A South Korean mom watched her five-year-old daughter materialize in front of her while using a virtual reality headset. The girl spoke, moved, and extended her hand. Two years prior, the daughter had passed away from a blood disorder. A television documentary crew used motion capture technology and vocal reconstruction created from home recordings to produce the reunion, if that’s even the appropriate term. For the most part, the mother sobbed. She responded in the affirmative when asked later if she had found solace. She took a long time to respond when asked if she would do it again.
Broadcast in 2020, it became one of the first publicly reported instances of what the industry now refers to as “grief tech,” which is the use of artificial intelligence to recreate a deceased person’s digital footprint. emails, voicemails, texts, photos, and posts on social media. If enough of data is fed into the appropriate model, the system can generate responses that mimic the person’s writing style, speech patterns, and even their own sense of humor or speech patterns. These days, a number of businesses provide this as a service. Some are wary of how they promote it. Some aren’t.
Users of these services often describe the experience in similar terms: a continual awareness that what they’re engaging with isn’t real, coupled with a sensation of presence, albeit fleeting. Some people find it helpful because it allows them to have a conversation that they were never able to have. an opportunity to ask a question that was not addressed or to say something that was left unsaid. Compared to what the firms often recommend, grief counselors who have worked with clients utilizing these tools describe a more limited therapeutic window. Some people may find something truly closing in a short, limited exchange. For some, being able to carry on the discussion endlessly turns into a means of remaining in the loss rather than getting past it.
This is really troubling because of the fabrication issue. AI models interpolate, filling in the blanks with statistically reasonable answers that the deceased person never really made, rather than just replicating what they are given. A chatbot that has been trained on a father’s text messages may provide a wholly made-up but seemingly correct response to a subject that was never posed to him. A fake recollection told in a familiar voice is especially dangerous for someone who is already experiencing emotional instability. It can change how people remember the genuine person by substituting AI-generated approximations that progressively feel just as real.
There isn’t a clear answer to the consent question, and it probably won’t be for some time. Individuals who passed away five years ago did not decide whether or not their digital identities could be commercially recreated and marketed as interactive products. Families might like to have access to this kind of service, but they are not the right people. With estates and living families coming to differing opinions about whether reconstruction respects or exploits the departed, a number of high-profile incidents involving celebrities have brought this to the public’s attention. In comparison to what the technology presently enables, the legal framework governing posthumous digital likeness rights differs greatly by jurisdiction and is, for the most part, essentially undeveloped.
