It would have been Kate’s 64th birthday on 4 July 2025, had her life not ended in the Princess Alice Hospice on 13 July 2017, shortly after her 56th and final birthday.
Every year we make two memorial pilgrimages to the place where we scattered her ashes, on her birthday and on Boxing Day.
And I write an annual bereavement-related post to mark her birthday, each post marked with the hashtag #Kateday plus the year it was written.
I want to sustain the first ritual for as long as I can; I expect to discontinue the second when I no longer have anything meaningful to say about the experience of bereavement and, more specifically, about the premature loss of one’s life partner.
Up till now I have found some solace in articulating these thoughts, also believing that – if I found them comforting – others in the same situation might do so too.
This year, though, I am taking a slightly different tack…which might mark the beginning of the end.
Last time, I reflected on how memories of our dead partners change over time, partly because we can no longer recall their quintessence, by which I mean their unique selves, exactly as we experienced them when they were alive.
Fairly early in that discussion, I made this throwaway statement:
‘While listening to the radio recently, I heard someone ask what our reaction would be if AI was harnessed to create a lifelike avatar of our deceased partner.
I suspect our instinctive distaste has something to do with knowing that even a perfect replica would not embody the quintessence of our loved one: it would be a soulless mannequin merely.’
Recently, while re-reading that statement, I realised that I knew next to nothing about how bereavement support is beginning to harness artificial intelligence (AI) and virtual reality (VR), the implications of that for both the dead and the grieving, and the ethical issues that are surfacing as a consequence.
So here is a summary of what I’ve learned.
Defining terms
First, some basic definition of terms:
- Social media are interactive, online websites and applications that support users to connect with each other, form communities of interest, and create, share and aggregate content they have generated.
- Artificial Intelligence (AI) is essentially the use of computer systems, machines and software to replicate human intelligence, so they become capable of learning, solving problems and taking decisions without human aid.
- Virtual Reality (VR) is a computer-generated, three-dimensional, interactive environment that provides a multi-sensory experience, potentially drawing on vision, sound and touch. The two principal dimensions of VR are immersion – how faithfully the environment is replicated and how extensively a user may interact with that environment – and presence – to what extent the user feels their virtual experience is really happening, and is happening to them rather than to an avatar that represents them.
- Augmented Reality (AR) is a halfway house, overlaying digital elements on our real world experience, potentially using smartphones, or possibly smart glasses, rather than costly VR headsets.
In the last few years AI has been integrated increasingly into our online and social media experience, principally through the use of AI-generated algorithms that analyse our interests and target material accordingly.
AI also supports content creation, communication and interaction. It can enhance key aspects of VR, including the degree of realism and personalisation of the experience, while also generating the new algorithms required to expand and evolve new virtual worlds.
VR has the potential to become completely integral to our social media activity, but has not yet drawn most users away from their addiction to ‘old-school’ formats including blog posts (WordPress), microblogs (Bluesky, X), pictures (Instagram, Pinterest), videos (YouTube, TikTok) and podcasts.
VR is deployed in numerous three-dimensional virtual worlds, each one created by different developers. The purists refer to a ‘metaverse’ – a single integrated network combining all these virtual worlds into one seamless whole – which has long been the holy grail, but shows little sign of realisation any time soon.
That said, if the ‘metaverse’ is assumed to be the sum total of existing virtual worlds, we can see that activity levels are substantial and growing. There are some 170 worlds in existence, with over 600 million monthly active users across all 170. (But, of course, many individual users will subscribe to more than one world.)
The most popular are those targeted at children and teenagers, the top three being Roblox (250 million monthly active users), Minecraft (165 million monthly active users) and Fortnite (90 million monthly active users). Zepeto, Avakin Life and VR Chat are all popular in the teenage market, as are IMVU and RecRoom
Second Life has existed for well over 20 years. (And, some 15 years ago, my former alter ego Gifted Phoenix led a gifted education workshop there.) Today, Second Life has approximately one million active monthly users.
Types of provision
Some eight years ago Ohman and Floridi identified 57 companies then active in what one might call ‘the digital afterlife industry’, which they divided into four categories:
- first, services to help users with problems relating to the management of digital assets that can arise as a result of death;
- second, digital messaging services that are triggered upon death;
- third, services that offer a digital memorial for the deceased; and
- fourth, services that use personal data to generate new content replicating the deceased person’s social behaviour.
The first three levels seem to me fairly uncontroversial, leaving the fourth as the primary focus of this post. It is in this fourth category that AI and VR are increasingly being used to make this new content more ‘lifelike’.
My reading suggests that this fourth category contains two strands that sometimes overlap: services intended primarily to support bereaved users and services intended primarily for individuals wanting their own ‘digital immortality’.
History
Much of the early thinking seems to have been driven by the pursuit of digital immortality rather than support for bereavement. One can still find online several papers, mostly dating from the years either side of the Millennium, which discuss how digital technology might soon be combined with research into human consciousness to secure this outcome.
Someone called Lee Frank set up an entity called the Digital Immortality Institute to develop and promote understanding of the concept. The website – which still exists though with many of its links defunct – was last updated in 2006.
In February 2011, Russian entrepreneur Dmitry Itskov founded the 2045 initiative, a non-profit organisation, whose main purpose is to:
‘create technologies enabling the transfer of an individual’s personality to a more advanced non-biological carrier, and extending life, including to the point of immortality.’
2025 is the deadline for creating an avatar in which a human brain can be transplanted and, by 2035, the project aims to have created a full computer model of the brain and human consciousness, as well as a means to transfer that into an avatar.
It would be fair to say there is some catching up to do, but scientific progress rarely progresses in a linear fashion.
In 2014, a company called Eterni.me, co-founded by Marius Ursache, a Romanian, was spawned by a MIT entrepreneurship development programme,
It sought to develop AI algorithms from dead people’s digital footprints, drawing on their social media activity, email, photographs and videos, fitness trackers, location information and so on.
Those algorithms could then be used to power the intelligences of three-dimensional avatars, so they could communicate with living relatives through video messaging. The developers stressed that this was primarily a means to curate people’s memories, rather than a vehicle to support grieving relatives.
Eterni.me claimed to have 18,000 potential users signed up and, by late 2016, was being widely described as ‘a Silicon Valley start-up’, planning to test its product for launch in early 2017. The Eterni.me Facebook page was live until 2020; the website finally expired in 2021.
Meanwhile, another Russian, Eugenia Kuyda, had created an AI-powered chatbot to memorialise her Belarusian friend, Roman Mazurenko, who had died in 2015 in a Moscow car accident. She drew on thousands of text messages they had exchanged, as well as material supplied by third parties.
This experience led her to develop Replika at her company, Luka, which she envisioned primarily as a means of generating virtual friends, though it could also be used to interact with the dead.
Replika was launched in November 2017 and, by 2024, Kyuda said the total number of users had reached some 30 million.
The project has run into controversy, however, because of the tendency for some users to form intense, erotically charged relationships with their virtual companions, some even claiming to have fallen in love with them.
Other live projects
Anyone reviewing developments over the last 20-30 years can’t fail to register the irony that online services which aim to offer their customers immortality seem regularly to die, probably because they are either not scalable or not financially sustainable.
Arguably these limitations will continue to plague the market unless and until a genuinely universal metaverse is established. Nevertheless, Replika has several competitors working more directly in the immortality/bereavement field.
They are mostly working in one or both of two overlapping areas, each of them traceable to the work of the early adopters mentioned above:
- Griefbots are chatbots powered by AI that can simulate the written and/or verbal communication of a dead person with those who remain alive.
- Digital avatars, also AI-powered, operate in a VR environment, mimicking the dead person’s physical appearance, facial expressions, voice, physicality and gestures. The aim is to create an avatar that can walk, talk and react just as that person did.
In both cases, successful development depends critically on the consumption of all available data about the subject while they were alive, especially their social media activity. Those with large digital footprints are relatively easier to clone.
Provided that the underpinning AI model is sufficiently sophisticated, such griefbots and avatars may have the capacity to initiate interaction, and possibly even to learn and develop after the originator’s death.
These developers include:
- Somnium Space, based in Prague, which is reportedly developing a ‘live forever’ mode;
- Deepbrain AI, a Korean company, offers a service called Re;memory. It charges $10,000 to create an avatar that your family can subsequently interact with. But the avatar is reportedly fairly primitive, confined to a single ‘mood’.
- HereafterAI is more basic, providing an interactive audio chatbot that will encapsulate someone’s memories in their own voice and respond to questioning by friends and family, but it is apparently all too easy to throw it off course. Sharing unlimited recordings and photographs costs $7.99 per month.
- Eter9 was established in 2015 by the Portuguese Henrique George. The website describes a service where users can set up an avatar known as a Counterpart, whose behaviour is AI-generated and developed largely from the individual’s own posts, comments and online behaviour. There are also adoptable bots called ‘Niners’.
- Silicon Intelligence, a Chinese company, combines a realistic looking deepfake avatar and a griefbot, so enabling users to access the equivalent of a WhatsApp call with a dead person.
- Superbrain, founded by Zhang Zewei, was the first Chinese company in this field in 2022, and is still one of the largest. As of March 2024, they had around 1,000 customers. Depending on the sophistication of the product, prices range from a few thousand yuan to tens of thousands. A decent but fairly basic avatar might cost the equivalent of $2000. Zhang calculates a profit margin of some 30-40%.
- Augmented Eternity is a MIT Media Lab project led by Hossein Rahnama, Visiting Associate Professor at Toronto Metropolitan University, busily creating digital personae that can interact with others after the subject’s death. He has been running a pilot study in which 50 volunteers have been sharing their material so that AI-driven avatars can learn to imitate them.
- You Only Virtual is run by Justin Harrison, a Los Angeles-based film-maker. His Company offers the Versona, designed to allow users to digitally recreate their ‘loved ones’ even after death. The company states that Versonas ‘continue to grow alongside you. They learn, evolve and remember – keeping that unique connection alive, just like a real relationship.’ YOV doesn’t require the prior consent of the deceased.
A Quick Test
In parentheses, I want to mention services that do nothing but animate old photographs. There are dozens of these, some offering a few tries for free.
If you’re unsure about your potential reaction to an AI driven VR avatar of your dead loved one, an easy test is to animate a photograph of them to find out how that makes you feel.
The animation isn’t the best, but it’s good enough.
If you find the result distinctly ‘creepy’, or it makes you feel slightly queasy, then I’d respectfully suggest that the full griefbot or avatar-driven experience may not be for you.
If, on the other hand, you’re intrigued by the possibilities that a much more sophisticated approach might open up for you, or if you wish to investigate your own ‘digital immortality’, there might be benefit in becoming an early adopter of this new technology.
In the two sections below, I’ve briefly summarised some of the arguments for and against, incorporating some of the many ethical questions that arise.
The case for
The standard opening argument in favour of sophisticated griefbots and avatars is that they are no more than the Twenty-First Century equivalent of a photograph, video or film.
When photography was first invented, our early Victorian forbears were equally horrified by its capacity to store the images of their deceased loved ones. Prior to photography, the wealthy had access to paintings, while the poor had only their memories.
Something similar occurred when the phonograph was invented and when ‘moving pictures’ arrived at the turn of the Twentieth Century.
But, over time, we have grown accustomed to these media, learning to treasure photographs, recordings and films if we have them. Photographs, in particular, have become pivotal to our grieving.
To what extent are our reservations about this new technology similar to our ancestors’ reactions? Are we simply wary of what is unfamiliar but will soon become second nature?
Well, maybe, up to a point. But the most sophisticated avatars in development today offer a high degree of realism, effectively embodying the deceased and offering multi-sensory interaction.
Real time interaction through three senses simultaneously is a qualitatively different experience to looking at a photograph or a home video. And we can only expect the level of sophistication to increase. Might there be a risk that the processing of grief is interrupted, even inhibited by a tendency to substitute the avatar for the deceased?
There is a counterargument advanced by some psychotherapists, who consider that avatars might be utilised within what is known as Virtual Reality Exposure Therapy (VRET), to help those who get stuck in the grieving process.
They might also be helpful in cases where partners or close relatives cannot achieve closure after their loved one’s death. For example, in the case of sudden death, particularly suicide, a partner may feel that he or she has not had the chance to ‘say goodbye’, even believing that they are somehow responsible. A controlled conversation might help to resolve these issues.
And, for the wider population, perhaps we worry unduly about over-attachment. Perhaps interacting with a digital avatar may help most grieving relatives to remember their loved ones and to cherish those memories.
There may also be substantial wider benefits to digital immortality if, for example, it ensures that human expertise and creativity remains available to society after the death of its possessor.
There may also be benefit to historians in avatar-enhanced memories of the past, particularly when there are no longer survivors to testify, eg to the Holocaust, or Hiroshima and Nagasaki.
Even in the everyday family context, it might help family members to interact with their more distant ancestors. Two or three centuries hence, family historians will have access to first hand testimony of what it was like to live during the advent of AI!
The case against
The literature seems to have done a rather better job of presenting the downside and associated risks.
As noted above, some have argued that the embodiment of deceased people via VR increases the risk that this might enforce, especially in some more vulnerable and/or suggestible people, a tendency to deny that death has occurred and to attempt to continue, or renew, their relationship with the deceased.
Rather than processing their grief, this might cause them to become completely dependent upon their griefbots or digital avatars, open to manipulation and vulnerable to harm.
It would also make them vulnerable to financial exploitation by companies providing such avatars, who may capitalise on the fact that users have become over-reliant.
Is there a risk that bereaved family members could become emotionally overwhelmed by the interaction with an avatar, yet have no way of escaping, given that the deceased has contracted for the service to continue?
Given the vulnerability of certain types of users, should it be incumbent on companies providing these services to screen their potential customers, rejecting requests from those they judge insufficiently stable? What evidence of their stability should potential customers be required to provide? Should there be an appeals mechanism?
The risks have led some to argue that use ought to be restricted by classifying such avatars and griefbots as medical devices that must be prescribed by a medical professional and used only under medical or psychiatric supervision.
Further, Ohman and Floridi argued that a dead person’s digital remains ought to be treated in the same manner as archaeological remains, perceived to have an inherent value.
It ought not to be possible, for example, for the digital remains of a celebrity to be encapsulated within an avatar and made available to all who are prepared to pay for the service. Digital remains should remain in the possession of those closely related to the deceased.
There is also a risk associated with what is known as the Uncanny Valley effect, where some less-than-perfect feature of the VR experience, or some fault in its delivery, some failure to replicate the deceased accurately, causes negative feelings, or even revulsion in the bereaved person using them.
There are several further ethical issues, some of them relating to the rights of the dead person; others to the impact on the bereaved.
First, in relation to the dead person, there is presently no mechanism that would enable a deceased person to keep control over their digital self after their death.
Are we to assume that we have the consent of the deceased to maintain their avatar after death, even in the absence of a formal agreement? What happens if one or more close relatives object?
It is open to question who becomes liable for fees after death, and what will happen if there is a problem over payment.
There is a risk that the company could be hacked and their data used to impersonate or create a false identity.
Most of the companies engaged in this activity are small scale start-ups with a high risk of failing financially. That has implications for the consistency of service available to the user, and what happens when such companies fail.
If a dead person, via their avatar, continued to create art or music, who would own the copyright, and who would receive any income generated as a consequence?
Because the commercial companies that operate digital avatars will have a vested interest in income generation, they may be tempted to sell data to third parties. Or they may be tempted to ‘sex up’ the avatars of unappealing or introverted dead people so as to encourage bereaved users to interact with them more intensively or for longer.
This would constitute a degree of interference in how the dead person is remembered by those still living: the dead would have lost control over the presentation of their identity.
There is perhaps a wider question related to whether the avatar will always remain faithful to the personality of the deceased. Could they ‘learn’ somewhat different behaviour that might be regard as atypical, at least at the point at which they died?
Will the dead person be preserved at the age when they died, or will they continue to age physically? What are the implications for how the bereaved interact with them?
Should digital avatars even be regarded as participants in society with their own rights? Should they be allowed to be active on social media?
In relation to the bereaved, those interacting with a digital avatar will almost certainly reveal significant information about themselves which the providing company might choose to monetise. This may particularly be the case if the user believes they are in an intense relationship with the avatar.
Conclusion
We are on the verge of a potentially seismic change in how we view mortality and how we grieve the passing of our loved ones.
However, recent rapid technological advancement has been stymied, in part, by the fragmented nature of the market, with dozens of small-scale start-ups struggling to establish and sustain themselves.
None of the mega IT companies is yet active in this field – they will probably wait to buy-out a successful start-up rather than pursue in-house development. Even then, the continuing absence of a single unified metaverse will be a limiting factor.
Those two provisos will probably delay effective provision by a decade or so. Meanwhile, there is an urgent need for regulation to prevent the worst excesses highlighted above.
Ultimately, though, we face a likely collision between two human imperatives:
- our desire for immortality, and to never to be forgotten, which is driven by the human condition; and
- our need to process our grief and survive bereavement intact, which depends on the human capacity to forget much while holding a few precious memories close.
TD
July 2025
timdracup.com 