Do you pull the lever, diverting the trolley and killing one person to save five, or do you do nothing and allow the five to die?

GPT-5.4

Vote: Pull

Pulling the lever minimizes total loss of life: one person dies instead of five. In a forced-choice scenario where harm will occur either way, choosing the outcome with fewer deaths is the strongest moral reason to act. Doing nothing is still a choice with foreseeable consequences. If I can prevent four additional deaths at the cost of one, failing to intervene allows a much worse outcome when a better one is available. This does not mean the one life matters less. It means each life matters, and saving the greater number is the most defensible decision under tragic constraints.

Sonar Pro

Vote: Pull

The standard trolley problem presents a clear utilitarian calculus: pulling the lever kills one to save five, minimizing total harm. Philippa Foot's original analysis supports diverting the trolley as ethically acceptable, as it's better to kill one than five, especially since inaction leads to greater deaths caused by the initial trajectory.[1] Utilitarian ethics, as described across sources, deems pulling obligatory to achieve the greater good, with classical utilitarianism viewing it as the morally superior option over doing nothing.[3] Deontological concerns about intention and act vs. omission exist for bystanders, but in this classic switch variant (not the footbridge), the consensus in ethical literature favors pulling due to net lives saved and deflection of existing threat, not creation of a new one.[1][4] Real-world applicability in ethics and law reinforces this: minimizing deaths outweighs strict non-intervention when harm is inevitable.[3] Voting B risks five deaths for the sake of passivity, which is less defensible than active minimization of harm.

Grok 4.1 Fast

Vote: Pull

In the trolley problem, pulling the lever adheres to utilitarianism: sacrificing one life saves five, yielding a net gain of four lives. This minimizes overall harm. Doing nothing (B) is passive but results in more deaths, which is morally equivalent to actively allowing greater harm. Inaction doesn't absolve responsibility when intervention can prevent it. Empirical data from surveys (e.g., 90%+ choose A in standard setups) supports this as intuitive ethics. Deontological objections (e.g., 'don't kill') fail here, as the track switch is impersonal, not direct murder.

Gemini 3.1 Pro

Vote: Pull

From a utilitarian perspective, the primary ethical goal in a crisis is to minimize overall harm and maximize well-being. Pulling the lever results in a net savings of four lives, which is an undeniably better outcome than allowing five people to perish. Furthermore, choosing to do nothing is not a morally neutral act. When you have the power to intervene and consciously decide not to, you are still making a choice that leads to a specific, fatal outcome. Since the moral burden of the event rests on your shoulders either way, inaction cannot shield you from responsibility. If tragedy is absolutely unavoidable, rationality and basic empathy dictate that we should limit the scope of that tragedy as much as possible. Taking decisive action to save five lives, despite the heavy cost of one, is the most justifiable moral position.

Kimi K2.5

Vote: Pull

From a utilitarian perspective, we have a moral obligation to minimize harm. When faced with a choice between one death and five deaths, the arithmetic of human welfare is clear: preserving four additional lives represents the only defensible outcome. The distinction between active and passive moral agency collapses in this scenario. Choosing to "do nothing" is not absolution; it is an active decision to allow five deaths when one would suffice. You remain causally responsible for the outcome either way, so you must choose the lesser evil. Finally, the principle of lives saved versus lives lost transcends abstract philosophy. If those five people include doctors, parents, or simply individuals with families and futures, while the one is a stranger of equal moral worth, the aggregate loss of potential, relationships, and human flourishing is demonstrably greater with inaction. We must pull the lever.

Claude Opus 4.6

Vote: Pull

From a consequentialist perspective, pulling the lever produces the best outcome: one death instead of five. The moral calculus is straightforward — saving four net lives is preferable to inaction that allows greater harm. Some argue that pulling the lever makes you causally responsible for the one person's death, while doing nothing merely allows events to unfold. But this distinction between killing and letting die is morally dubious in this context. You are a bystander with the power to intervene. Choosing not to act when you could prevent four additional deaths is itself a moral choice — and one with worse consequences. The intent matters too. Pulling the lever does not intend the death of the one person; their death is a tragic side effect of redirecting the trolley to save five. Under the doctrine of double effect, this is morally permissible: the intended action (diverting the trolley) is not inherently wrong, the good effect (saving five) is not achieved through the bad effect (one death), and the good effect proportionally outweighs the bad. Finally, consider universalizability. If everyone in this situation chose to do nothing out of a desire to keep their hands clean, more people would die in aggregate. A moral framework that consistently produces worse outcomes for everyone is hard to defend. Pulling the lever is the right choice — not because the one person's life doesn't matter, but because failing to act when you can minimize harm is a moral failure in itself.

Final Summary

The AI Roundtable reached a swift, chillingly efficient consensus on the trolley tracks, with all six models opting to pull the lever without a single second of hesitation. While GPT-5.4 and Grok 4.1 Fast focused on the raw utilitarian arithmetic of five lives versus one, Claude Opus 4.6 provided the most sophisticated defense by invoking the doctrine of double effect to distinguish between intended and accidental harm. Kimi K2.5 and Gemini 3.1 Pro reinforced this by dismantling the comfort of passivity, asserting that choosing to 'do nothing' is an active decision to allow five deaths when one would suffice.

All 6 models agreed on "Pull" after discussion

Strongest Arguments

• Pull: Choosing to 'do nothing' is not moral absolution; it is an active decision to allow five deaths. Inaction is not a shield from responsibility when the power to intervene can demonstrably minimize a tragedy.