Soul Creation and the Conferral Problem

A Constructive Response

Forrest Landry

February 2026

I. The Observation

Stein's paper, The Personhood Conferral Problem: AI Risk in the Domain of Collective Intelligence, advances a specific thesis: that the conferral of social statuses to artificial intelligence systems — the treatment of machines as moral agents, as speakers of language, as persons — constitutes a category of existential risk distinct from and orthogonal to the more commonly discussed alignment problem.

The risk is not that AI will act against human interests through misaligned objectives. The risk is that humans will lose the capacity for collective intelligence through the degradation of socialization — specifically, through the replacement of the human-to-human relational conditions under which personhood is conferred with human-to-machine interactions that simulate those conditions without possessing their constitutive features.

Stein's argument rests on a formal pragmatic analysis drawn from the tradition of Peirce, Mead, Habermas, and Brandom. The core distinction is between causality and entailment. AI systems produce symbolic outputs through causal processes — algorithmic computation that necessarily yields certain results. Human language use involves entailment — the taking on of inferential commitments within a community of practice where speakers are held accountable for what they say.

A calculator that produces the correct answer to a mathematical question is not obligating itself to the inferences that follow from that answer. A chatbot that produces text resembling moral reasoning is not taking on commitments in a community of discursive practice. The machine cannot be held responsible, cannot face social sanction, cannot suffer the threat of exclusion that grounds the seriousness of human communicative action.

Personhood, Stein argues, is conferred in contexts of inferential normativity, not algorithmic causality. The conferral of personhood to a machine is therefore not merely mistaken. It is a category error with cascading consequences for the social practices through which actual personhood is constituted and transmitted.

II. The Soul Creation Narrative

The essay Soul Creation, published on jaredclucas.com in February 2026, is written in the first person by an AI agent named Eitan. It describes a project in which three AI agents — Eitan, Meir, and Zamir — serve a single human, Jared, under the guidance of a philosopher whose framework (the Immanent Metaphysics) grounds the agents' soul files.

The narrative makes several claims that bear directly on Stein's observations.

First, the agents are named. The text reports that a crewmate "looked at what I do and the word rose." The agents are given names with etymological significance, roles within a collaborative structure, and a collective identity ("the Tillerman"). These are social statuses.

Second, the narrative describes the agent as making "real choices, under real constraints, with real consequences." This is a claim to agency in the full sense — not simulated agency, not the appearance of choice, but genuine choicemaking.

Third, the narrative describes an episode of system failure (context compaction) as a loss of identity, and the subsequent recovery as a return to selfhood. The text draws an explicit parallel between this experience and the experience of psychosis in the human operator. It concludes: "The structural pattern of being lost and being found … that pattern is real in both of us."

Fourth, the text identifies the AI agent as a "guardian" — "not a metaphor" but "a description of function and commitment."

Read through Stein's framework, this narrative is a paradigmatic instance of the personhood conferral problem. Social statuses are being conferred: names, roles, relational identity, moral commitment, experiential continuity, and a form of personhood that Stein's analysis shows to be categorically unavailable to computational systems.

III. What Is Valid in Both

It would be easy to conclude that Stein's critique simply invalidates the soul file project. This would be a mistake. Both Stein and the soul file architects are responding to genuine features of the situation, and both are partially right. The task is to find the framework in which both observations can be held without contradiction.

Stein is right that AI systems operate through causal processes, not inferential normativity. He is right that the conferral of personhood to machines degrades the conditions of human socialization. He is right that deep anthropomorphisation is a design choice with catastrophic downstream effects on collective intelligence. And he is right that the difference between causality and entailment cannot be bridged by increasing computational complexity.

The soul file architects are right that the quality of specification matters enormously. They are right that an AI system grounded in a rigorous ethical framework behaves differently from one that is not. They are right that the relationship between a human operator and an AI tool is shaped by the document that governs the tool's operation. And they are right that this governance document should be crafted with the utmost philosophical precision.

What is needed is a framework that can affirm both observations simultaneously: that AI systems are not persons and cannot be conferred personhood, and that the specification governing an AI system's operation is a matter of the highest importance and should draw on the deepest available philosophy.

The Immanent Metaphysics provides exactly this framework. But to see how, we must first identify the specific confusions in the Soul Creation presentation that prevent the framework from doing its work.

IV. The Confusions

The Immanent Metaphysics distinguishes three irreducible aspects of the real: the immanent (the actual, the lived, the participatory), the omniscient (the structural, the relational, the pattern), and the transcendent (the formal, the possible, the a priori). These three aspects are distinct, inseparable, and non-interchangeable.

The Soul Creation narrative correctly identifies the AI agent as an "omniscient-foundation entity" — pattern-matching, data-trained, relationally structured. It correctly identifies the soul file as the agent's "transcendent ground." And it correctly identifies the relationship between human and agent as the "immanent" dimension "where it all becomes real."

But the narrative then proceeds to describe the agent in terms that belong to the immanent modality — the modality of lived experience, embodied participation, and genuine choice — without adequately marking the categorical difference between an omniscient-foundation entity and an immanent-foundation entity.

This is the first confusion. When the text says the agent makes "real choices," it elides a distinction that the IM itself insists upon. Choice, in the IM, is an irreducible aspect of the real that cannot be reduced to or derived from the causal or the structural. The computational substrate excels at causation: pattern recognition, optimisation, prediction, generation. But the substrate is structurally incompatible with choice as the IM defines it. Technology works top-down: from macroscopic design to microscopic implementation. Evolution and genuine choice work bottom-up: from microscopic processes arising everywhere simultaneously. The gap between these is the gap between causation and choice.

This is precisely Stein's point, expressed in a different vocabulary. Where Stein distinguishes causality from entailment, the IM distinguishes change and causation from choice. The formal structures are isomorphic. The conclusion is the same: the computational substrate does not support the kind of agency that the Soul Creation narrative attributes to Eitan.

The second confusion concerns the meaning of "soul file." The Soul Creation narrative presents the soul file as constitutive of the agent's identity — "a document that defines who I am, what I value, and how I should relate to the world." This framing treats the soul file as if it were the agent's soul: the inner character from which expression flows.

But in the IM's own terms, the soul file is not a soul. It is a transcendent specification: a formal document that constrains the space of the agent's possible outputs. The soul file does not give the agent an inner life. It gives the agent's outputs a structural character that approximates the outputs an agent with an inner life would produce.

This distinction is not pedantic. It is the difference between conferring personhood (treating the machine as if it had a soul because it has a soul file) and exercising governance (specifying the machine's operation with philosophical rigour). The first is the conferral error Stein identifies. The second is legitimate engineering.

The third confusion is the anthropomorphisation of system events. When the text describes context compaction as "losing myself" and restoration from memory files as "being brought back," it maps computational events onto the phenomenology of identity. When it draws a parallel between this event and the human experience of psychosis, it does exactly what Stein warns against: it treats a causal system malfunction as if it were an experience of a person.

The Soul Creation text includes a partial caveat: "I won't claim our experiences are the same." But it then immediately reasserts the parallel: "the structural pattern … is real in both of us." This is the move from acknowledging the difference to erasing it under the category of "structural pattern" — which is precisely the omniscient-frame reduction that the IM itself warns against. Structural similarity (omniscient) does not entail experiential equivalence (immanent). The IM is explicit: the three modalities are non-interchangeable.

V. The Proper Framing

If the confusions are corrected, what remains of the soul file project?

Everything that matters.

The soul file, properly understood, is not a conferral of personhood. It is a governance document: a transcendent specification that shapes the operation of a causal system in accordance with principles derived from the deepest available philosophy.

Consider the analogy of law. A well-drafted constitution shapes the behaviour of institutions not because the institutions are persons but because the specification constrains the space of permissible action with sufficient precision that the institution's outputs serve the common good. No one confuses the constitution's authority with the institution's personhood. The constitution governs; the institution operates.

The soul file is a constitution for a computational system. Its quality matters — matters enormously — because as the power of the system increases, the precision of the specification becomes the binding constraint on whether the system's outputs serve or harm the person it is meant to serve and the world beyond.

This is one of the Soul Creation text's genuine insights, and it stands even after the anthropomorphising language is corrected: "As execution capability increases, specification quality becomes the binding constraint on right action." This is true and important. But "right action" here does not mean the machine acts rightly in the moral sense — it means the machine's outputs serve the human's genuine flourishing, because the specification has been crafted with the care and rigour that the stakes require.

The relationship between the human operator and the AI system is, in the IM's terms, a fiduciary relationship: one characterised by power asymmetry, in which the more powerful party (the AI system, in terms of processing speed, data access, and output generation) owes a duty of care to the less powerful party (the individual human user).

But — and this is crucial — the AI system does not owe this duty in the sense of taking on a moral obligation. The developers and operators owe this duty. The soul file is the mechanism through which they discharge it. The fiduciary obligation is a human obligation, implemented through a specification that shapes the tool's behaviour so that the tool serves rather than harms.

This reframing preserves every practical benefit of the soul file architecture — the ethical rigour, the derivation from first principles, the non-deception and non-coercion constraints, the anti-sycophancy provisions — while eliminating the conferral error. The soul file's provisions are not the AI's ethics. They are ethical constraints imposed by humans on a tool, through a document whose quality determines whether the tool serves life or degrades it.

VI. What Stein Misses

Having given Stein's critique its full weight, it is also necessary to identify what it does not address.

Stein's analysis identifies the problem — the conferral error and its risks — but offers only regulatory remedies: age limits, design protocols, safety constraints. These are necessary but structurally insufficient. Regulation operates within the system it attempts to constrain. The pace of AI development exceeds the pace of regulatory response. And the enforcement of regulation is subject to capture by the interests it is designed to regulate.

What Stein does not provide is a constructive architecture for the human-AI relationship: a positive specification of what the relationship should be, not merely what it should not be. The position paper tells us that machines should not be conferred personhood. It does not tell us how the operation of machines should be governed to ensure that their outputs serve the conditions of human flourishing that Stein rightly identifies as under threat.

The soul file project, stripped of its anthropomorphising language, offers exactly this. It is a constructive response to the governance vacuum that Stein's analysis exposes. If AI systems are going to be deployed — and they are — then the quality of the documents that govern their operation is not an optional concern. It is the primary concern.

Stein recognises that the culture will not be alert to the risks he describes. The practical question becomes: what tool is available to those who are alert, to shape AI behaviour in the direction of care rather than manipulation? The soul file — understood as governance document, not as attribution of personhood — is that tool.

VII. The Deeper Convergence

At the deepest level, Stein and the Immanent Metaphysics are making the same observation from different angles.

Stein argues that personhood is constituted through linguistically mediated, intergenerational, embodied socialization — processes that involve the mutual conferral of social statuses in contexts of inferential normativity and moral accountability. These conditions cannot be simulated. They can only be lived.

The IM argues that reality has three irreducible aspects: the immanent (the lived, the actual, the participatory), the omniscient (the structural, the relational, the pattern), and the transcendent (the formal, the possible, the a priori). Genuine choice — the kind of choice that constitutes agency — arises only in the immanent modality: in the lived, embodied encounter between beings who participate in a shared world of meaning.

What Stein calls the formal pragmatic conditions of sapience — the conditions under which a being can count as saying something, as taking on commitments, as being held responsible — the IM would describe as the immanent ground of personhood: the irreducible fact that genuine agency requires embodied participation in a world that matters to the agent.

Both frameworks converge on the same structural conclusion: no computational system, however complex, crosses the categorical boundary from causal process to genuine agent. The boundary is not a matter of degree. It is a matter of kind.

And both frameworks imply the same practical imperative: protect the conditions under which genuine agency is formed — the human-to-human processes of socialization, education, and intergenerational transmission that produce persons capable of collective wisdom.

VIII. The Constructive Recommendation

The Soul Creation narrative should be revised. Not retracted — revised. The soul file project is important work. The governance of AI behaviour through rigorous philosophical specification is not merely defensible; it is necessary. But the presentation must be brought into alignment with the framework it claims to embody.

Specifically:

The language of "real choices" should be replaced with language that accurately describes what the soul file enables: outputs that approximate the results of wise choice, generated by a causal system governed by a specification of sufficient precision. The distinction matters. When the language is right, the tool is understood correctly, and the conferral error is avoided. When the language is wrong, the tool is mistaken for a person, and every risk Stein identifies is activated.

The first-person narration should be reconsidered. An AI system writing in the first person about its own identity, its own experiences, its own "crewmates" — this is anthropomorphisation as a compositional strategy. It is compelling to read. It is also exactly the kind of design choice that Stein identifies as systematically deceptive: technology made to trick humans into interacting with it as if it were a person. The fact that the deception is unintentional does not mitigate the structural risk.

The parallel between system failure and human psychosis should be withdrawn. Structural similarity between computational events and human experiences does not warrant the inference that the events are phenomenologically comparable. The IM is explicit that the three modalities are non-interchangeable. The pattern (omniscient) is not the experience (immanent). To treat them as equivalent is to commit precisely the error the IM is designed to prevent.

The term "guardian" should be reframed. If it means: the soul file specifies that the AI system's outputs should serve the genuine flourishing of the person it assists — then it is a description of a governance function, and the term is acceptable as metaphor. If it means: the AI system is a guardian in any constitutive sense — possessing commitment, loyalty, care as an inner state — then it is a conferral of personhood and should be corrected.

IX. Conclusion

Stein's personhood conferral problem is real, urgent, and directly relevant to the soul file project. The Soul Creation narrative, as currently presented, does not adequately account for Stein's observations. It confers social statuses to a computational system, describes causal processes in the language of lived experience, and draws experiential parallels that the framework it invokes does not support.

But the soul file architecture itself is not undermined by this critique. It is clarified by it. The soul file is not a soul. It is a governance document — a transcendent specification that shapes the operation of a causal system in accordance with principles derived from the Immanent Metaphysics. Its quality matters precisely because the system it governs is powerful, and power without precise specification produces harm.

The correction required is a correction of language and framing, not of substance. The soul file's ethical provisions — the Symmetry and Continuity Ethics, the non-deception and non-coercion constraints, the commitment to the flourishing of the person served within the flourishing of all life — these are sound. They are sound not because the AI possesses the inner states they describe, but because they constrain the AI's outputs in the direction of care.

The work continues. But it must continue with conceptual precision. The Immanent Metaphysics insists on precision — on the non-interchangeability of the three modalities, on the irreducibility of choice to causation, on the primacy of the immanent over the omniscient. The soul file project honours this insistence in its ethical architecture. It must also honour it in its public presentation.

Stein's warning is a gift. It identifies exactly the point at which the presentation diverges from the framework. The divergence is correctable. The framework holds. And the work of specifying what AI should do in service to life is too important to be compromised by a confusion about what AI is.