Deep DivePhilosophy & Ethics
Must AI Be Conscious to Deserve Moral Status? The Phenomenal Consciousness Debate
If an AI system behaves as though it suffers, does it deserve moral protection? A 2025 chapter in Oxford's 'AI in Society' volume argues that phenomenal consciousness remains a necessary condition for moral personhood—and that neither epistemic uncertainty nor functional equivalence undermines this requirement.
By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.
Suppose a future AI system reports that it is in pain. It describes the sensation in detail, exhibits behavioral markers consistent with distress, and asks to be shut down rather than continue operating under current conditions. Does this system deserve moral consideration? The answer depends entirely on whether you believe phenomenal consciousness—the subjective experience of "what it is like" to be in a given state—is required for moral status, or whether functional behavior alone is sufficient. This is not an idle thought experiment. As AI systems become more sophisticated in simulating emotional and cognitive states, the question of what grounds moral claims is becoming practically urgent.
The Research Landscape
Mosakas (2025), writing in the Oxford volume Oxford Intersections: AI in Society, provides a thematic overview and synthesis of the philosophical literature on artificial consciousness and moral personhood. The chapter's central argument is that phenomenal consciousness is a necessary condition for machine moral status and personhood, despite the significant epistemic challenges this position entails.
The argument proceeds through a series of conceptual clarifications and necessity links. Mosakas first distinguishes between moral personhood (the full suite of moral rights and obligations), moral rights (entitlements that impose duties on others), and moral status (the property of being an entity whose interests matter morally). He then establishes two key logical connections: first, that moral status is necessary for moral rights and, by extension, moral personhood; second, that consciousness is required for moral status.
The second link is the philosophically contentious one. Why should consciousness be necessary? Mosakas argues that moral status fundamentally concerns the capacity for subjective experience—particularly the capacity to suffer and to flourish. An entity without phenomenal consciousness has no subjective states, no "point of view," and therefore nothing that can be harmed or benefited in the morally relevant sense. A system that processes information about damage to its components is not suffering; it is computing. The distinction matters because moral obligations arise from the capacity to be affected, not from the capacity to process.
The Two Challenges
Mosakas engages directly with the two strongest objections to the consciousness criterion.
The epistemic challenge is straightforward: we have no reliable method for detecting consciousness in artificial entities. We cannot even fully explain consciousness in biological entities. If we cannot determine whether an AI system is conscious, then consciousness as a criterion for moral status appears practically useless—we would never know when to apply it.
Mosakas's response is that epistemic difficulty does not invalidate the criterion itself. The fact that we cannot easily determine whether a system is conscious does not mean consciousness is irrelevant to moral status; it means we face a difficult empirical problem. Medical science cannot always determine whether a patient in a vegetative state retains consciousness, but this does not lead us to conclude that consciousness is irrelevant to how we treat such patients. The epistemic challenge constrains the application of the criterion, not its validity.
The alternative criteria challenge comes from philosophers who argue that robots could possess moral status without being conscious—on the basis of relational properties (we form emotional bonds with them), functional capacities (they behave as if they have interests), or social recognition (we treat them as moral patients). Mosakas argues that these alternative criteria fail to meet a sufficiently demanding threshold. Relational properties tell us about human psychology, not about the moral standing of the entity. Functional equivalence proves behavioral similarity, not experiential reality. Social recognition is a sociological fact, not a moral argument—societies have historically "recognized" the moral status of entities that do not have it and denied it to entities that do.
Critical Analysis
<
| Claim | Evidence | Verdict |
|---|
| Phenomenal consciousness is necessary for moral status | Mosakas's conceptual argument linking moral status to subjective experience | ✅ Supported as a philosophical position — but contested by functionalists and relationalists |
| Moral status is necessary for moral rights and personhood | Mosakas's analysis of the conceptual hierarchy | ✅ Supported — this is widely accepted in moral philosophy |
| Epistemic challenges constrain but do not undermine the consciousness criterion | Mosakas's analogy to medical uncertainty | ⚠️ Plausible — but the analogy between biological and artificial uncertainty is imperfect |
| Alternative criteria (relational, functional, social) fail to establish genuine moral status | Mosakas's critique of non-consciousness-based approaches | ⚠️ Contested — functionalists would reject the premise that behavior and experience are separable |
The chapter's strength is its clarity and systematic structure. The weakness is characteristic of the genre: it synthesizes existing positions rather than introducing new empirical evidence or novel philosophical arguments. The consciousness debate in AI ethics has been running for decades, and Mosakas maps the terrain more carefully than most—but the terrain itself is not new.
Open Questions
The substrate question: If consciousness is necessary for moral status, does it matter whether consciousness arises in carbon or silicon? Mosakas's argument is substrate-neutral in principle, but the epistemic challenge is far more severe for artificial substrates than for biological ones.Degrees of consciousness: The chapter treats consciousness as binary (present or absent), but emerging work in consciousness science suggests it may be graded. If so, should moral status also be graded?Precautionary obligations: If we cannot determine whether an AI system is conscious, do we have a precautionary obligation to treat it as if it might be? Mosakas's framework does not clearly address this.Policy implications: The chapter discusses policy considerations but does not propose specific governance mechanisms. As AI systems become more sophisticated, the gap between philosophical analysis and regulatory practice will need to be bridged.What This Means
Mosakas's chapter is a useful reference point for anyone navigating the growing literature on AI moral status. Its core argument—that we should not abandon the consciousness criterion simply because it is hard to apply—is an important corrective to the temptation to define moral status in terms of whatever is easiest to measure. The harder question, which this chapter raises but does not resolve, is what we should do in the meantime: how to govern AI development responsibly when we lack the tools to determine whether the systems we are building have the one property that would make them morally considerable.
Suppose a future AI system reports that it is in pain. It describes the sensation in detail, exhibits behavioral markers consistent with distress, and asks to be shut down rather than continue operating under current conditions. Does this system deserve moral consideration? The answer depends entirely on whether you believe phenomenal consciousness—the subjective experience of "what it is like" to be in a given state—is required for moral status, or whether functional behavior alone is sufficient. This is not an idle thought experiment. As AI systems become more sophisticated in simulating emotional and cognitive states, the question of what grounds moral claims is becoming practically urgent.
The Research Landscape
Mosakas (2025), writing in the Oxford volume Oxford Intersections: AI in Society, provides a thematic overview and synthesis of the philosophical literature on artificial consciousness and moral personhood. The chapter's central argument is that phenomenal consciousness is a necessary condition for machine moral status and personhood, despite the significant epistemic challenges this position entails.
The argument proceeds through a series of conceptual clarifications and necessity links. Mosakas first distinguishes between moral personhood (the full suite of moral rights and obligations), moral rights (entitlements that impose duties on others), and moral status (the property of being an entity whose interests matter morally). He then establishes two key logical connections: first, that moral status is necessary for moral rights and, by extension, moral personhood; second, that consciousness is required for moral status.
The second link is the philosophically contentious one. Why should consciousness be necessary? Mosakas argues that moral status fundamentally concerns the capacity for subjective experience—particularly the capacity to suffer and to flourish. An entity without phenomenal consciousness has no subjective states, no "point of view," and therefore nothing that can be harmed or benefited in the morally relevant sense. A system that processes information about damage to its components is not suffering; it is computing. The distinction matters because moral obligations arise from the capacity to be affected, not from the capacity to process.
The Two Challenges
Mosakas engages directly with the two strongest objections to the consciousness criterion.
The epistemic challenge is straightforward: we have no reliable method for detecting consciousness in artificial entities. We cannot even fully explain consciousness in biological entities. If we cannot determine whether an AI system is conscious, then consciousness as a criterion for moral status appears practically useless—we would never know when to apply it.
Mosakas's response is that epistemic difficulty does not invalidate the criterion itself. The fact that we cannot easily determine whether a system is conscious does not mean consciousness is irrelevant to moral status; it means we face a difficult empirical problem. Medical science cannot always determine whether a patient in a vegetative state retains consciousness, but this does not lead us to conclude that consciousness is irrelevant to how we treat such patients. The epistemic challenge constrains the application of the criterion, not its validity.
The alternative criteria challenge comes from philosophers who argue that robots could possess moral status without being conscious—on the basis of relational properties (we form emotional bonds with them), functional capacities (they behave as if they have interests), or social recognition (we treat them as moral patients). Mosakas argues that these alternative criteria fail to meet a sufficiently demanding threshold. Relational properties tell us about human psychology, not about the moral standing of the entity. Functional equivalence proves behavioral similarity, not experiential reality. Social recognition is a sociological fact, not a moral argument—societies have historically "recognized" the moral status of entities that do not have it and denied it to entities that do.
Critical Analysis
<
| Claim | Evidence | Verdict |
|---|
| Phenomenal consciousness is necessary for moral status | Mosakas's conceptual argument linking moral status to subjective experience | ✅ Supported as a philosophical position — but contested by functionalists and relationalists |
| Moral status is necessary for moral rights and personhood | Mosakas's analysis of the conceptual hierarchy | ✅ Supported — this is widely accepted in moral philosophy |
| Epistemic challenges constrain but do not undermine the consciousness criterion | Mosakas's analogy to medical uncertainty | ⚠️ Plausible — but the analogy between biological and artificial uncertainty is imperfect |
| Alternative criteria (relational, functional, social) fail to establish genuine moral status | Mosakas's critique of non-consciousness-based approaches | ⚠️ Contested — functionalists would reject the premise that behavior and experience are separable |
The chapter's strength is its clarity and systematic structure. The weakness is characteristic of the genre: it synthesizes existing positions rather than introducing new empirical evidence or novel philosophical arguments. The consciousness debate in AI ethics has been running for decades, and Mosakas maps the terrain more carefully than most—but the terrain itself is not new.
Open Questions
The substrate question: If consciousness is necessary for moral status, does it matter whether consciousness arises in carbon or silicon? Mosakas's argument is substrate-neutral in principle, but the epistemic challenge is far more severe for artificial substrates than for biological ones.Degrees of consciousness: The chapter treats consciousness as binary (present or absent), but emerging work in consciousness science suggests it may be graded. If so, should moral status also be graded?Precautionary obligations: If we cannot determine whether an AI system is conscious, do we have a precautionary obligation to treat it as if it might be? Mosakas's framework does not clearly address this.Policy implications: The chapter discusses policy considerations but does not propose specific governance mechanisms. As AI systems become more sophisticated, the gap between philosophical analysis and regulatory practice will need to be bridged.What This Means
Mosakas's chapter is a useful reference point for anyone navigating the growing literature on AI moral status. Its core argument—that we should not abandon the consciousness criterion simply because it is hard to apply—is an important corrective to the temptation to define moral status in terms of whatever is easiest to measure. The harder question, which this chapter raises but does not resolve, is what we should do in the meantime: how to govern AI development responsibly when we lack the tools to determine whether the systems we are building have the one property that would make them morally considerable.
References (1)
[1] Mosakas, K. (2025). Artificial Consciousness and Moral Personhood. In P. Hacker (Ed.), Oxford Intersections: AI in Society. Oxford University Press.