9D Taxonomy of Digital Afterlife Tech

• Digital afterlife technologies are AI systems representing deceased individuals, characterized by key dimensions such as timing, consent, data source, and interactivity.
• The taxonomy maps digital ghosts onto a design cube, detailing practical use cases in legacy preservation, grief support, legal testimony, and entertainment.
• It guides ethical design and regulatory frameworks by highlighting risks like high-fidelity deception and autonomous agency in digital ghost simulations.

Digital afterlife technologies—commonly known as “digital ghosts”—comprise artificial agents, avatars, and simulacra that simulate, represent, or extend the presence of deceased individuals using data-driven AI systems. These technologies now span commercial, institutional, and private settings, impacting domains from bereavement support and legacy preservation to legal testimony and entertainment. To rigorously classify, assess risk, and guide design, researchers have proposed a nine-dimensional taxonomy that delineates the functional, ethical, and operational axes along which digital ghosts vary (Spitale et al., 25 Nov 2025, Morris et al., 14 Jan 2024).

1. Axes of the Nine-Dimensional Taxonomy

The Spitale & Germani (2025) taxonomy (Spitale et al., 25 Nov 2025) and Morris & Brubaker (2024) (Morris et al., 14 Jan 2024) converge on a nine-dimensional design space, each dimension capturing distinct technical or normative points of divergence among systems:

1. Timing of Creation / Deployment Timeline: Whether a digital ghost is constructed and/or activated pre-mortem (with the living subject’s participation) or solely post-mortem. Pre-mortem development enables intentional legacy planning, while post-mortem builds raise questions regarding legitimacy and consent.
2. Consent / Provenance: The form, timing, and provenance of authorization for creating, training, or deploying the ghost. This includes explicit consent from the individual, surrogate authorization (family or estate), or unilateral/implicit creation by third parties.
3. Source(s) of Data: The origin and type of data used to construct the agent—ranging from direct user input (interviews, video), to digital traces (emails, social media), to curated archives (public speeches, testimonials).
4. Interaction & Interactivity: The modalities supported (text, audio, video, multimodal) and degree of generativity or user engagement—ranging from non-interactive memorials to fully generative, adaptive conversational systems.
5. Fidelity & Disclosure / Anthropomorphism Paradigm: The degree of behavioral, vocal, and affective mimicry (fidelity) and whether system status is clearly disclosed to users. This axis includes the distinction between “reincarnation” (posing as the person) and “representation” (an overt simulation).
6. Purpose: The principal intent—a dimension spanning grief support and remembrance, legacy archiving, education, entertainment, legal instrumentation, and more.
7. Audience & Access / Multiplicity: Who can access or interact with the ghost (private, restricted, or public) and whether multiple versions exist for distinct audiences.
8. Governance & Ownership: The infrastructural control—whether digital ghosts are governed and overseen by direct heirs, commercial entities, or public/institutional bodies.
9. Autonomy & Behavioral Agency / Agentic Capability: The extent to which the ghost acts reactively, initiatively, or autonomously (including performing real-world actions).

No quantitative metrics are provided for these axes; the typology remains categorical, but future work may introduce continuous scales such as embodiment fidelity or agency scores (Morris et al., 14 Jan 2024).

2. Criteria, Sub-Dimensions, and Exemplars

Each axis in the 9D taxonomy is populated by explicit subcategories, often matched to real-world services and prototypes.

Dimension	Key Subcategories	Exemplar Systems / Prototypes
Timing	Pre-mortem, Post-mortem	HereAfter AI (pre), Project December (post)
Consent	Explicit, Surrogate, None	Legacy Avatars (explicit), Court Pelkey (surrogate), Griefbots (none)
Data Source	Direct, Traces, Curated	HereAfter (direct), Project December (traces), Deepfakes (curated)
Interactivity	Non/Semi/Fully Interactive	StoryFile (non), HereAfter (semi), Replika (full)
Fidelity/Disclosure	Low/Medium/High; Transparent/Opaque	Replika (medium, often labeled), Deepfakes (high, opaque)
Purpose	Grief, Legacy, Entmt., Education, Legal	HereAfter (grief), StoryFile (legacy), Celebrity Sims (entmt.), Court avatars (legal)
Audience & Access	Private, Restricted, Public	Family legacy bots (private), Fan-site bots (public), Museum avatars (restricted)
Governance	Family/Estate, Corporate, Institutional	Estate-run HereAfter, Corporate Replika, Court Pelkey
Agency	Reactive, Limited, Generative	Scripted video (reactive), Memory-flash bots (limited), Chatbots (high agency)

*Entmt. = Entertainment.

Tables such as this (cf. Spitale & Germani Table 1 (Spitale et al., 25 Nov 2025)) show that each technology is formally situated at a unique intersection in the taxonomy.

3. Conceptual Frameworks and Interaction of Dimensions

Each axis is underpinned by a conceptual metric or framing. For example, Spitale & Germani posit a simple risk function for fidelity and disclosure:

$$\text{Ethical risk} \propto \text{Fidelity} \times (1 - \text{Disclosure})$$

The risk of psychological harm or deception increases when fidelity is high but users are unaware they are engaging with a simulation. Similarly, interactivity and agency are linked to the risk of pathological attachments; high agency (e.g., bots initiating messages or generating new "memories") multiplies emotional entanglement risk (Spitale et al., 25 Nov 2025).

The nine axes together comprise a high-dimensional “design cube,” where every digital afterlife technology realizes one trajectory or region. Concrete archetypes—Legacy Avatars, Private Griefbots, Celebrity Sims, Institutional Avatars—map to distinct vector positions across these dimensions (Spitale et al., 25 Nov 2025, Morris et al., 14 Jan 2024). Some dimensions are effectively binary (e.g., timing), while others (e.g., agency, fidelity, data source) are ordinal or categorical with spectrum-like variation.

4. Ethical and Legal Implications

Each axis highlights unique ethical tensions:

• Timing & Consent: Post-mortem, third-party, or implicit creation can effectively override individual autonomy and risk infringing residual personhood and dignity. Surrogate (family/estate) consent provides only partial mitigation.
• Data Source: Use of uncurated digital traces risks exposing private or unintended content, whereas direct, premortem user-contributed data minimizes this risk.
• Fidelity & Disclosure: High-fidelity ghosts—especially those lacking clear disclosure—can foster harmful illusions or even manipulation. Transparent design with explicit AI labeling mitigates such risks.
• Purpose, Audience & Access: Grave ethical disparities arise between uses for therapeutic/grief support and commercial or entertainment purposes. Public deployment and multiple ghost variants expand the sphere of risk.
• Governance & Agency: Commercial governance can lead to monetization/abuse. High agency bots (e.g., those performing autonomous actions) elevate the potential for manipulation, dependency, or societal impact (Spitale et al., 25 Nov 2025).

These ethical vectors are not equally weighted across dimensions. A system that is post-mortem, high-fidelity, opaque in disclosure, public-access, corporate-governed, and fully agentic occupies a high-risk sector of the taxonomy.

5. Application: Archetypes and Full Taxonomic Mapping

Taxonomic mapping provides an analytic tool for design, risk assessment, and regulatory calibration.

Type	Timing	Consent	Data Source	Interaction	Fidelity/Disclosure	Purpose	Audience	Governance	Agency
Legacy Avatars	Pre	Explicit	Direct	Semi-int.	High/Transp.	Legacy	Private	Estate	Low
Private Griefbots	Post	None	Traces	Full-int.	Med./Varies	Grief	Private	Informal	High
Celebrity Sims	Post	Unclear	Archives	Semi-int.	Med./Opaque	Entmt.	Public	Corp.	Mod.
Institutional Avatars	Pre/Post	Surrogate	Curated	Semi-int.	Med./Transp.	Legal/Ther.	Restricted	Institution	Low

This framework enables comparison and risk forecasting for emerging systems. A plausible implication is that strict ethical design occupies a narrow region: premortem, explicit-consent, direct-data, transparent, private, estate/steward-governed, and reactive agents. Technologies falling outside these safe zones require explicit technical or regulatory guardrails (Spitale et al., 25 Nov 2025, Morris et al., 14 Jan 2024).

6. Prescriptive Guidance and Future Research Directions

Designing ethical digital ghosts involves aligning developmental axes with normative safeguards:

• Premortem intent and explicit consent
• Limited, authorized data sourcing
• Clear, unambiguous disclosure and calibrated fidelity
• Private, opt-in audience profiles
• Family or estate-controlled governance
• Minimized agency (reactive, not generative or autonomous)
• Restricted, purpose-specific deployments (not broad commercialization or ambiguous entertainment use)

Current research catalogs, but does not quantify, these differentiators. Expansion of the taxonomy could introduce formal metrics—embodiment fidelity, audience segmentation, composite agency scores—enabling more nuanced evaluation (Morris et al., 14 Jan 2024). The taxonomy underpins a research agenda spanning risk/benefit modeling, grief outcomes, security, identity, and posthumous rights.

This taxonomy offers a scaffolding for policy formulation, technical standard-setting, and empirical paper, enabling both the systematic classification and the ethical advancement of AI-mediated digital afterlife technologies (Spitale et al., 25 Nov 2025, Morris et al., 14 Jan 2024).