some of these concerns. To the extent that Confucian philosophy focuses on non-medicalized, social, non-individualized public rituals that, while somewhat arbitrary, serve (like traffic laws) to help coordinate social responses to the loss of a group member, it can serve as an important corrective to the more individualized, private grief-centered accounts of digital ghost ethics dominant in the contemporary debate, especially those that treat grief as something to be gotten over, underscoring the value of mourning and thus opening up space for digital ghosts to be incorporated into structured commemorative and supportive rituals.
Digital ghost technologies are criticized for their potential to invite overreliance in times of grief (Fabry & Alfano, 2024; Lindemann, 2022), and compared to addictive substances whose use must be supervised and prescribed by a medical professional. Worries are raised about their inherent risk of inaccuracies, due to their underlying technologies, LLMs which themselves are inherently unreliable (Bao & Zeng, 2024). People worry that they lead us to think of our loved ones as replaceable, to instrumentalize or even zombify our dead (Fabry & Alfano, 2024), or to disregard the significance of death itself (Bao & Zeng, 2024).
Confucian philosophy, however, both cautions against some of the underlying assumptions behind these worries, and suggests strategies that can blunt their force. Their emphasis on relationality and emotional engagement suggest that worries about “too much” grief or dependence get something deeply wrong about our ethical and emotional landscape, and caution against the sanitized, individualist quasi-medical concerns associated with many modern worries about overreliance. Mourning technologies, even when in some sense “unnatural” or “inaccurate”, can matter when and because they help preserve important emotional capacities, including our care for the dead, as invitations to shape and nurture our responses rather than perfectly mimic the deceased, blunting concerns about accuracy. This makes room for digital ghosts as a kind of affective scaffolding for the bereaved, but scaffolding that does different work than that explored by, for example, Kreuger and Osler (2022).
Furthermore, their emphasis on social ritual helps protect against some of these concerns. For example, the Confucian philosopher Xunzi’s discussion of funeral and mourning rituals offers several striking points of connection to the modern phenomenon of digital ghosts. Rituals, he says, are nurturing. They help us to balance emotions and their expression by providing thoughtfully organized structures for experiences of various kinds. This is especially important, he argues, for experiences involving the beginning and end of life. “Ritual is that which takes care to order living and dying. Birth is the beginning of people, and death is the end of people. When the beginning and end are both good, the human way is complete” (Xunzi, 2014, p. 206) Responding appropriately to death is part of responding appropriately to people. If digital ghosts are to be appropriate, they need to help us express our emotions in a balanced way that nurtures the mourners and shows respect for the end of life.
Notably, this is not a solo affair, but a collective practice of evidencing respect for people. Part of this respectfulness, he thinks, involves making death a public affair. Responding to the death of a person should involve not just family and friends, but all those in the neighborhood and district. This should, he claims, extend not just to the funeral itself but to mourning afterward. To the extent that chatbots allow us to pretend “there has never been a funeral” (Xunzi, 2014, p. 208), they do something bad by failing to let us mourn. But to the extent that they allow for public and social engagement with the loss, they may have important work to do in two tasks: helping to keep the memory of the person alive, and keeping us from moving on too quickly. This both identifies important activities that chatbots, used as affective scaffolds, can perform, ones that are intrinsically social, while also introducing social resources and living connections that can help vulnerable people to make decisions with the support and reflective opportunities offered by a caring community, rather than going it alone.
This potential, in turn, is consistent with empirical evidence from users’ interactions with real digital ghosts, where distinctions between the simulations and the simulated are not just recognized, but often crucial to how they engage and what they find valuable about it, from expressing longing, to expressing frustrations about their difficult relationships with the deceased, to sharing memories with subsequent generations (Xygkou et al., 2023).
Recommended Citation
Elder, Alexis, "AI, Grief and Mourning: Connecting with Digital Ghosts" (2025). Value and Responsibility in AI Technologies. 6.
https://repository.gonzaga.edu/ai_ethics/2025/general/6
AI, Grief and Mourning: Connecting with Digital Ghosts
Wolff Auditorium, Jepson Center
Chaired by Tim Clancy, Ph.D. (Gonzaga University)
The advent of Large Language Models has made it possible for people to produce generative AI chatbots that imitate the conversational style of the dead, fine-tuning the system’s outputs to mimic speech patterns by drawing on their text messages, social media posts, emails, and other digital remains. What should we think about this practice? And how should we think about it?
Much of the literature on digital ghosts is either focused on their psychological impacts to individuals, or on what we might think of as straightforwardly ethical concerns, focusing on the specific obligations and concerns we may have in interacting with the dead, without much regard for social practices. But this elides an important distinction that emerges in historical literature on care of the dead: the difference between grieving and mourning. Grief is relatively immediate, unintentional, emotional, and experienced by individuals. Mourning takes longer, is more proactive, and is ritualized, with a public, social face.
Shifting our focus from grieving to mourning motivates a turn to Confucian philosophy, notable for its detailed defense of ritual in the service of supporting both social networks and individuals’ emotional capacities. It advocates embedding personal, individualized grief in formal mourning practices, contingent but emotionally potent social rituals.
To the Confucian, much of the current discussion suffers from an excessive focus on digital ghosts’ impact on individuals, overlooking the social coordination issues that may arise in mourning our beloved dead, ones that may both exacerbate and blunt some of these concerns. To the extent that Confucian philosophy focuses on non-medicalized, social, non-individualized public rituals that, while somewhat arbitrary, serve (like traffic laws) to help coordinate social responses to the loss of a group member, it can serve as an important corrective to the more individualized, private grief-centered accounts of digital ghost ethics dominant in the contemporary debate, especially those that treat grief as something to be gotten over, underscoring the value of mourning and thus opening up space for digital ghosts to be incorporated into structured commemorative and supportive rituals.
Digital ghost technologies are criticized for their potential to invite overreliance in times of grief (Fabry & Alfano, 2024; Lindemann, 2022), and compared to addictive substances whose use must be supervised and prescribed by a medical professional. Worries are raised about their inherent risk of inaccuracies, due to their underlying technologies, LLMs which themselves are inherently unreliable (Bao & Zeng, 2024). People worry that they lead us to think of our loved ones as replaceable, to instrumentalize or even zombify our dead (Fabry & Alfano, 2024), or to disregard the significance of death itself (Bao & Zeng, 2024).
Confucian philosophy, however, both cautions against some of the underlying assumptions behind these worries, and suggests strategies that can blunt their force. Their emphasis on relationality and emotional engagement suggest that worries about “too much” grief or dependence get something deeply wrong about our ethical and emotional landscape, and caution against the sanitized, individualist quasi-medical concerns associated with many modern worries about overreliance. Mourning technologies, even when in some sense “unnatural” or “inaccurate”, can matter when and because they help preserve important emotional capacities, including our care for the dead, as invitations to shape and nurture our responses rather than perfectly mimic the deceased, blunting concerns about accuracy. This makes room for digital ghosts as a kind of affective scaffolding for the bereaved, but scaffolding that does different work than that explored by, for example, Kreuger and Osler (2022).
Furthermore, their emphasis on social ritual helps protect against some of these concerns. For example, the Confucian philosopher Xunzi’s discussion of funeral and mourning rituals offers several striking points of connection to the modern phenomenon of digital ghosts. Rituals, he says, are nurturing. They help us to balance emotions and their expression by providing thoughtfully organized structures for experiences of various kinds. This is especially important, he argues, for experiences involving the beginning and end of life. “Ritual is that which takes care to order living and dying. Birth is the beginning of people, and death is the end of people. When the beginning and end are both good, the human way is complete” (Xunzi, 2014, p. 206) Responding appropriately to death is part of responding appropriately to people. If digital ghosts are to be appropriate, they need to help us express our emotions in a balanced way that nurtures the mourners and shows respect for the end of life.
Notably, this is not a solo affair, but a collective practice of evidencing respect for people. Part of this respectfulness, he thinks, involves making death a public affair. Responding to the death of a person should involve not just family and friends, but all those in the neighborhood and district. This should, he claims, extend not just to the funeral itself but to mourning afterward. To the extent that chatbots allow us to pretend “there has never been a funeral” (Xunzi, 2014, p. 208), they do something bad by failing to let us mourn. But to the extent that they allow for public and social engagement with the loss, they may have important work to do in two tasks: helping to keep the memory of the person alive, and keeping us from moving on too quickly. This both identifies important activities that chatbots, used as affective scaffolds, can perform, ones that are intrinsically social, while also introducing social resources and living connections that can help vulnerable people to make decisions with the support and reflective opportunities offered by a caring community, rather than going it alone.
This potential, in turn, is consistent with empirical evidence from users’ interactions with real digital ghosts, where distinctions between the simulations and the simulated are not just recognized, but often crucial to how they engage and what they find valuable about it, from expressing longing, to expressing frustrations about their difficult relationships with the deceased, to sharing memories with subsequent generations (Xygkou et al., 2023).