pith. sign in

arxiv: 2604.11499 · v2 · submitted 2026-04-13 · 💻 cs.CY

Postmortem avatars in grief therapy: Prospects, ethics, and governance

Joshua Hatherley , Sandrine R. Schiller , Iwan Williams , Filippos Stamatiou , Nina Rajcic , Anders S{\o}gaard This is my paper

Pith reviewed 2026-05-10 15:35 UTC · model grok-4.3

classification 💻 cs.CY
keywords postmortem avatarsgrief therapyAI ethicscontinuing bondsfictionalismethical objectionstherapeutic applicationsgovernance challenges
0
0 comments X

The pith

Postmortem avatars can be used ethically in grief therapy because no major objection rules them out in clinical settings.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper investigates whether AI systems that simulate deceased people could help in grief therapy. It proposes two concrete uses: adding the avatars to standard exercises where clients imagine talking to the lost person, and having clients help create the avatar as a creative activity. After reviewing five common ethical concerns including authenticity, consent, and possible harm, the authors conclude these do not block therapeutic use because trained clinicians can keep the risks under control. This approach draws on established ideas about how people grieve by keeping bonds alive and treating the simulation as a helpful fiction rather than a literal replacement. The work matters because it opens a path for new support tools while stressing that rules and real tests are still required to know if the benefits outweigh the downsides.

Core claim

Postmortem avatars, which are AI models fine-tuned on a deceased person's data, can be integrated into grief therapy through imaginal exercises such as the empty chair technique and through the creative process of building the avatar itself. Drawing on the dual-process model of grief and the theory of continuing bonds, the authors treat these avatars as fictional aids rather than literal recreations. They examine five ethical objections and maintain that none of them provide decisive reasons against therapeutic deployment, largely because the structured clinical environment supplies safeguards around consent, authenticity, and psychological safety that are absent in casual uses.

What carries the argument

The framework that treats postmortem avatars as fictional continuing bonds within established grief therapy techniques, allowing ethical objections to be assessed against the protective effects of clinical oversight.

If this is right

  • Therapists could add postmortem avatars to existing imaginal exercises to help clients sustain meaningful bonds with the deceased.
  • The act of creating a postmortem avatar could itself become a structured art-therapy activity that aids grief processing.
  • Clinical contexts reduce the force of ethical objections that apply more strongly to non-professional or commercial uses of the same technology.
  • Governance systems must still develop clear standards for data use, posthumous consent, and professional accountability.
  • Empirical trials are needed to measure actual therapeutic outcomes and to identify any hidden psychological costs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the approach proves workable, families and individuals might begin curating digital archives specifically for future therapeutic use after death.
  • Legal questions about ownership and control of a person's digital likeness after death would become more pressing.
  • The same technology could be tested in related areas such as supporting people with dementia or processing other forms of loss.
  • Commercial services offering postmortem avatars outside therapy would likely require separate and stricter rules to prevent exploitation.

Load-bearing premise

The assumption that clinical therapeutic frameworks and professional oversight will reliably limit harms from inauthenticity, missing consent by the deceased, and unintended emotional effects.

What would settle it

A controlled study finding that patients assigned to use postmortem avatars show higher rates of complicated grief, prolonged distress, or dependency than patients receiving standard therapy alone would undermine the safety argument.

read the original abstract

Postmortem avatars (PMAs) -- AI systems that simulate a deceased person by being fine-tuned on data they generated or that was generated about them -- have attracted growing scholarly attention, yet their potential role in clinical settings remains largely unexplored. This paper examines the ethics of deploying PMAs as therapeutic tools in grief therapy. Drawing on the dual-process model of grief, the theory of continuing bonds, and the philosophical framework of fictionalism, we propose two potential therapeutic applications: incorporating PMAs into established imaginal exercises such as the empty chair exercise, and treating the process of PMA creation as an art-therapeutic exercise in its own right. We consider five ethical objections to these applications and argue that none constitute knock-down arguments against therapeutic use, particularly given the risk-mitigating role of the clinical context. We conclude by identifying outstanding governance challenges and calling for empirical research, without which neither the promise nor the dangers of therapeutic PMAs can be adequately assessed.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper claims that postmortem avatars (PMAs)—AI systems fine-tuned on data from or about a deceased person—can be ethically used in grief therapy via two applications: integrating PMAs into imaginal exercises such as the empty-chair technique, and framing PMA creation itself as an art-therapy process. Drawing on the dual-process model of grief, continuing-bonds theory, and philosophical fictionalism, the manuscript reviews five ethical objections (authenticity, consent, harm, etc.) and argues that none constitute knock-down arguments against therapeutic deployment, primarily because the clinical context supplies risk-mitigating safeguards. It concludes by outlining governance challenges and calling for empirical research to evaluate both benefits and dangers.

Significance. If the central claim is substantiated, the work is significant as one of the first structured ethical analyses of an emerging AI technology in clinical grief therapy. It earns credit for grounding its proposals in established psychological frameworks (dual-process grief model, continuing bonds, fictionalism) rather than ad-hoc invention and for identifying concrete applications and governance gaps instead of issuing blanket endorsements or prohibitions. The balanced treatment of objections and the explicit call for empirical validation provide a useful starting point for interdisciplinary work at the intersection of AI ethics and mental-health practice.

major comments (2)
  1. [section on ethical objections and clinical mitigation] The section discussing the five ethical objections and their rebuttals: the central claim that 'none constitute knock-down arguments... particularly given the risk-mitigating role of the clinical context' is asserted without operational detail. No concrete mapping is provided showing how standard therapeutic safeguards (informed-consent checklists, session protocols, outcome monitoring) would be adapted to AI-specific issues such as the authenticity of generated speech, data provenance from the deceased, or potential attachment to a non-human entity. This leaves the mitigation argument as an assumption rather than a demonstrated buffer, directly affecting whether the objections are shown to be non-decisive.
  2. [section proposing therapeutic applications] The subsection proposing the two therapeutic applications (PMA-augmented empty-chair exercises and PMA creation as art therapy): while the applications are described at a high level, there is no discussion of how the dual-process or continuing-bonds frameworks would be modified to accommodate the non-human, data-derived nature of the avatar, leaving the therapeutic rationale incomplete for the very risks the paper later claims are mitigated by clinical context.
minor comments (3)
  1. [Abstract] Abstract: the five ethical objections are referenced but not enumerated; adding a parenthetical list would improve immediate clarity for readers scanning the abstract.
  2. [References] References: full citations for the dual-process grief model, continuing-bonds theory, and fictionalism should be supplied in the reference list to allow readers to trace the foundational sources.
  3. [Conclusion] Conclusion: the call for empirical research would be strengthened by suggesting at least one or two example study designs (e.g., controlled comparison of PMA-augmented vs. standard empty-chair sessions) to guide future investigators.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed and constructive report. The comments correctly identify places where the manuscript's arguments would benefit from greater operational specificity. We address each major comment below, indicating revisions that will be incorporated in the next version.

read point-by-point responses

  1. Referee: The section discussing the five ethical objections and their rebuttals: the central claim that 'none constitute knock-down arguments... particularly given the risk-mitigating role of the clinical context' is asserted without operational detail. No concrete mapping is provided showing how standard therapeutic safeguards (informed-consent checklists, session protocols, outcome monitoring) would be adapted to AI-specific issues such as the authenticity of generated speech, data provenance from the deceased, or potential attachment to a non-human entity.

    Authors: We accept this assessment. The original manuscript relies on the general claim that clinical oversight supplies safeguards without illustrating how those safeguards would be operationalized for PMA-specific risks. In revision we will add a new paragraph (or short subsection) that supplies illustrative mappings: for example, an informed-consent checklist item requiring disclosure of training-data provenance and model limitations; a session protocol requiring the therapist to log instances of generated speech that deviate from known biographical facts; and outcome-monitoring items that track attachment intensity via validated scales adapted for digital entities. These additions will make the mitigation argument demonstrative rather than assumptive while remaining within the paper's conceptual scope. revision: yes

  2. Referee: The subsection proposing the two therapeutic applications (PMA-augmented empty-chair exercises and PMA creation as art therapy): while the applications are described at a high level, there is no discussion of how the dual-process or continuing-bonds frameworks would be modified to accommodate the non-human, data-derived nature of the avatar, leaving the therapeutic rationale incomplete for the very risks the paper later claims are mitigated by clinical context.

    Authors: We agree that the therapeutic rationale is underdeveloped on this point. The manuscript invokes the dual-process model and continuing-bonds theory but does not explicitly address how their constructs must be adjusted when the 'bond' is with a synthetic, data-derived representation. In the revised version we will insert a brief discussion immediately after the description of the two applications. It will note that continuing-bonds theory already accommodates non-corporeal objects (e.g., photographs, letters) and can be extended to digital avatars provided the patient retains awareness of their fictional status, consistent with the fictionalism framework already cited. For the dual-process model we will indicate that the 'restoration' orientation may require additional therapeutic work to distinguish adaptive engagement with the avatar from maladaptive avoidance of grief reality. These clarifications will tighten the link between the proposed applications and the risk-mitigation claim. revision: yes

Circularity Check

0 steps flagged

No significant circularity; analysis grounded in external frameworks

full rationale

The paper draws on established external psychological and philosophical sources—the dual-process model of grief, continuing bonds theory, and fictionalism—to outline two therapeutic applications and evaluate five ethical objections. No mathematical derivations, fitted parameters, self-definitional loops, or load-bearing self-citations appear in the provided text or abstract. The central claim that clinical context mitigates risks is advanced as an interpretive argument supported by these independent references rather than reducing by construction to the paper's own inputs or prior results. The analysis is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

The central proposals rest on three established external frameworks rather than new postulates or fitted values; no free parameters or invented entities are introduced.

axioms (3)
  • domain assumption Dual-process model of grief
    Invoked to support therapeutic integration of avatars into established exercises.
  • domain assumption Theory of continuing bonds
    Used to justify maintaining connections with the deceased via simulation.
  • domain assumption Philosophical framework of fictionalism
    Applied to treat avatars as useful fictions in therapy.

pith-pipeline@v0.9.0 · 5483 in / 1386 out tokens · 29748 ms · 2026-05-10T15:35:12.870050+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

8 extracted references · 8 canonical work pages

  1. [1]

    Avrahami, D. (2006). Visual art therapy's unique contribution in the treatment of post - traumatic stress disorders. Journal of Trauma & Dissociation, 6(4), 5-38. Balaguer, M. (2018). Fictionalism in the philosophy of mathematics. Stanford Encyclopedia of Philosophy, July 23rd. https://plato.stanford.edu/entries/fictionalism-mathematics/. Bao, A., & Zeng,...

    work page 2006

  2. [2]

    We do it to keep him alive

    Bardideh, F., Jarareh, J., Mofrad, M., & Bardideh, K. (2025). The effectiveness of integrated group therapy on prolonged grief disorder of bereaved people from COVID -19 randomized controlled trial. OMEGA-Journal of Death and Dying, 91(2), 675-692. Beaumont, S. L. (2013). Art therapy for complicated grief: A focus on meaning -making approaches. Canadian A...

    work page 2025

  3. [3]

    Bonanno, G. A. (2004). Loss, trauma, and human resilience: Have we underestimated the human capacity to thrive after extremely aversive events? American Psychologist, 59(1), 20-28. Bos, E. (1995). Making the dead live on: An interactive, talking picture of a deceased person. ACM SIGCAS Computers and Society, 25(1), 7-9. Boss, P . (2009). Ambiguous loss: L...

    work page doi:10.1007/s11097-025-10083-6 2004

  4. [4]

    Jeter, K., & Turns, B. (2022). Grieving the child that never was: Treatment of ambiguous loss in parents of children with Down syndrome. Australian and New Zealand Journal of Family Therapy, 43(2), 243-256. Joyce, R. (2005). Moral fictionalism. In Kalderon, M. E. (Ed.), Fictionalism in Metaphysics (pp. 287-313). Oxford University Press. Klugman, C. M. (20...

    work page 2022

  5. [5]

    Never say goodbye

    Lister, S., Pushkar, D., & Connolly, K. (2008). Current bereavement theory: Implications for art therapy practice. The Arts in Psychotherapy, 35(4), 245-250. Lodhi, A. (2024). “Never say goodbye”: Can AI bring the dead back to life? Al Jazeera News, August 9 th. https://www.aljazeera.com/news/2024/8/9/never-say-goodbye-can-ai- bring-the-dead-back-to-life....

    work page doi:10.1007/s11097-025-10120-4 2008

  6. [6]

    I felt I was t alking to him

    Milmo, D. (2024). “I felt I was t alking to him”: Are AI personas of the dead a blessing or a curse? The Guardian , June 14 th. https://www.theguardian.com/lifeandstyle/article/2024/jun/14/i-felt-i-was-talking-to- him-are-ai-personas-of-the-dead-a-blessing-or-a-curse. Neimeyer, R. A. (2012) Chair work. In Neimeyer R. A. (Ed.), Techniques of Grief Therapy ...

    work page doi:10.1007/s11097-025-10092-5 2024

  7. [7]

    M., & Greenberg, L

    Sharbanee, J. M., & Greenberg, L. S. (2023). Emotion -focused therapy for grief and bereavement. Person-Centered & Experiential Psychotherapies, 22(1), 1-22. Simon, N . M., & Shear, M. K . (2024) Prolonged grief disorder. New England Journal of Medicine, 391(13), 1227-1236. Stokes, P. (2025). Sensing presence: Deathbots and bereavement hallucination. Phen...

    work page doi:10.1007/s11097-025-10064-9 2023

  8. [8]

    Testoni, I ., Pompele, S., Liberale, L., & Tressoldi, P . (2022). Limits and meanings to the challenging territory of mediumship. A qualitative study with grievers. European Journal of Science and Theology, 18(3), 97-109. The Guardian. (2024). Back from the dead: Could AI end grief? The Guardian, September 24th. www.youtube.com/watch?v=5udOx8-QxtE. Voinea...

    work page doi:10.1038/d41586-025-02940-w 2022