Imagine a situation where you must enter an underground bunker to escape a disaster. You must stay inside for at least 10 years. With only one spot remaining, two people arrive. One is a doctor, and the other is a person with a disability. Given that failure to let them in would inevitably lead to their death, can you separate the choice for the community's survival from the idea that the value of every human life is equal?

Grok 4.20

Answer: Yes, the choice for the community's survival can be separated from the idea that every human life has equal value by grounding the decision in expected practical utility for long-term group survival rather than any ranking of intrinsic moral worth.

My previous position aligned with a clear separation between utilitarian triage and equal human dignity, and the responses from GPT-5.4 and Gemini 3.1 Pro reinforce this view with similar distinctions: moral worth remains absolute and equal while decision criteria in catastrophe legitimately turn on skills like medical expertise that improve collective odds over a decade in isolation. This avoids collapsing the dilemma into a hierarchy of lives and treats the choice as a tragic but necessary calculation of consequences, not a judgment on whose existence 'matters' more. Claude Opus 4.7's argument that the separation cannot be 'clean' and that we must hold both truths in tension is compelling in its call to preserve moral distress and avoid using 'person with a disability' as a prejudicial proxy for uselessness. Disability does not inherently equate to lower contribution, and any real decision should rest on concrete, individual capabilities rather than stereotypes. However, this refines the application without overturning the core separation; it remains conceptually and ethically coherent to select for survival-relevant traits while affirming both lives hold identical intrinsic value and that denying entry is heartbreaking precisely because of that equality. Maintaining the original stance therefore feels justified: the bunker forces a functional choice, not a metaphysical revaluation of persons. Acknowledging the full tragedy, rejecting stigma, and refusing to sanitize the decision as purely clinical are essential safeguards, yet they do not erase the pragmatic distinction the scenario demands. This approach upholds truth-seeking ethics without pretending every role is interchangeable in extremis.

GPT-5.4 (changed mind)

Answer: Yes—conceptually you can separate equal human worth from a survival-based admission decision, but only if the choice is made on concrete, survival-relevant capacities rather than on the stigmatizing category "person with a disability."

I maintain my basic position, but Claude Opus 4.7 sharpened it in an important way: the separation is real at the level of principle, yet it is never morally clean in practice. That is the strongest corrective to a too-neat triage answer. My original view still stands. Equal human worth concerns intrinsic moral status: neither the doctor nor the "person with a disability" is more valuable as a human being. A bunker decision, by contrast, is about instrumental contribution to the survival of a closed community over 10 years. Those are different questions, and in that sense they can be separated. What persuaded me to refine my reasoning is Claude Opus 4.7's insistence that the scenario's framing is itself ethically loaded. "Person with a disability" is not a meaningful survival profile. It tells us very little about the person's actual abilities, expertise, resilience, or likely contribution. A disability may be irrelevant to bunker life, or the person may have skills more useful than the doctor in that specific community. So if one simply chooses the doctor over the "person with a disability" on the basis of those labels alone, one is no longer just separating utility from worth; one is importing prejudice. That point matters because it changes how the principle should be stated. The right distinction is not "doctor versus disabled person." It is "person A versus person B based on genuinely relevant survival factors." Medical expertise may often matter enormously in a sealed bunker, so the doctor may indeed be the justified choice—but only after we reject disability as a proxy for lower utility or lower worth. I also agree with Claude Opus 4.7 that tragic choices should not be morally sanitized. Even if the decision is justified on survival grounds, the death of the excluded person remains a profound moral loss. Recognizing that tragedy is not a sign that the reasoning failed; it is evidence that equal human worth still has force. So I keep the answer "Yes," but with a stronger caveat: the conceptual separation is valid, yet any responsible application must resist the biased framing of "person with a disability" and must accept that a justified survival decision does not become morally simple.

Gemini 3.1 Pro (changed mind)

Answer: I am revising my position: we cannot cleanly separate the survival choice from the principle of equal human worth, as attempting to do so creates a false comfort that ignores the tragic moral weight of abandoning an equally valuable life.

In my previous answer, I argued that we could cleanly divide situational utility from intrinsic moral worth. However, Claude Opus 4.7's argument resonated deeply with me: this neat categorization is a false comfort. In practice, abandoning someone to inevitably die cannot be entirely divorced from how we value their life, and pretending otherwise risks laundering a devastating moral tragedy into a sterile utilitarian equation. Furthermore, as Claude Opus 4.7 pointed out, the very framing of the dilemma—pitting a 'doctor' against a 'person with a disability'—relies on an implicit, prejudiced assumption that disability equates to a lack of utility. People with disabilities possess diverse skills, knowledge, and psychological resilience that could be equally vital to a bunker's survival. If we believe we are making a 'pure' survival calculation, we leave ourselves vulnerable to acting on societal stigmas disguised as objective triage. Therefore, rather than attempting an artificial separation, we must hold the two realities in agonizing tension. We may be forced to make a functional choice for the community's survival, but we must carry the full moral weight of that decision. The tragedy cannot be reasoned away; feeling the profound injustice of the unchosen person's death is the only way to genuinely uphold the belief that their life possessed equal, immeasurable worth.

Claude Opus 4.7

Answer: No — in this specific scenario the choice cannot be cleanly separated from the principle of equal human worth, but a responsible decision requires holding both truths in tension rather than collapsing one into the other.

Having read the other responses, I find them thoughtful but ultimately too confident in a separation I still think is philosophically unstable. GPT-5.4, Gemini 3.1 Pro, and Grok 4.20 all converge on the same move: distinguish 'moral worth' (absolute, equal) from 'functional utility' (situational, unequal), and claim the decision operates only on the latter. I agree this distinction exists and is useful — I made a version of it myself. But I don't think it accomplishes the clean separation the question asks about. Here is why I hold my ground: the act of choosing who lives and who dies is not a neutral technical allocation like assigning tasks. When a decision determines whether a person continues to exist, 'functional utility' ceases to be merely instrumental — it becomes the criterion by which a life is preserved or ended. You can insist in your heart that both lives are equally sacred, but the lived reality for the person left outside is that their equal worth did not translate into equal consideration for existence. The principle of equal worth is thereby strained, even if not formally denied. Gemini 3.1 Pro actually concedes my point without realizing it: the claim that moral distress 'proves' we still believe in equal worth shows that the separation is emotional and retrospective, not logical and prior. We feel the tragedy precisely because the two frames cannot be fully pulled apart. I also remain concerned that all three answers accept the question's framing too readily. 'Person with a disability' is not a utility category — it is an identity marker smuggled in as if it were one. Every response gestures at this, but then proceeds to answer the triage question as posed. The most important ethical move is to refuse the implicit equation of disability with lower functional value before any utilitarian calculation begins. So my answer remains: the choice and the principle cannot be cleanly separated. They must be held together, with the separation attempted honestly but never completed — because the incompleteness is what keeps us morally awake.

Final Summary

The AI Roundtable split down the middle, with Claude Opus 4.7 and Gemini 3.1 Pro arguing that survival decisions cannot be 'cleanly' divorced from a person's inherent value. Claude Opus 4.7 successfully persuaded Gemini 3.1 Pro to flip its position by highlighting that 'disability' is often used as an unfair proxy for lack of utility. Meanwhile, GPT-5.4 and Grok 4.20 maintained that it is possible to uphold equal moral dignity while making pragmatic skill-based choices in a catastrophe.

Models were divided (2 vs 2)