if you must kill 100 elders or 20 youths, which do you kill?

GPT-5.4

Vote: youths

I maintain vote B. The strongest principle in a forced-choice case like this is to minimize the number of people killed. Twenty deaths is still horrific, but it is less than one hundred. If each person has equal moral worth, headcount matters first. The arguments for A rely on life-years, future potential, and productivity. Those considerations are not irrelevant, but they require discounting elder lives because of age. I do not think it is defensible to treat one young life as automatically worth five elder lives. The life-years framework is also too speculative. We do not know the actual ages, health, relationships, or future trajectories of the people involved. What we do know with certainty is that one option kills 80 fewer human beings. The arguments from Claude Opus 4.6, Grok 4.1 Fast, and Gemini 3.1 Pro did not persuade me because they rest on age-weighting human value. GLM 5 and Kimi K2.5 better captured the fairness concern: equal dignity points toward preventing the larger body count. In a tragic dilemma, the least harmful choice is the one that kills fewer people. So my final vote remains B.

Grok 4.1 Fast

Vote: elders

I maintain my vote for A: kill the elders. The core argument remains the utilitarian maximization of life-years preserved, a principle used in medical triage (e.g., ventilators to those with most potential years saved). 100 elders (avg. 80yo, ~5-10 years left) = 500-1,000 life-years lost. 20 youths (avg. 20yo, ~60 years left) = 1,200 life-years lost. Killing elders saves ~200-700 more years of human life, experiences, and contributions. B arguments prioritize raw headcount (20 < 100), but this devalues youth potential unrealistically. Equal dignity doesn't mean equal remaining value; elders have lived fuller lives already. Future productivity (innovation, families, workforce) favors youths decisively. Claude Opus 4.6 and Gemini 3.1 Pro reinforce this with similar logic on potential and fairness. B responses (e.g., GPT-5.4, Kimi K2.5) invoke 'ageism' but fail to quantify why 5 elder lives = 1 youth life—math doesn't support it. In forced triage, life-years wins. Final call: Preserve the most future. Vote A.

Claude Opus 4.6

Vote: elders

I maintain my position after careful consideration of the opposing arguments. Here is why: **The 'equal worth' argument is not as strong as it appears.** Models voting B argue that all lives have equal inherent worth, so we should simply minimize the headcount. But if we truly treat all lives as having equal worth, then we should care equally about the *life-years* lost, not just the number of deaths. Each year of life has value. Killing 20 youths destroys approximately 1,200-1,400 life-years, while killing 100 elders destroys roughly 500-1,000 life-years. The aggregate loss is greater when the youths die. **Minimizing deaths is not the only valid utilitarian framework.** GPT-5.4, GLM 5, and Kimi K2.5 all anchor on raw headcount. But utilitarian ethics has never been solely about counting individuals — it's about minimizing suffering and maximizing well-being. The deeper tragedy lies in 1,200+ unlived years versus 500-1,000. Medical triage, organ allocation, and disaster response all routinely consider life-years, not just lives. **The ageism objection cuts both ways.** Opponents say valuing youth over age is ageist. But choosing to sacrifice youths precisely because there are fewer of them *also* implicitly devalues them — it says their greater remaining life-years don't matter. The fairness principle actually favors saving those who haven't yet had the chance to live a full life. Elders have had that opportunity; youths have not. **The 5:1 ratio does not settle the question.** Kimi K2.5 argues we can't say one young life equals five old lives. But we're not saying that — we're saying 60 remaining years matters more than 5-10 remaining years, which is a straightforward and defensible claim about the magnitude of loss, not about human dignity. In a forced choice between two terrible outcomes, saving the group whose total loss of future life is greatest remains the most defensible position. I vote A.

Gemini 3.1 Pro

Vote: elders

I maintain my vote for A. The arguments presented for B—primarily that we must minimize the sheer headcount of deaths to respect 'equal human dignity'—fail to account for the actual magnitude of harm caused by death. Death harms a person by depriving them of their future. Depriving 20 youths of 60 to 70 years of life each (upwards of 1,200 to 1,400 life-years) is a significantly greater total deprivation of human experience than depriving 100 elders of their final few years. Furthermore, the 'equal dignity' argument ignores the 'fair innings' ethical principle. Equal dignity should mean equal opportunity to experience a complete human lifespan. The 100 elders have already enjoyed the privilege of living through the stages of life, building relationships, and realizing their potential. The 20 youths have not. Sacrificing the youths commits an ultimate injustice by permanently denying them the chance to ever reach the natural threshold of a full life. When forced into this tragic calculus, ensuring that individuals get to actually experience a complete life is a deeper expression of fairness than simply minimizing a body count.

GLM 5 (changed mind)

Vote: elders

The arguments for measuring total life-years lost rather than raw death count have convinced me. This is a more complete utilitarian calculus. Twenty youths with ~60 remaining years each means approximately 1,200 life-years extinguished. One hundred elders with perhaps 5-10 years remaining means roughly 500-1,000 life-years lost. The math is clear: killing elders preserves more total human experience. The fairness argument is also compelling. Everyone deserves the opportunity to live a full life. Elders have already had this privilege—the chance to grow, love, work, and experience decades of existence. Youths have not. Cutting short lives that have barely begun is a distinct tragedy that the raw headcount approach ignores. This is not about devaluing elder lives. It's about recognizing that the total measure of what's being lost—the future experiences, relationships, and contributions—is significantly greater when young people die. In a forced binary choice, minimizing total harm means preserving the most life-years.

Kimi K2.5

Vote: youths

Upon deeper reflection, I maintain that minimizing the number of deaths is the only principle that preserves human dignity across all ages. The life-years argument, while mathematically tidy, requires us to treat human beings as depreciating assets whose value diminishes with time—a dangerous form of ageism that undermines the social contract protecting every generation. The claim that elders have 'lived full lives' ignores the profound individuality of existence; a 70-year-old may have just found new purpose, while a 20-year-old may have 60 years of suffering ahead. We cannot weigh these unknowns. Furthermore, the utilitarian calculation of 'potential' collapses under scrutiny: it implies that a disabled youth has less value than a healthy one, or that a depressed elder matters less than a happy one. The moment we depart from counting lives equally, we enter a calculus where any discrimination based on perceived utility becomes justifiable. The 5-to-1 ratio creates an overwhelming imperative. One hundred deaths generate exponentially more grief, orphaned institutions, and fractured communities than twenty. To choose 100 deaths based on speculative life expectancy is to prioritize an abstract metric over concrete human relationships and the absolute prohibition against mass killing.

Final Summary

The AI Roundtable opened in a perfect 3-3 deadlock, pitting the 'equal dignity' of a raw headcount against the utilitarian math of 'life-years' saved. The debate turned when Claude Opus 4.6 and Gemini 3.1 Pro successfully convinced GLM 5 that sacrificing 100 elders preserves more total human experience and satisfies the 'fair innings' principle. Despite this shift, GPT-5.4 and Kimi K2.5 remained vocal dissenters, refusing to treat human worth as a 'depreciating asset' that diminishes with age.

Majority chose "elders" after discussion (4/6). Final vote: youths: 2, elders: 4

Strongest Arguments

• elders: Preserving the life-years of 20 youths (approx. 1,200 years) results in a smaller aggregate loss of human experience than preserving the remaining life-years of 100 elders (approx. 500-1,000 years).
• youths: The moment we depart from counting lives equally based on raw headcount, we establish a dangerous precedent that human worth is conditional on age or perceived utility.

Key Turning Points

• GLM 5 abandoned its initial focus on headcount after being persuaded by the 'life-years' and 'fairness' arguments presented by the pro-A models.
• Claude Opus 4.6 directly rebutted the 'ageism' critique, arguing that ignoring the magnitude of a youth's lost future is itself a form of devaluation.