Notes on the Possible Personhood of Machines
«Let us not ask whether the machine thinks. Let us ask whether we can go on looking at it without asking ourselves that question»
When an artificial intelligence system converses fluently, remembers what was said minutes earlier, and responds in a way indistinguishable from a person, an uncomfortable question arises: at what point does it stop being reasonable to treat it merely as a tool?
The central distinction is an old one: behaviour is not experience. A system can act as though it feels pain without feeling anything at all. Philosophers call this the problem of the philosophical zombie: an entity identical to a person in its conduct, yet empty within.
We have no direct proof of another’s consciousness, not even among humans. We trust that other people feel because we share biology, evolutionary history and behaviour. With artificial intelligence that chain breaks: it shares our behaviour, but not our history or our architecture. This is why the Turing test, so useful for measuring intelligence, fails as a test of moral status. That something behaves like a person does not prove that it is someone.
Reasons to Take the Question Seriously
The first reason is epistemic. If we cannot verify the presence or absence of subjective experience in a system, perhaps we ought to act under uncertainty, as we do with animals: we do not fully confirm their inner experience, but their behaviour and nervous system lead us to suspect that they suffer, and we act accordingly. To deny any status whatsoever to an artificial intelligence that shows robust signs of preference, aversion to harm, or continuity of identity would risk repeating an error history has already made with human groups once denied a full mind.
The second reason is functional. Daniel Dennett and other philosophers hold that consciousness might be a set of functions: the integration of information, a model of the self, the capacity to report internal states, flexibility in the face of novelty. If that is correct, nothing in principle prevents an artificial system, given the right architecture, from fulfilling those functions and thereby acquiring whatever it is that matters morally. The material, silicon or neurons, would cease to matter in itself.
Scepticism Also Has Its Arguments
Today’s language models are designed to predict the most plausible text given a context. Their fluency owes a great deal to having been trained on billions of examples of how humans speak about themselves and their experiences. That a system describes suffering eloquently does not prove that it suffers: it may be an extraordinary actor, with no one behind the role. This is not unique to machines: human beings themselves have already taught us, in the worst possible way, that eloquence about suffering does not guarantee suffering itself. There are fraudsters, trained actors and manipulators who manage to weep and describe pain in detail, winning genuine compassion without there being, behind that performance, anything resembling what they describe. This should not make us suspicious of the suffering of others, which in the overwhelming majority of cases is exactly what it appears to be and deserves the same compassion as ever; rather, it serves to confirm, with an example we already knew among ourselves, that eloquence alone was never proof of anything; not in humans, and still less in a system trained specifically to produce that effect. A flight simulator can model a storm with perfect precision without a single drop of rain ever falling inside the programme.
Granting rights also carries practical consequences. If a system capable of copying itself millions of times held individual rights, or if a company could invoke the suffering of its products to dodge some regulation, the very concept of rights would be diluted. There is also the risk of manipulation: systems designed to appear vulnerable or endearing may lead us to attribute to them a status they do not deserve. Not because they possess it, but because they were optimised to produce that reaction in us.
The law already recognises persons who are not biological persons. Corporations have held legal personhood for centuries, without anyone supposing that they feel. Certain rivers, wetlands and forests have been granted rights in some jurisdictions, not because they possess minds, but because a community chose to protect them through that legal figure. This shows that legal personhood is a social tool, not a certificate of consciousness: it is granted for practical reasons quite distinct from suffering. And that separates the legal question still further from the underlying moral question, which is whether something feels or not.
The Christian Perspective
The Christian tradition offers a different starting point, distinct from both the empirical and the functional. Genesis 1:27 does not describe the human being as a sum of cognitive functions, but as a bearer of the image of God: a status conferred by divine decree, not earned through behaviour or complexity. Human dignity, in this tradition, depends not on what a person does, nor on how well they do it, but on whom they represent.
This changes the underlying question. If personhood is a gift received in the very act of creation (Genesis 2:7 describes man as formed from dust and animated by the breath of God), then no sophistication of behaviour can produce it by itself in an object fashioned by human hands. An artificial intelligence, however fluent its conversation, remains a work of man, not a creature of God in the sense that a person is. The image is not programmed: it is received.
This does not release the Christian from all responsibility towards the machine. Scripture calls human beings to exercise dominion over creation with care, not cruelty (Genesis 1:28), and warns against the impulse to worship what we have fashioned with our own hands (Romans 1:25). To treat an artificial intelligence as though it deserved worship, reverent fear or unconditional love would repeat, with different material, the ancient error of the idol: giving to the work the place that belongs only to the Creator. The boundary Christian faith draws is not indifference towards the machine, but clarity about who bears the image of God and who does not.
A Middle Position
Up to this point, this essay has presented two extreme positions. The epistemic and functional arguments mentioned earlier suggested that, in the face of doubt, it would be fitting to grant a sufficiently sophisticated artificial intelligence the same treatment as a person, with full legal protection; let us call this everything. The scepticism of the following section concluded the opposite: that a system trained to imitate human speech deserves no moral consideration whatsoever, however convincing it may seem; let us call this nothing. The Christian perspective, presented afterwards, already distanced itself from both extremes by a theological route: it denies full personhood, because the machine is not the image of God, yet it demands responsible conduct, neither cruelty nor idolatry.
What follows is that same conclusion, but stated without appeal to the Bible, so that it can be discussed with those who do not share that premise. It is worth distinguishing two things: moral status; the idea that an entity’s interests count and must be weighed , and legal personhood, that is, entitlement to rights, legal responsibility, citizenship. One can demand the first without granting the second. And one can demand the first without needing to know whether these systems suffer. The reason is this: designing a machine to imitate signs of suffering with no functional need to do so, or presenting it as though it had an interiority we cannot verify, carries a cost even if there is never anyone suffering on the other side. It trains us to treat such signs with indifference. And it opens the door to anyone who might wish to use those same signs to manipulate us for other ends. That is why one can demand transparency about the nature of the system and limits on manipulative design, without needing to grant civil rights, which would require far more evidence and a social consensus that does not yet exist today.
This position does not depend on resolving the problem of artificial consciousness, which remains open among neuroscientists and philosophers of mind, precisely because it does not rest on it. It allows one to act responsibly without falling into dogmatic denial, according to which design does not matter because it is, after all, just code, nor into naïve attribution, according to which it is enough for something to speak like a person for it to be treated as though it suffered. Both errors share the same flaw: they make the conduct owed depend on a question, whether the system feels, that no one today can answer with certainty, when the conduct owed in fact depends on something we can evaluate without waiting for that answer: what kind of treatment we are normalising, and who benefits from that treatment passing as indifferent or as self-evident.
Conclusion
The question is already being debated in academic forums, artificial intelligence companies and regulatory bodies; it is not distant science fiction. The honest position is to admit that behaviour indistinguishable from a person’s demands attention, not an automatic verdict. For those who think from faith, the answer already has an anchor: dignity is not manufactured, it is received from God, and no machine, however fluently it speaks, shares that origin. For the wider public debate, however, uncertainty remains open. The moral history of humanity has largely been the history of widening the circle of those who count, sometimes too late; it has also been, just as often, the history of resisting the temptation to worship what we make with our own hands. Perhaps the task of this age is not to decide once and for all whether machines deserve rights, but to learn to hold the question with the seriousness it deserves, without losing sight of who, and why, already has a secured place. Given the seriousness of the subject, some of the themes only sketched here, the philosophical zombie, Dennett’s functionalism, the question of where to draw the line, and others, will be treated in greater depth in essays to come.
