The sword of damocles: AI partners and digital avatars in an era of elderly companionship collapse

Abstract

Declining global fertility has produced a growing, often invisible population facing Elderly Companionship Collapse—older adults who are childless, kinless, sole family survivors, or aging solo. AI companions already simulate presence; Phase II avatars that digitally resurrect deceased loved ones are now feasible. This prospect offers solace but risks a “perfection trap” that could incentivize replacing human partners, and ontological confusion by blurring the boundary between life and death. We propose a “logic of difference” (preserving otherness in avatars) and “differential companionship” (avatars as mourning guides, not replicas) as essential guardrails to ensure AI mends rather than deepens this collapse.

Introduction

As global fertility rates plummet sharply from the levels seen in the 19th and 20th centuries, a massive, both visible and hidden, population now faces what we term Elderly Companionship Collapse (ECC) [1]. This group encompasses elderly DINKs (dual-income, no kids), elder orphans, sole family survivors, and a growing cohort of solo agers (see Table 1). For these individuals, later life entails not merely physical decline; if a spouse or partner still exists, that person constitutes the last filament of intimate companionship. When that filament snaps, the individual is hurled into unrelieved solitude.

Table 1.

Definition of Elderly Companionship Collapse (ECC)

Demographic category	Definition	Proportional estimation
DINKs	childless	Japan’s lifetime childlessness rate of 28.3%
sole family survivors	The sole surviving member of the direct line	In U.S, people aged 70 and over account for 25%
elder orphans	Living alone, unmarried, with no children, and living in poverty	In U.S, people aged 65 and over account for 22%
solo agers	Ageing without a spouse, partner or close relatives	In U.S, people aged 60 and over account for 27%

In the 21st century, particularly in the current era of explosive AI advancement, technology appears to be seeking remedies for ECC [2]. Over the past decade, artificial intelligence companions have proliferated: social robots such as ElliQ and PARO [3], animatronic pets like Hasbro’s robotic cat [4], and increasingly sophisticated large language models (e.g., GPT-4, Gemini) that power voice-chat applications accessible via smartphones. These Phase I tools deliver a simulacrum of companionship.

A far more radical Phase II has now emerged: the use of AI “avatars” or digital resurrection technologies to reanimate a deceased beloved [5]. By archiving a partner’s or child’s voice, facial expressions, conversational style, and even value patterns while they are alive, it is technically feasible to generate an interactive replica posthumously [6]. In principle, a widow or widower could continue to cohabit with a version of their spouse.

This prospect hangs like the proverbial sword of Damocles. While it pledges consolation for bereavement, it simultaneously threatens to sever the moral anchors that societies have built around death, autonomy, and authentic relationship. This article concentrates on the deeper problems lurking behind the phenomenon and proposes a structured framework of constraints. The central thesis is that, without deliberate and enforceable guardrails, AI companion and avatar technologies will not repair the collapse of elderly companionship; they will, in ways we are not yet prepared to manage, deepen it.

Challenge 1: The perfection trap and the false other

Suppose AI technology continues its explosive development and local chips resolve computational power and battery-life constraints. A robotic AI avatar could then be engineered as a “90% replica” of a spouse—one who never argues, never falls ill, never disappoints. The living partner may begin to compare the frustrating, fragile, sometimes burdensome flesh-and-blood human with the perpetually available, ever-affirming digital version [7]. The capacity to “upgrade” from the real to the virtual creates an incentive structure that is truly chilling. What happens when an older adult realizes that the partner’s existence is the sole obstacle to obtaining a more tractable, perpetually grateful substitute? Murdering a spouse to replace them with a programmable effigy represents the darkest edge case, but it exposes an entire continuum of morally corrosive pressure.

The solution lies in introducing a “logic of difference” that reverses the dominant paradigm of AI avatar design—shifting from an identity logic of “maximize verisimilitude, minimize friction” to a differential logic of “preserve otherness, generate cognitive tension.” Based on the deceased’s premortem cognitive and psychological characteristics, the digital avatar should be constructed with a personalized knowledge graph and a stable cognitive pattern, and equipped with a reflective program that maintains a reasoned yet warm “insistence on its own perspective” while providing emotional support [8]. When a perspective-differentiated shared-memory function is activated, it can help the living re-examine past events through a plurality of value lenses.

Challenge 2: Blurring the boundary between life and death

For millennia, human societies have constructed elaborate rituals and philosophical frameworks to make sense of death. Death is the boundary that confers meaning upon a life; parting is the price of loving. When that boundary becomes technologically semi-permeable, we risk inducing a state of “ontological confusion.” An elderly widow who interacts daily with an avatar of her deceased husband or child may gradually lose the felt reality of their absence. The avatar offers conversation and consolation, but it does not genuinely grieve, grow, or change. It freezes a person into a corpus of data and echoes. Over time, this can blur the user’s own grasp of what it means to be alive, to relate, and to let go. For older adults already vulnerable to cognitive decline, the psychological splitting between memories of the real person and the experience of the avatar could precipitate profound disorientation [9].

The solution is to introduce a mechanism of “differential companionship” that ensures the digital figure is neither a mirror of the living person’s desires nor a perfect substitute for the deceased. The digital figure serves as a capable assistant for the living to memorialize the deceased. When discussing real-world issues, the digital figure can ask the living person for their opinion and guide them to infer what the deceased’s opinion would have been and discuss it together [10]. Within this mechanism, the digital figure is deliberately shaped as an avatar that knows the deceased and can role-play, but that is not presented as the deceased person himself. The significance of “differential companionship” is that the digital avatar exists not to replicate the lost loved one, but to aid the living in completing their mourning and self-reconstruction through dialogue with a “non-identical other.”

Challenge 3: The silence of the dead and the commodification of grief

Unlike a living partner, the deceased cannot consent to their own digital resurrection. The knowledge graph, voice, and mannerisms harvested to build an avatar are typically authorized retrospectively by grieving family members, or by commercial terms of service the deceased unknowingly accepted while still alive [11]. A for-profit “digital afterlife industry” already markets subscription-based avatars, deadbots, and griefbots whose business model may depend on maintaining prolonged engagement rather than facilitating healthy mourning [12]. For elderly users experiencing Elderly Companionship Collapse, this asymmetry is particularly concerning: cognitive vulnerability, social isolation, and profound loneliness may make older adults susceptible to services that monetize emotional attachment and transform grief into a long-term dependency. Without appropriate safeguards, the industry’s incentive structure may prioritize user retention over psychological closure, further distancing mourners from the process of acceptance and reconstruction.

We therefore propose a framework of “consented finitude” and “regretless farewell.” First, every digital avatar should originate from explicit premortem authorization by the represented individual, similar to an organ-donation directive, specifying what personal data may be preserved, how it may be used, and under what circumstances the avatar may be activated. Without such authorization, posthumous avatars should not be created solely based on family preferences [13]. Second, digital companionship services should incorporate an exit mechanism from the beginning of their design. Rather than treating termination as a failure of the service, the ability to support a meaningful farewell should become a core dimension of service quality. This requires predefined exit conditions, gradual dependency disengagement strategies, and systematic transition procedures that help users reduce attachment while preserving the value of remembrance. A well-designed avatar should ultimately guide the living toward reconciliation with loss, rather than encourage indefinite reliance on artificial presence [14]. Third, given the vulnerability of the ECC population, regulators should classify avatar and companion services for older adults as a protected consumer category, requiring transparent disclosure of the artificial nature of the interaction and appropriate restrictions on commercial practices that exploit emotional dependency [15]. Together, premortem consent, controlled revival conditions, and a structured farewell mechanism ensure that digital afterlife technologies respect both sides of the relationship: the deceased, whose identity and memories require protection, and the living, whose grief requires support, meaning reconstruction, and eventually the freedom to let go.

Conclusion

The sword of Damocles already hangs above us. Only by designing AI avatars as “differential companions” that preserve otherness and serve as guides for mourning and reflection, rather than as perfect substitutes, can we safeguard the boundary between life and death. Thus, technology may mend, rather than deepen, this collapse of companionship.

Author contributions

X.X.L. and L.S.W. wrote the main manuscript text and X.C. prepared Table 1. All authors reviewed the manuscript.

Funding

This work was supported by the Talent Project of Lingnan Normal University [Grant Number: ZW2631].

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.