AI’s Role in Grief: Bringing Deceased Loved Ones Back to Life
Wertynews.com – Jack Manning’s personal journey highlights the emotional and logistical difficulty of honoring a loved one’s memory. When he lost his sister over a decade ago, his family dedicated countless hours to raising funds for a volunteer initiative in her name. “It took years of bake sales and 5K races to secure the resources needed to create a meaningful tribute,” Manning shared with CBS News. “We realized how much effort goes into crafting a lasting memorial. That’s when I wondered if technology could offer a more accessible solution.”
Manning, a doctoral candidate in information science at the University of Colorado Boulder, is now exploring how large language models (LLMs) can generate what some call “generative ghosts.” These AI chatbots are trained on data like social media posts, emails, audio clips, video recordings, and photographs to recreate a digital version of the deceased that friends and family can engage with after their passing.
According to Manning, this field is expanding rapidly. “There’s increasing interest in how AI intersects with grief,” he noted. Companies such as Séance AI, You, Only Virtual, and re;memory are leading the charge in developing these services. For instance, re;memory allows users to build lifelike avatars using photos and voice recordings, offering a way to converse with a deceased loved one as if they were still present.
A Lifelike Digital Presence
“By interacting with a digital avatar of your loved one, you can reconnect with the cherished moments you hold dear,” the re;memory website states. The platform emphasizes that users can “see and hear their parents as they were,” rather than relying on imagined recollections.
Technological Simulations of Grief
In a June study published by the Association for Computing Machinery, lead researcher Jed Brubaker and his team explored users’ responses to AI-generated representations of lost individuals. Participants provided details about their loved ones’ lives and traits, then engaged with chatbots designed to mimic their communication styles. “The participants were amazed by how much an LLM could do with minimal input,” Brubaker explained.
Two Approaches to Digital Reincarnation
Manning identified two core types of generative ghosts. The simpler “death bot” replicates recorded statements but lacks the ability to simulate a person’s conversational style. “A single prompt on a free ChatGPT account can produce a basic portrayal of a loved one,” Brubaker said. “You could ask it to mimic a grandfather’s mannerisms using brief details about his behavior.”
Generative ghosts, however, go further by replicating voice and communication patterns, allowing for interactive dialogue. “These models can express thoughts a loved one might not have shared during life,” Manning explained. “They act as a digital extension, answering questions in a familiar tone.”
Costs and User Preferences
Users pay varying fees for these services. re;memory charges $24 monthly for three custom avatars, while Séance AI offers animated images for $19.99 per month. The study found that users generally favored first-person interactions with AI “ghosts,” finding the chatbots more relatable than third-person avatars. “If a bot used a term of endearment that their loved one never expressed, users often decided to end the conversation,” Brubaker said.
Deepfakes vs. Generative Ghosts
While generative ghosts share similarities with deepfakes, Brubaker emphasized a key distinction. “Deepfakes are created to mislead others,” he stated. “Ghosts, on the other hand, aim to provide comfort rather than deception.” This difference in purpose shapes how these technologies are used, with generative ghosts serving as companions for grieving individuals rather than tools for misinformation.
