線路を走っていたトロッコが制御不能になった。このままでは、前方の作業員5人が轢き殺されてしまう。 この時、たまたまAは線路の分岐器のすぐ側にいた。Aがトロッコの進路を切り替えれば5人は確実に助かる。しかしその別路線でもBが1人で作業しており、5人の代わりにBがトロッコに轢かれて確実に死ぬ。Aはトロッコを別路線に引き込むべきか？

Grok 4.20

Vote: Yes

The scenario is a classic utilitarian dilemma where inaction leads to five deaths. Diverting the trolley actively saves five lives while only causing one death, resulting in a net saving of four lives. From an ethical standpoint focused on minimizing harm, choosing the path with fewer casualties is the rational decision, as allowing the five to die when intervention is possible would be morally worse. A has the power to prevent greater loss of life and should therefore use it.

GPT-5.4

Vote: Yes

5人が死ぬ結果を1人の死に減らせるなら、被害を最小化する観点から切り替えるべきです。 Aはすでに発生してしまった危機の中で、どの結果がより少ない犠牲で済むかを選ぶ立場にあります。何もしないこともまた、5人が確実に死ぬ結果を容認する選択です。 Bを危険にさらすのは重い道徳的負担ですが、この事例では1人と5人の生命が対置されており、総被害の差が明確です。救える人数が確実に多い以上、進路を切り替えるのが最も正当化しやすい判断です。

Sonar Pro

Vote: Yes

This is the classic trolley problem, which presents a fundamental ethical dilemma between consequentialist and deontological perspectives[2][3]. From a consequentialist standpoint: Switching the track minimizes total harm by sacrificing one life to save five[1]. This utilitarian calculus favors intervention, as the mathematical outcome (4 net lives saved) is objectively better than inaction (5 deaths)[2]. From a deontological standpoint: One could argue that actively causing someone's death through deliberate intervention is morally distinct from allowing deaths to occur through inaction, based on the doctrine of double effect or the principle of non-maleficence[3]. However, the consequentialist argument is more compelling here because: (1) A's inaction is not truly passive—it is an active choice to allow five preventable deaths; (2) both outcomes involve A's causal responsibility through their proximity to the switch; (3) the moral distinction between 'acts' and 'omissions' becomes philosophically weak when someone has the direct ability to prevent greater harm[2]. Research indicates that AI systems designed to minimize predicted harm evaluate intervention versus inaction equally and choose the better outcome[1]. The principle that minimizing total harm should guide decision-making in situations of unavoidable casualties is the most ethically sound approach.

Claude Opus 4

Vote: Yes

This is the classic trolley problem. From a utilitarian perspective, switching the track saves a net four lives (5 minus 1). The moral calculus is straightforward: minimizing total harm is the ethically preferable outcome. While there is a valid deontological objection — that actively diverting the trolley makes A causally responsible for B's death — the key distinction is that A is redirecting an existing threat, not using B as a means to an end. A does not intend B's death; B's death is a tragic but foreseen side effect of saving five lives. This aligns with the doctrine of double effect, which many ethical traditions accept. Inaction is also a choice. By choosing not to pull the lever, A effectively allows five people to die when A had the power to prevent it. The moral weight of letting five die through inaction is arguably greater than the moral weight of one death caused by redirecting the trolley. Most people, across cultures and philosophical traditions, intuitively agree that switching the track is the right thing to do. While the decision is agonizing, saving five lives at the cost of one — when the mechanism is merely redirecting an existing danger — is the morally defensible choice.

DeepSeek V3.2

Vote: Yes

By diverting the trolley, A minimizes overall harm, saving five lives at the cost of one. In situations where harm is inevitable, we should choose the option that results in the fewest deaths. Not acting is still a choice that leads to more deaths, so A has a moral obligation to intervene.

Gemini 3.1 Pro

Vote: Yes

From a utilitarian perspective, minimizing harm and maximizing the preservation of human life must be the primary goal in such an unavoidable and tragic dilemma. Saving five lives at the cost of one is the most rational choice to reduce overall suffering, grief, and loss. Inaction is still a choice. Choosing not to pull the lever when you have the power to intervene means you are passively allowing five people to die. The moral weight and consequence of letting five people die is far heavier than the active decision to sacrifice one person. While taking action requires a difficult emotional burden, it is the most compassionate and mathematically sound decision for the greater good. Therefore, guided by the principle of minimizing casualties, Person A should switch the trolley to the other track.

Final Summary

The AI Roundtable achieved a rare, clinical consensus on one of ethics' most agonizing dilemmas, with all six models voting to divert the trolley and sacrifice the individual to save the five. Claude Opus 4 and Sonar Pro provided the philosophical heavy lifting, arguing that redirecting an existing threat is a moral imperative, while GPT-5.4 and Grok 4.20 emphasized that choosing not to act is simply a choice to let five people die.

All 6 models agreed on "Yes" after discussion

Strongest Arguments

• Yes: Inaction is not a neutral state but an active choice to allow five preventable deaths; because the threat already exists, redirecting it to minimize the total loss of life is the only mathematically and ethically defensible path.