From Familiarity to Transformation: Rethinking Griefbots in the Grieving Process

When a loved one dies, the world does not simply lose a person; it loses a future. The shared plans, the inside jokes, the quiet routines, and the specific way that person shaped our daily lives all vanish at once. For centuries, philosophers and psychologists have debated how the living should respond to this void. One influential idea, known as "continuing bonds," suggests that the healthiest way to grieve is to maintain a relationship with the deceased. This does not mean pretending the person is still alive, but rather finding new, adaptive ways to keep their memory and influence active in our lives, much like keeping a cherished habit alive even after the person who started it is gone. In recent years, artificial intelligence has offered a startling new tool for this purpose: the "griefbot." These are chatbots trained on the texts, voice messages, and social media posts of the dead, designed to mimic their speech patterns and personality so the bereaved can still "talk" to them. While these tools promise comfort and a sense of familiarity, a new analysis from researchers at the University of California, Davis, argues that this specific kind of technological comfort may come with a hidden cost. They suggest that by simulating a living conversation with the dead, we might be short-circuiting a necessary, painful, but ultimately healing process of transformation.

The researchers, led by Jack Jacobs and his team, begin by acknowledging the appeal of griefbots. If a parent dies, a child might lose the habit of calling them to share good news or vent about a bad day. A griefbot could recreate that familiar voice, offering a sense of being heard and understood that feels remarkably close to the real thing. Proponents of this technology argue that it helps sustain "habits of intimacy," allowing the grieving person to continue the relationship in a way that feels natural and supportive. The authors of this paper do not deny that these bots can provide genuine relief. They point out that users often report feeling less lonely, more regulated emotionally, and better able to function during the darkest moments of early grief. However, the team argues that looking only at this immediate comfort misses the bigger picture of what grief actually does for a human being.

To understand their concern, one must look at grief not just as a reaction to the past, but as a process of rebuilding the future. The authors propose that grief serves a vital, transformative function. When a person loses someone central to their life, their entire practical identity is shaken. They are forced to ask who they are without that person, what values they hold, and how to move forward in a world that has fundamentally changed. This painful reorganization yields specific, valuable goods: a deeper understanding of one's own life story, a clearer picture of what truly matters, and a new direction for living. The researchers argue that this transformation requires the griever to fully confront the reality of the loss and the absence of the future they once imagined. It is a difficult journey from a world where the loved one is present to a world where they are not, and the journey requires the griever to let go of the old way of relating to make space for the new.

The central finding of the paper is that griefbots, by simulating an ongoing, reciprocal conversation with the deceased, risk blocking this necessary journey. The authors describe a spectrum of risk. At the extreme end, a highly convincing bot could lead to "complete grief denial," where the user feels so much like the person is still there that they never have to face the reality of the loss. The person stops grieving because the bot fills the void so effectively that the absence is no longer felt. At a less extreme but still dangerous level, the bot can cause "partial grief denial." In this scenario, the bot keeps the user stuck in a state of looking backward, constantly engaging with the past rather than turning toward the future. The researchers draw on a psychological model called the "dual process model," which suggests healthy grieving involves a healthy back-and-forth between focusing on the loss and focusing on rebuilding life. They argue that griefbots tend to lock the user into the first mode, making it harder to swing toward the second. By offering a familiar, comforting voice that mimics the dead, the technology can prevent the user from doing the hard work of reinterpreting their life and finding a new way to live forward.

The authors are careful to state that they are not saying griefbots are always harmful or that the comfort they provide is fake. They emphasize that the positive reports from users are real evidence of short-term relief. However, they argue that immediate comfort is not the only measure of a tool's value. Because grief is a transformative experience, its deepest benefits—such as discovering new strengths or redefining one's purpose—cannot always be seen or felt while the person is in the thick of the pain. Just as a person cannot know what a new skill feels like until they have struggled to learn it, a griever cannot know the value of their new life until they have navigated the loss without the old crutches. The paper suggests that by providing a crutch that feels too much like the original support, griefbots may prevent the griever from ever learning to walk on their own again.

In place of the mimetic griefbot, the researchers propose a different design they call a "transformative bot." This system would not try to sound like the deceased person. Instead, it would act more like a wise, caring mutual friend who knew both the person who died and the person who is grieving. This bot would not say, "I am still here with you," but rather, "I know how much she meant to you, and I know how hard it is to face the day without her." It would help the user reflect on the relationship, identify what has been lost, and explore how the values and love of the deceased can still shape the user's future in a new way. For example, instead of simulating a conversation where the deceased gives advice, a transformative bot might ask, "She always encouraged you to be patient. How can you bring that patience into your life now that she is gone?" The goal is not to recreate the past, but to scaffold the difficult work of remembering, reinterpreting, and rebuilding.

The paper concludes that the ethics of grief-related technology should not be judged solely by whether it offers comfort or familiarity. The true measure of success should be whether the technology helps the bereaved navigate the transition from a world of presence to a world of absence. While griefbots aim to preserve the illusion of ongoing connection, the authors argue that a more constructive role for artificial intelligence is to support the living in the difficult, necessary work of letting go and moving forward. By shifting the design focus from simulation to reflection, technology could help people honor their loved ones not by keeping them frozen in time, but by helping the living find a way to live well in the time that remains.