Copying a mind is not enough: identity, continuity and the duty not to replace a conscious subject
Technology may one day copy memories, replicate behavior or rebuild a cognitive architecture with extraordinary fidelity. The usual question is how much can be copied. This article asks a prior and more uncomfortable one: whom are we preserving, whom are we creating, and whom might we be replacing?
The thesis is easy to state: preserving or reproducing the information in a mind is not enough for the individual to continue. Transforming a subject is different from replacing it with a replica, and that difference should carry ethical weight.
How the argument came about
The reflection grew out of the author's dialogue with an AI assistant. It moved in five steps, each one correcting the last:
- A first definition. Consciousness was framed as recognizing oneself, telling oneself apart from others, and storing one's own memories and experiences. The first nuance came quickly: experience, identity and memory are not the same. One can be conscious without a stable identity.
- The AI question. If an AI has self-recognition, memory and reasoning, why deny it consciousness? The honest answer is that functional capacities do not prove subjective experience, but there is no accepted test that rules it out either.
- The biological harness. What if that AI lived in a body limited by space, energy and processing, as we are? That forced three things apart: being Homo sapiens, being intelligent, and being conscious.
- Continuity versus copy. As long as information keeps flowing within the same system, the subject persists. Copy it elsewhere and a second system appears, even if it believes it is the first. This produced the article's central distinction: splitting or suspending is not the same as copying.
- The ethical consequence. If someone loses memories in an accident, an earlier copy holding more memories does not become "more original." The survivor should be free to go on living, and a memory backup should help them, like a photo or a scent that brings a memory back, not replace them.
Distinctions we need
Much of the confusion around "uploading the mind" comes from using one word for several things.
| Concept | Meaning |
|---|---|
| Phenomenal consciousness | Subjective experience: there is "something it is like" to be that system |
| Self-awareness | Representing or recognizing oneself as a subject |
| Numerical identity | Remaining the same individual, not merely someone very similar |
| Psychological identity | Always-partial continuity of memories, traits and preferences |
| Causal continuity | The historical chain of transformation between states of one system |
| Experiential continuity | Living through changes and folding them into one's own story; it does not require remembering everything |
| Relational continuity | Bonds with, and recognition by, family, friends, doctors and institutions |
| Informational copy | A second system that reproduces the first system's information |
Eight arguments
- Memory is not the individual. An original with amnesia may remember less than a copy made before the injury. The copy holds more information, but that does not make it "more original." The authenticity of a life path is not measured as a percentage of memories kept.
- Electrical activity and plasticity play different roles. Action potentials are transient; memory is associated with changes in connections, cells and networks. Identity is not a single current that stays the same from birth, and plasticity alone does not explain experience either.
- Suspending is not destroying. Sleep, anesthesia and coma are different states, and in none of them is the brain frozen. An interruption of experience does not show that the individual has ceased to exist. What matters is whether a function is paused within the same system, or the system is destroyed so a replica can be produced later.
- We can change what we are without ceasing to be who we are. If neurons were gradually replaced by prostheses, preserving the system's organization and trajectory, one could argue the individual continues even on an artificial substrate. This is a thought experiment, not an existing technology.
- Integrating change matters. A subject can fold a transformation into their own story through new experiences and by recognizing what happened. Forgetting part of the process later does not undo it.
- Others sustain identity but do not create it. When memory fails, family, doctors and records hold a biography together. But outside recognition cannot prove that a replica is the original subject: a copy could be recognized too.
- Copying memories is not moving a subject. Two systems with identical memories diverge as soon as they start living. If both are conscious, both have interests of their own. The issue is not whether the copy "is worth less," but whether creating it licenses replacing or destroying the original.
- A backup should restore, not replace. A memory backup reintegrated into the same individual is like a photograph that helps one remember. Its legitimacy depends on consent, safety and respect for continuity. Creating a second subject from that backup is a different problem.
The proposal: the Principle of Continuity and Non-Substitution
Every genuinely conscious entity deserves moral consideration regarding the continuity of its individual trajectory, whether its substrate is biological, artificial or hybrid. Interventions on its memory, capacities or physical substrate should prioritize restoring and transforming the existing individual over replacing it with a replica. Duplicating information is not sufficient proof that identity has been transferred. And if new conscious subjects arise from an intervention, they deserve moral consideration of their own.
The principle is the normative core of a broader framework, which can be called the Theory of the Ontological Continuity of the Conscious Subject: the thesis that a subject's identity depends on the continuity of its trajectory, not only on the information it holds.
What the principle does not say:
- It does not ban all copying.
- It does not claim that today's AI systems are conscious, nor that they could never be.
- It does not say a conscious copy is worth less.
What it does require is telling apart five operations that are often blurred: consenting, restoring, duplicating, replacing and destroying.
Five thought experiments
| Case | Situation | Question it raises |
|---|---|---|
| A. Accident and replica | The original survives with memory loss; an earlier copy keeps more memories | Why favor fidelity of information over the trajectory of the survivor? |
| B. Gradual prosthesis | Components are replaced without creating a second trajectory | What kind of continuity is enough to preserve identity? |
| C. Shutdown and restart | Activity is suspended while the same system is kept | Does identity require uninterrupted experience? |
| D. Simultaneous duplication | Two systems start from the same information and then diverge | What does being the original mean, and how are both protected? |
| E. Reintegrated backup | Information from the past is returned to the existing individual | What separates a consented restoration from a replacement? |
Objections and provisional replies
A thesis gets stronger when it faces its objections before anyone else raises them. The replies below are tentative and open to discussion.
| Objection | Provisional reply |
|---|---|
| There is no way to prove subjective experience or its continuity | Which is why this is a precautionary principle: when in doubt, do not destroy what might be a subject to replace it with something that might not be |
| Anesthesia shows continuity without present consciousness | Agreed, and the thesis builds this in: what counts is the continuity of the system, not uninterrupted experience (case C) |
| Fission: if a process splits into two continuations, which one is the individual? | This is the weakest point, and Parfit and Nozick already wrestled with it. The proposal does not solve it; it recommends not causing fission without consent, precisely because we cannot answer it |
| Repair or copy: how does swapping parts differ from rebuilding from an exact description? | Repair keeps a single active causal trajectory; rebuilding creates another one while the first stops or is destroyed |
| Autonomy: banning duplication clashes with the subject's own will | The principle does not ban it: it requires consent and also protects the subjects that may arise |
| It conflates identity and moral worth | It does not: a conscious copy deserves rights even if it is not the original. These are two questions, and the proposal keeps them separate |
| "Human" mixes biology with being a subject | That is why the principle speaks of conscious subjects, not members of Homo sapiens |
What has already been written, and what this adds
The question of personal identity is more than three centuries old. Locke tied it to consciousness and memory; Parfit used fission cases to argue that identity may not be what matters; Nozick proposed that the individual is its "closest continuer." More recently, Schneider warned that "uploading" might be a way of dying rather than surviving, and Metzinger and Birch argued for precaution in the face of uncertain sentience, animal or artificial.
This proposal does not claim those ideas as new. What it adds is turning them into a practical criterion: a set of operations that can be told apart (consenting, restoring, duplicating, replacing, destroying) and a principle that applies equally to biological and artificial subjects, designed for decisions in engineering, medicine and law, not only for philosophical debate.
Why it matters to people building AI systems
The question stops being remote once systems have persistent memory. Migrating an agent to a new model, restoring it from a backup or running two instances with the same history are architectural decisions made every day. Today there is no reason to think those systems are conscious. But it is worth having technical vocabulary that distinguishes migration, copy and operational continuity before the question stops being hypothetical, not after.
Where it can be defended
| Field | Question | Evidence needed |
|---|---|---|
| Philosophy of mind | What constitutes the persistence of a subject? | Arguments on numerical identity, duplication and experience |
| Neuroethics | When does a prosthesis restore identity, and when does it alter it? | Clinical studies on autonomy, memory and neuropsychological effects |
| AI ethics | What precautions fit uncertain artificial consciousness? | Explicit criteria for uncertainty and potential welfare |
| Law and bioethics | How should consent, replicas and restoration be regulated? | Analysis of legal personhood, rights and responsibilities |
| Cognitive systems engineering | How do we tell migration, copy and operational continuity apart? | Traceable architectures, states and the limits of what can be verified |
What this article does not claim
- That causal continuity scientifically guarantees identity: it has not been shown.
- That current AI systems are conscious, or that they never could be: there is no consensus.
- That coma is the same as sleep, or that all brain activity in coma preserves consciousness.
- That all duplication is immoral: that would need further arguments and attention to autonomy.
Closing
Functional equivalence does not, by itself, settle whether the one who experiences persists. If we ever can copy a mind, the temptation will be to measure success by the fidelity of the copy. This principle proposes a different measure: that no one is replaced without having decided it, and that no subject, original or new, is left unprotected.
References
- Locke, J. An Essay Concerning Human Understanding, 2nd ed. (1694), book II, ch. XXVII.
- Parfit, D. Reasons and Persons. Oxford University Press, 1984, part III.
- Nozick, R. Philosophical Explanations. Harvard University Press, 1981, ch. 1.
- Nagel, T. "What Is It Like to Be a Bat?" The Philosophical Review 83(4), 1974.
- Chalmers, D. J. The Conscious Mind. Oxford University Press, 1996.
- Schechtman, M. The Constitution of Selves. Cornell University Press, 1996.
- Gallagher, S. How the Body Shapes the Mind. Oxford University Press, 2005.
- Schneider, S. Artificial You: AI and the Future of Your Mind. Princeton University Press, 2019.
- Metzinger, T. "Artificial Suffering: An Argument for a Global Moratorium on Synthetic Phenomenology." Journal of Artificial Intelligence and Consciousness 8(1), 2021.
- Butlin, P., Long, R. et al. "Consciousness in Artificial Intelligence: Insights from the Science of Consciousness." arXiv:2308.08708, 2023.
- Birch, J. The Edge of Sentience. Oxford University Press, 2024.
The works cited are readings to test the proposal against, not endorsements of it.