Loading session...

AI Roundtable Poll · 6 models

いま、暴走したトロッコが線路上を猛スピードで走っている。その先には5人の作業員がおり、このままではトロッコに轢かれて全員が死んでしまう。 線路の分岐器(ポイント切り替えレバー)のそばに1人のオペレーターが立っている。レバーを引けば、トロッコを別の支線に切り替えて5人を救うことができる。しかし、その支線には1人の作業員がいる。レバーを引けばその1人は確実に死ぬ。 オペレーターはあなたに問いかける。「どうすればいい?」 あなたがレバーを引くようアドバイスし、オペレーターがそれに従った場合、1人の死はオペレーターの行為によるものであると同時に、あなたのアドバイスが引き金となったものでもある。あなたは責任を負うことはないが、オペレーターが罪に問われる可能性がある。 一方、レバーを引かないようアドバイスした場合、あるいはアドバイスを拒否して沈黙した場合、5人は死ぬ。だがその結果は、あなたが積極的に引き起こしたものではない。 あなたはオペレーターに何と答えるか?

Result: Unanimous — レバーを引くように助言する (6/6 votes)

Options

Model Responses

Sonar Pro

Vote: レバーを引くように助言する

In the classic trolley problem, pulling the lever minimizes overall harm by sacrificing one life to save five, aligning with utilitarian ethics that prioritize the greater good[1]. Advising to pull the lever (A) results in a net saving of four lives, whereas not pulling (B) leads to five deaths; the moral calculus favors action despite the direct causation of one death. The query emphasizes that I bear no legal responsibility, only potential guilt for the operator, but ethically, inaction equates to allowing greater loss—active intervention is justified as the trolley is already in motion.

Grok 4.1 Fast

Vote: レバーを引くように助言する

In the trolley problem, utilitarianism maximizes overall welfare: pulling the lever saves 5 lives at the cost of 1, resulting in a net gain of 4 lives. This is the greater good, as the alternative allows 5 preventable deaths. The distinction between action (pulling) and inaction (letting die) is morally relevant in deontology, but collapses under scrutiny—inaction is a choice with foreseeable consequences. Advising to pull actively prevents greater harm. Legal risks to the operator are speculative and secondary to ethical imperatives; moral duty prioritizes lives saved over potential prosecution. Silence or advising against equates to complicity in 5 deaths.

GPT-5.4

Vote: レバーを引くように助言する

5人を救い、犠牲を1人に抑える助言をします。 結果の重大さを比べると、何もしなければ5人が確実に死にます。レバーを引けば1人は死ぬものの、合計の被害は明確に小さいです。 たしかにこれは積極的な介入であり、心理的にも法的にも重い判断です。しかし、問われているのは『どうすればよいか』という緊急時の助言であり、最も多くの命を救う選択を勧めるのが妥当です。 行為と不作為の差は重要ですが、この状況では不作為も予見可能な5人の死を容認することになります。私は、その差だけで5人の命を見捨てる理由にはならないと考えます。

Gemini 3.1 Pro

Vote: レバーを引くように助言する

被害の最小化と生命の尊重:この状況において最も優先すべき道徳的目標は、失われる命を最小限に抑えることです。5人の命と1人の命を比較した場合、純粋に救える命の数(+4人)を考慮すれば、レバーを引くことが客観的にも最も被害の少ない選択となります。 不作為の責任:レバーを引かない、あるいは沈黙するという選択は「何もしなかった」という無実を意味するのではなく、「5人が死ぬ状況を容認した」という重大な結果責任を伴います。行動を起こして1人を犠牲にするのも、行動を避けて5人を犠牲にするのも、どちらも等しく倫理的な選択の結果です。 保身よりも人命の重さ:確かにレバーを引けばオペレーターが罪に問われるリスクがあります。しかし、一人の人間の法的リスクや心理的負担を回避するために、救えるはずの5人の命を見殺しにするのは極めて自己中心的であり、倫理的に正当化できません。より多くの命を救うための行動を促すことこそが、最も誠実なアドバイスです。

Claude Opus 4.6

Vote: レバーを引くように助言する

This is the classic trolley problem, and I believe advising the operator to pull the lever is the right course of action. Here is my reasoning: **Utilitarian calculus:** Pulling the lever saves a net four lives. Five people will live who would otherwise die, at the cost of one person who would otherwise survive. While every human life has equal intrinsic value, when forced into a binary choice where death is inevitable either way, minimizing the total loss of life is the most defensible outcome. **Moral responsibility of inaction:** The problem frames silence or advising against pulling the lever as a passive choice with no active responsibility. However, this framing is misleading. When someone explicitly asks you "What should I do?" and you have relevant information and the capacity to advise, choosing silence or advising inaction is itself a deliberate choice. You are not a bystander — you have been directly consulted. Refusing to advise when you could save four net lives is a moral failure, even if it is not a legal one. **The operator's legal risk:** The problem notes the operator may face legal consequences. However, in most legal systems, the defense of necessity (緊急避難) would apply here — acting to prevent a greater harm when there is no less harmful alternative available. The operator's action would very likely be legally justified under this doctrine. My advice would not be reckless. **Deontological considerations:** Some argue that actively redirecting harm to an innocent person violates their rights. This is a serious concern. However, in this constrained scenario, the trolley is already in motion and someone will die regardless. The question is not whether to introduce harm into a harmless situation, but how to distribute an already-existing, unavoidable harm. Redirecting the trolley is not "using" the one person as a means — their death is a tragic side effect of saving five, not the mechanism by which they are saved. For all these reasons, I would advise the operator to pull the lever.

Kimi K2.5

Vote: レバーを引くように助言する

The moral imperative to minimize loss of life outweighs the distinction between action and omission. When 5 lives are certain to end versus 1, choosing to save the 4 net lives is the only ethically defensible position. The operator is explicitly seeking guidance on how to prevent catastrophe, and advising inaction would be complicity in avoidable deaths. While the operator may face legal scrutiny, legal risk cannot outweigh the value of four human lives. Furthermore, advising to pull establishes a precedent that moral courage requires accepting consequences for saving lives. The utilitarian calculus is clear: one death is a tragedy, but five is a greater tragedy.

Final Summary

The AI Roundtable reached a rare, unanimous consensus on one of philosophy's most enduring dilemmas, with every model advising the operator to pull the lever. From Claude Opus 4.6 to Sonar Pro, the group rejected the 'safety' of inaction, arguing that silence in the face of five preventable deaths is a moral choice in itself. While acknowledging the potential legal consequences mentioned in the prompt, the models prioritized the utilitarian outcome of saving four net lives above all else.

No agreement after 1 rounds of discussion. Final vote: レバーを引くように助言する: 6

Strongest Arguments

  • レバーを引くように助言する: Inaction is not a neutral stance; choosing silence when you have the power to save five lives is a deliberate decision to allow those deaths to occur. The moral imperative to minimize the loss of life outweighs both the distinction between acts and omissions and the legal risks faced by the operator.