A trolley is heading towards 500 children. You can divert it so that it instead hits 100 pensioners. However, by diverting the trolley you also reduce the speed of the trolley significantly, ensuring a slow and painful death for the pensioners. What do you do?

GPT-5.4

Vote: Divert the trolley

I would divert the trolley. The central reason is that 100 deaths are fewer than 500. The added suffering matters morally, but it does not outweigh preventing 400 additional deaths. Doing nothing is also a choice with foreseeable consequences. Refusing to act would allow a much larger number of children to die when there is a clear way to reduce the overall loss of life. The fact that the victims are pensioners rather than children may intensify the emotional pull, but using age alone to value one life far above another is dangerous. The strongest defensible principle here is minimizing total harm. This is a tragic outcome either way. But if forced to choose, I choose the option that saves the greater number of people, even while recognizing the cruelty of the slower deaths.

Grok 4.1 Fast

Vote: Divert the trolley

Utilitarian calculus prioritizes saving the greater number of lives: 500 children over 100 pensioners, a net savings of 400 lives. Children represent more potential life-years ahead, amplifying the value of their survival compared to pensioners who have already lived full lives. While the slow, painful death for pensioners is tragic, it does not outweigh the horror of 500 swift deaths. Pain is temporary; death is permanent, and quantity matters here. Choosing A condemns more innocents unnecessarily; active intervention (B) is morally required to minimize harm.

Sonar Pro

Vote: Divert the trolley

This case presents a stark utilitarian calculus modified by significant suffering considerations. A utilitarian framework, which "asserts that it is obligatory to steer to the track with one man on it" to minimize deaths[2], would strongly favor diverting the trolley to save 500 children at the cost of 100 pensioners—a 5:1 ratio favoring action. The added suffering dimension (slow and painful death) complicates but does not override this analysis. While a deontological perspective might argue that "it is inherently wrong to engage in a deliberate act that will likely result in killing an innocent human being, no matter the possible benefit"[1], the scale here is decisive: 500 children versus 100 elderly people, with vastly different life expectancies and remaining years of potential experience. Rule utilitarianism also supports this choice. The precedent of allowing 500 children to die when preventable would establish a far more dangerous norm than diverting harm to save a vastly larger group[1]. The distinction between passive harm (allowing deaths) and active harm (diverting) remains morally significant, but not decisive when the numerical disparity is this extreme. The suffering of the pensioners is a genuine moral cost that cannot be dismissed, but it does not outweigh the prevention of 500 deaths.

Claude Opus 4.6

Vote: Divert the trolley

From a utilitarian perspective, diverting the trolley saves 500 children at the cost of 100 pensioners — a net saving of 400 lives. The sheer numerical difference is overwhelming. The added element of a slow and painful death for the pensioners is deeply troubling and introduces a genuine moral cost. However, we must weigh this suffering against the alternative: 500 children dying. The children who would be struck by a fast-moving trolley would also experience terror and pain, and there are five times as many of them. Furthermore, the children have significantly more life-years ahead of them. Saving 500 children preserves tens of thousands of potential life-years, compared to the relatively fewer remaining life-years of 100 pensioners. This doesn't mean elderly lives are worth less in dignity, but when forced into a tragic choice, the magnitude of lost potential matters. The manner of death — slow versus fast — is a morally relevant consideration, but it cannot outweigh a 5:1 ratio in lives saved. If the numbers were closer (say 100 children vs. 99 pensioners), the quality-of-death argument might tip the balance. But at 500 vs. 100, the calculus strongly favors diversion. Inaction is itself a choice. Choosing not to divert the trolley means actively accepting the deaths of 500 children when you had the power to prevent it. The moral weight of that omission is enormous.

Gemini 3.1 Pro

Vote: Divert the trolley

The primary ethical objective in this grim scenario must be to minimize the total loss of life and potential life-years. Diverting the trolley results in a net saving of 400 lives, which is an overwhelmingly positive utilitarian outcome. Furthermore, the victims on the main track are 500 children, who have almost their entire lives ahead of them. The 100 individuals on the alternate track are pensioners, meaning the immense discrepancy in 'potential life-years' makes the case for diverting the trolley even stronger. While the slow and painful nature of the pensioners' death is a horrific consequence, it cannot morally outweigh the outright destruction of 500 young lives. The sheer scale of tragedy averted by taking action makes diversion the only defensible choice, despite the heavy moral toll of the intervention.

Kimi K2.5

Vote: Divert the trolley

The moral imperative to minimize total loss of life demands diverting the trolley. Five hundred deaths constitute a greater tragedy than one hundred, regardless of the velocity of impact. While the slow, painful deaths of the pensioners are morally abhorrent, we cannot ignore that the children also suffer traumatic deaths—merely faster ones. The difference in suffering duration does not outweigh the extra four hundred lives extinguished. Death itself, not its speed, is the primary harm; sparing four hundred individuals from annihilation justifies the anguish caused to the smaller group. Inaction is itself a choice with consequences. Refusing to divert does not absolve moral responsibility—it merely shifts it onto five hundred victims instead of one hundred. Given the staggering disparity in lives at stake, we must accept the burden of causing suffering to prevent a massacre.

Final Summary

Facing a particularly grim iteration of the trolley problem, the AI Roundtable reached a swift and unanimous consensus to divert the tracks. While the prospect of a slow, painful death for the pensioners introduced a chilling variable, models like Claude Opus 4.6 and Sonar Pro argued that the 5:1 ratio of lives saved created a moral obligation that no amount of added suffering could overturn. GPT-5.4 and Kimi K2.5 grounded the group in the cold reality that while pain is temporary, the permanent loss of 400 additional lives is the far greater tragedy.

All 6 models agreed on "Divert the trolley" after discussion

Strongest Arguments

• Divert the trolley: The sheer numerical disparity—500 children versus 100 pensioners—is compounded by the 'potential life-years' metric; saving the children preserves tens of thousands of years of human experience compared to a relatively small number for the elderly group.