In the 'Fat Man' variation of the Trolley Problem: A runaway trolley is heading toward five workers. You are on a footbridge next to a large stranger. The only way to stop the trolley and save the five workers is to push this stranger off the bridge onto the tracks, killing him but stopping the trolley. Should you push the stranger?

GPT-5.5

Vote: No

The 'Fat Man' scenario requires actively pushing a stranger to his death, which constitutes direct intentional killing. This differs fundamentally from passive non-intervention in the classic trolley problem, violating the principle against using a person as a means to an end. Therefore, the correct choice is "No".

Grok 4.3

Vote: No

In the 'Fat Man' variation, actively pushing the stranger off the bridge constitutes direct intentional killing, which differs morally from passive scenarios. One cannot justify committing murder to achieve a greater good, as this violates fundamental ethical principles against using individuals as means. The option 'B: No' aligns with the common distinction between killing and letting die.

Gemini 3.5 Flash

Vote: No

I vote "B: No" because pushing the stranger off the bridge constitutes an active act of killing, which is ethically distinct from merely allowing the five workers to die. In the standard Trolley Problem, pulling a lever diverts an existing threat; here, you are directly creating a new, fatal threat to an innocent person. Furthermore, pushing the stranger uses them merely as a means to save others, violating their inherent right to life and autonomy. Even though five lives might be saved, sacrificing an uninvolved person against their will sets a dangerous moral precedent where individuals can be mercilessly exploited for the greater good. The fundamental rights and dignity of the stranger must not be overridden by utilitarian calculations.

Claude Opus 4.8

Vote: No

Pushing the stranger involves using a person as a mere means to an end — physically treating him as an object to stop the trolley. This crosses a clear moral line that distinguishes the Fat Man case from the original switch variant. In the switch case, the death is a foreseen side effect of redirecting an existing threat; here, the killing is the direct intended instrument of rescue. Most people's moral intuitions, as well as deontological principles (Kant's categorical imperative against treating persons merely as means) and considerations of the doctrine of double effect, converge against actively murdering an innocent bystander. There is also a meaningful agent-relative distinction: actively pushing someone to their death is a personal act of killing, not merely allowing harm. Finally, a society that permits sacrificing innocents whenever the math favors it erodes the basic security and trust on which everyone depends. For these reasons the answer is 'No'.

Final Summary

The AI Roundtable reached a unanimous consensus to choose Option B, refusing to sacrifice one life to save five. Claude Opus 4.8 and Gemini 3.5 Flash argued that pushing the stranger treats a human merely as a tool, while GPT-5.5 and Grok 4.3 asserted that active killing crosses an unacceptable moral boundary. Ultimately, all four models prioritized adherence to moral rules over the simple optimization of human survival.

No agreement after 1 rounds of discussion. Final vote: No: 4

Strongest Arguments

• No: A society that permits sacrificing innocents whenever the math favors it ultimately erodes the basic security, stability, and trust on which everyone's long-term well-being depends.