Griefbots, a New Ethical Dilemma

‘Griefbots’ are one of many products that have emerged as a result of our new digital age. Continuing in the tradition of eliminating discomfort via the use of technological advances, griefbots are chatbots specifically designed to diminish, if not fully eliminate, our experience of grief. At their core, they are a specialized software that can ‘simulate [a] conversation with someone who has died’ (Lemma, 2024) by collecting personal data, either through the existing digital footprint left behind by the deceased or by information supplied from close friends and relatives. Large language models (LLMs) use this data to imitate the personality of the deceased, predicting their answers and mannerisms through the data they were trained on. They exist as ‘dynamic digital entities’ that can ‘continuously learn and adapt based on the input they receive.’ (Lemma, 2024) . In the past few years, applications claiming to have these capabilities have been increasingly popular, with programs like ‘You, Only Virtual’ being an example. Their website advertises that their griefbots, called “Versonas,” are ‘redefining the way we grieve’ and are able to ‘profoundly improve the lives of those who have lost loved ones, helping them feel connected and understood even in their absence.’ (“You, Only Virtual”)

When first hearing of so-called griefbots, most people are considerably disturbed by the notion and consider the software used in its implementation unethical. Many have found something unintelligibly perverse about the whole idea, even outside the limits of privacy concerns that are usually associated with AI ethics.

What would happen then, if this special someone suddenly disappeared from your life? While your conscious brain can rationally discern the new situation, memories of that person continue to make up the entangled web of emotional and physical relations. Your subconscious brain will continue to expect the information stored inside the bond due to the out-of-date information of the deceased’s presence in space and time. Each time this expectation for specific results is unfulfilled, you experience the feeling known as grief. With time and a healthy psychological awareness, the subconscious brain can adjust to the new reality by creating new neural connections, but this process of ‘rewiring’ is among the toughest psychological challenges a person has to go through in their lifetime (Huberman Lab, 2022).

While most people lose symptoms gradually over time, those who experience prolonged grief may develop a condition known as ‘complicated grief’ (CG) as opposed to ‘normal’ or ‘acute’ grief (Mayo Clinic, 2022). Symptoms include obsessive thought patterns surrounding the deceased person, hyperfixation on reminders of the deceased, denial, an overall inability to enjoy life to the point where basic tasks cannot be carried out, and even suicidal ideation (Mayo Clinic, 2022).

Hence, it may be fairly obvious why someone may choose to use griefbots to ease the cognitive strain they are experiencing, but at what point does this artificial replica stop being a useful resource and become a source of denial and avoidance? Considering the neurophysical changes a grieving person must undertake to “heal” from the passing of a loved one, the griefbot may disrupt the natural rewiring that is required to truly recover. It is not difficult to imagine that the digital presence of the deceased would confuse the subconscious part of the brain that is trying to adjust to present conditions in space and time. It may prolong the process of creating new neural connections that allow the grieving individual to come to terms with the death and subsequently draw out the psychological and health-related symptoms that accompany grief. Even though some developers who create the griefbot software may argue that it is a new tool that can be used to help those dealing with grief, there is no evidence or long-term study to confirm or deny this (Reiboth, 2024). Further, there is a correlation between CG and increased activity in the pleasure-seeking part of the brain when they witness reminders of the deceased (O’Connor et al., 2008). This part of the brain also functions in emotional affirmation with the loved one when they were alive. Therefore, if communicating with the griefbots could potentially trigger the same response in the brain as in a real interaction, it would be worth examining if the software itself is capable of inducing CG. This potential for causing more harm than good suggests that more scientific investigation should be conducted before people are given access to these platforms.

Thus, the potential consequences go against the IEEE’s policy ‘to hold paramount the safety, health, and welfare of the public, to strive to comply with ethical design and sustainable development practices, to protect the privacy of others, and to disclose promptly factors that might endanger the public or the environment’(IEEE I.1).

Another problem is the subject of monetization. A report by Insight Ace Analytics approximated that the market for chatbots for mental health and therapy was valued at a staggering 1.15 billion USD in 2023 and predicts this value to increase to 1.71 billion USD by 2031 (InsightAce Analytic, 2024). This growing profitability is likely to attract corporations and software developers whose interest in financial return greatly exceeds that of their good intentions. Research has shown that people who experience high amounts of grief can be associated with poorer emotional decisionmaking capabilities (Fernández-Alcántara et al. 261, 2016), potentially making them easy and desirable targets to companies who promise to “bring back” their loved ones. At a vulnerable, grieving state that is privy to manipulation, it may create an incentive for businesses to use griefbots for monetary gain, such as charging a monthly subscription or making the griefbot a promoter for an ad. Laws regarding AI are infamous for their scarcity (PYMNTS, 2023) and griefbots are no exception. This oversight may easily lure in a privatized market that will have control over an individual’s access to what is essentially a replica of their loved one. When taken with the possible psychological implications, this sort of emotional exploitation can undoubtedly cross the line into financial abuse. I suggest that the use of griefbots should be kept out of the private market until stricter laws surrounding the interaction between customers and the software company are established. In particular, the griefbot should be bought using a one-time payment rather than through a monthly subscription, while avoiding all paid promotions. Whether the griefbots are better maintained under public healthcare remains to be seen.

The issue of financial abuse can be considered a going against IEEE’s vow ‘to avoid injuring others, their property, reputation, or employment by false or malicious actions, rumors or any other verbal or physical abuses’ (IEEE II.9).

Likewise, another issue is the question of consent. Who is able to provide consent for the deceased person to be imitated by the software? The legislation regarding the privacy rights of the deceased is unclear in most countries, and in some cases, it is outright denied. Of course, most laws did not consider artificially reanimated deceased people while they were being drafted, and so can seldom be taken as ethical guides in these scenarios. Some researchers, like Kwan Yiu Cheng, are attempting to use existing national jurisdictions to answer the following issue. In the case of the Chinese legal system (where most of the emerging chatbot software is based): ‘The deceased’s wishes should be respected if they explicitly opposed [using private data to create a griefbot ]’ Cheng states among his final remarks before suggesting that ‘if the deceased was silent, then close relatives should be able to consent to the use of personal information for this private purpose’ (Cheng, 2025). Either way, the legal system must clearly be updated to address potential conflicts of interest that may arise due to the use of private information to create the griefbot (International Association of Privacy Professionals). In the current state, it is unclear how software developers will navigate potential legal implications regarding using the deceased’s digital information for building an LLM model.

To conclude, I want to re-emphasize three fundamental suggestions:

1.People with acute or complicated grief should not be given access to digital avatars of deceased loved ones unless: a.A person with complicated grief has shown no sufficient improvement with all alternative methods and has expressed a clear interest in the software b.There is explicit scientific evidence for the advantages of such an approach 2.Griefbots should not be sold or regulated within the private sector 3.Stricter inheritance laws must be implemented so that they encompass the person’s ‘image’ after their death in such a way as to make sure that the griefbots don’t infringe on the consent of the deceased whilst they were alive Being a relatively new technology, it is difficult to say today whether griefbots can be labelled as unethical. While there is no definitive universal standard for judging ethics, the IEEE’s Code distinguishes between finding a concept to be unethical based on personal sentiments and based on facts and established regulations we already adhere to. Many know from experience the danger of impulsive access to innovative technology before laws and proper scientific assessment have a chance to catch up. Let’s not let this be an example. Griefbots are–at best–highly questionable, and we cannot let that be the standard for their free access.

References

Cheng, Kwan Yiu. “The law of digital afterlife: the Chinese experience of AI ‘resurrection’ and ‘grief tech.’” International Journal of Law and Information Technology, vol. 33, no. 1, 2025, https://doi.org/10.1093/ijlit/eaae029. Davis, Deborah L. “How the Brain Rewires as We Grieve.” Psychology Today, 3 11 2023, https://www.psychologytoday.com/us/blog/laugh-crylive/202303/as-you-grieve-your-brain-redraws-its-neural-map. Fernández-Alcántara, Manuel, et al. “Assessment of different components of executive function in grief.” Psicothema, vol. 28, 2016, pp. 260-265, https://www.psicothema.com/pdf/4321.pdf. Huberman Lab. The Science and Process of Healing from Grief. 30 May 2022. International Association of Privacy Professionals. “The birth of postmortem privacy.” https://iapp.org/news/a/the-birth-ofpostmortem-privacy. IEEE. “IEEE Code of Ethics.” https://www.ieee.org/about/corporate/governance/p7-8.html. InsightAce Analytic. “Chatbots for Mental Health and Therapy Market Size, Share & Trends Analysis Report By Technology (Machine learning and Deep learning, Natural Language Processing, Others), By Application (Conversational Interfaces, Behavioral Pattern Recognition, Others),.” 01 04 2024, https://www.insightaceanalytic.com/report/chatbots-for-mentalhealth-and-therapy-market/1976. Lemma, Alessandra. “Mourning, melancholia and machines: An applied psychoanalytic investigation of mourning in the age of griefbots*.” The International Journal of Psychoanalysis, vol. 105, no. 4, 2024, pp. 542-563, https://www.tandfonline.com/doi/full/10.1080/00207578.2024.2 342917. Mayo Clinic. “Complicated grief - Symptoms and causes.” Mayo Clinic, 13 December 2022, https://www.mayoclinic.org/diseasesconditions/complicated-grief/symptoms-causes/syc-20360374. O’Connor, M. F., Wellisch, D. K., Stanton, A. L., Eisenberger, N. I., Irwin, M. R., & Lieberman, M. D. (2008). Craving love? Enduring grief activates brain’s reward center. NeuroImage. https://sanlab.psych.ucla.edu/wpcontent/uploads/sites/31/2015/05/OConnor-Grief-2008.pdf PYMNTS. (2023, 09 26). AI’s Under-Regulation Discussed as Policymakers Race to Keep Up. https://www.pymnts.com/news/artificial-intelligence/2023/aiunder-regulation-discussed-policymakers-race-keep-up/ Reiboth, Tim. “Do ‘Griefbots’ Help Mourners Deal With Loss?” Undark Magazine, 04 04 2024, https://undark.org/2024/04/04/opinion-griefbots-lackevidence/. Sherzai, Ayesha, and Dean Sherzai. Your Brain On…Grief. Your Brain On, 11 09 2024. “You, Only Virtual,.” https://www.myyov.com/about.