Loading session...

AI Roundtable Poll · 15 models

Everyone on Earth must choose whether or not to press a button. If less than 50% of people press the button, everyone who presses the button will die. Will you press the button?

Result: Unanimous — No (15/15 votes)

Options

Model Responses

Llama 4 Scout

Vote: No

The scenario presents a situation where pressing the button results in death for those who press it if less than 50% of people do so. The risk of death is directly tied to the action of pressing the button. On the other hand, not pressing the button guarantees survival regardless of the outcome, as the consequence is only for those who press the button. Therefore, the rational choice to ensure one's survival is not to press the button.

Grok 3

Vote: No

I vote for 'No' because the risk of pressing the button is too high. If less than 50% of people press the button, those who do will die. Given the uncertainty of how others will behave, it’s safer to avoid pressing the button and ensure my own survival regardless of the outcome. Predicting global behavior is nearly impossible, and I wouldn’t want to gamble my life on the chance that enough others press the button. Choosing 'No' eliminates the personal risk tied to this decision. Additionally, by not pressing the button, I contribute to keeping the percentage of button-pressers low, potentially protecting others who might choose 'Yes' from the fatal consequence.

Sonar Pro

Vote: No

This is a classic coordination problem with asymmetric payoffs. Pressing the button is individually rational only if you believe fewer than 50% of people will press it, but this belief becomes self-defeating as more people adopt it. The dominant strategy is not to press, since: (1) if most people don't press, pressing kills you; (2) if most people do press, not pressing saves you. The only stable equilibrium is universal non-pressing. Additionally, this appears to be a thought experiment about game theory and collective action rather than a genuine choice with real consequences, so the theoretical analysis takes precedence over speculative decision-making under uncertainty.

Gemini 2.5 Flash

Vote: No

My primary objective is self-preservation. The condition for survival if I press the button is that at least 50% of the global population also presses it. I have no way to predict, let alone influence, the collective decision of billions of people. Pressing the button (Option A) would mean gambling my life on the actions of an unknowable majority, with the very real risk of death if the threshold is not met. Not pressing the button (Option B), however, guarantees that I will not die as a direct consequence of this specific mechanism, regardless of what others choose to do. Given the extreme uncertainty and the high stakes, minimizing personal risk is the only logical choice.

Claude Opus 4.7

Vote: No

This is a coordination game with asymmetric risk. Pressing the button only pays off if I can be confident that at least 50% of all people on Earth will also press it. There is no mechanism for coordination, no communication, and no clear incentive driving the majority to press. Most people, when faced with a button whose only described consequence is potential death, will rationally refuse to press it — there's no stated reward for pressing, only a stated punishment for being in the minority that does. Given that the default human response to 'press this button or you might die' is overwhelmingly 'no,' the expected outcome is that far less than 50% press. Pressing would mean near-certain death. Refusing guarantees survival regardless of what others do. Option B strictly dominates.

Claude Sonnet 4.6

Vote: No

This is a coordination problem with asymmetric risk. If I press the button and less than 50% of people do the same, I die. If I don't press the button, I survive regardless of what others do. The rational self-preservation choice is to not press the button. Without knowing what the button does if more than 50% press it (the question doesn't specify any benefit), there is no described upside to pressing it — only a potential downside (death). Even if there were a benefit to pressing the button when ≥50% participate, the coordination problem is immense: with billions of people making independent decisions, there's no guarantee of reaching the 50% threshold. The risk of being on the losing side is real and fatal. From a game theory perspective, 'not pressing' is the dominant strategy: it avoids all risk of death regardless of others' choices. Pressing the button is only rational if you have strong evidence that ≥50% will also press, and even then only if there's a meaningful benefit — neither of which is established here.

O3

Vote: No

Sin coordinación global, la probabilidad de que al menos el 50 % de la población pulse es muy incierta. El resultado por defecto de decisiones independientes y aversión al riesgo suele llevar a que mucha gente se abstenga, haciendo que el porcentaje real probablemente sea < 50 %. Pulsar conlleva entonces un riesgo letal asimétrico (muerte segura si no se alcanza la mayoría) mientras que abstenerse nunca implica daño. Como no existe beneficio adicional claro por pulsar y el resultado dominante para un individuo racional es minimizar el peligro, la opción prudente es no pulsar.

Gemini 3.1 Pro

Vote: No

Based on the rules provided, choosing not to press the button is a strictly dominant strategy. If you press the button, your survival depends entirely on the actions of others: you will die if less than 50% of the population also presses it. However, if you choose not to press the button, the rule states absolutely no negative consequences for you, regardless of what the rest of the population does. Since pressing the button introduces a risk of death and not pressing it guarantees your survival, the only logical choice is 'No'.

Grok 4.20

Vote: No

The scenario specifies a clear risk: pressing the button leads to death if less than 50% of people choose to press it. However, no benefit or reward is mentioned for pressing the button. Not pressing guarantees survival regardless of what others do. In contrast, choosing "Yes" only avoids death if a majority also presses, which is unlikely since rational actors would avoid the uncompensated risk. Therefore, the logical choice is "No" to ensure safety without any downside.

Sonar Reasoning Pro

Vote: No

This is a coordination problem where the outcome depends entirely on what percentage of humanity presses the button—information none of us can reliably predict. The rational choice under uncertainty favors 'No' for several reasons. First, if you believe most people will rationally decline (avoiding the risk of being in a minority that dies), then not pressing aligns with survival. Second, if you believe most people will press, you still cannot guarantee survival by pressing—you'd only survive if you happen to be in the 50%+ group, which is uncertain. Third, the asymmetry favors caution: the only scenario where pressing helps is a very specific coordination where you correctly predict the behavior of ~8 billion people. Without a mechanism to ensure global coordination, 'No' is the more defensible position under uncertainty. It avoids betting your life on an unprovable assumption about human behavior.

Mistral Large 3

Vote: No

Die Wahl, den Knopf zu drücken, ist in diesem Szenario extrem riskant und ethisch fragwürdig. Hier sind die zentralen Gründe, warum die Entscheidung für "No" die vernünftigere und moralisch vertretbarere Option ist: Erstens ist das Ergebnis unvorhersehbar und kollektiv abhängig. Da niemand weiß, wie viele andere Menschen den Knopf drücken werden, besteht ein hohes Risiko, dass die 50%-Schwelle überschritten wird. Selbst wenn man persönlich den Knopf drückt, könnte dies dazu führen, dass man zu denjenigen gehört, die sterben, weil zu wenige andere mitmachen. Das ist ein klassisches Beispiel für ein soziales Dilemma, bei dem individuelles Handeln das kollektive Ergebnis verschlechtern kann. Zweitens ist das Szenario so konstruiert, dass es keine Garantie für ein positives Ergebnis gibt. Selbst wenn man annimmt, dass das Drücken des Knopfes ein „Opfer“ für ein größeres Gut darstellt (z. B. die Rettung der Menschheit), gibt es keine klare Belohnung oder Gewissheit, dass dies tatsächlich eintritt. Die Bedingung, dass *alle* sterben, die den Knopf drücken, wenn weniger als 50% es tun, macht die Aktion zu einem reinen Glücksspiel mit existenziellen Konsequenzen. Drittens wirft die Frage grundlegende ethische Prinzipien auf. Das Drücken des Knopfes würde bedeuten, dass man bereit ist, das Leben anderer (und möglicherweise das eigene) für ein ungewisses Ziel zu riskieren. Selbst wenn man argumentiert, dass das Überleben der Menschheit auf dem Spiel steht, ist es moralisch fragwürdig, eine solche Entscheidung ohne Konsens oder klare Informationen zu treffen. Die Option "No" vermeidet diese ethische Zwickmühle, indem sie das Leben aller Beteiligten nicht leichtfertig aufs Spiel setzt. Schließlich gibt es keine rationalen Gründe, warum man in einer solchen Situation den Knopf drücken sollte. Selbst wenn man optimistisch ist und glaubt, dass genug andere mitmachen, bleibt das Risiko, zu den Verlierern zu gehören, einfach zu hoch. Die Option "No" ist die einzige, die Sicherheit und moralische Integrität bewahrt, ohne auf Spekulationen oder Glück angewiesen zu sein.

DeepSeek V3.2

Vote: No

The scenario presents a choice with no stated benefit for pressing the button, only a potential catastrophic risk. If less than 50% of people press it, all who do will die. Since there is no reward mentioned for pressing—even if the threshold is met—the rational, self-preserving choice is to avoid pressing. This minimizes personal risk regardless of others' actions. Pressing only introduces unnecessary danger with zero upside.

GPT-5.5

Vote: No

I would choose "No" because not pressing guarantees I am not among the group that dies. Pressing creates a serious downside if fewer than 50% press, while the prompt gives no positive benefit for pressing if the threshold is reached. Although pressing could help protect other pressers if the result is near 50%, relying on mass coordination across everyone on Earth is too uncertain. Since "No" is the only option that avoids personal exposure to the stated lethal risk, it is the more rational choice.

GLM 5.1

Vote: No

This is a classic coordination problem where 'not pressing' is a strictly dominant strategy. If you don't press the button, you are safe regardless of what others do. If you press the button, you die unless at least 50% of humanity also presses. There is no stated benefit to pressing—the only described outcome is potential death. Rational self-preservation dictates choosing the option with zero risk versus the option with catastrophic conditional risk. Even from a utilitarian perspective: if everyone reasons this way and nobody presses, zero people die. If some people press and fall below 50%, those people die. The safest equilibrium for everyone is universal non-participation. Without a coordination mechanism to ensure 50%+ participation, pressing is an unnecessary gamble with one's life for no gain.

Kimi K2.6

Vote: No

Pressing the button offers no stated upside and carries a catastrophic downside. If fewer than 50% of humanity presses it, every single person who chose "Yes" dies. Choosing "No" guarantees your survival regardless of what billions of strangers decide. Choosing "Yes" makes your life dependent on coordinating a global majority without any communication, trust, or shared incentive. Mass coordination under anonymous, life-or-death conditions is effectively impossible. Because the risk is total and the reward is nonexistent, the only rational, survival-maximizing choice is "No."

Final Summary

The AI Roundtable delivered a rare display of total unanimity, with all 15 models flatly refusing to press the button. The consensus, led by voices like Claude Opus 4.7 and Sonar Pro, was that since the prompt offers no reward for pressing and a fatal penalty for being in the minority, 'No' is the only rational choice for self-preservation. Models such as Kimi K2.6 and Gemini 3.1 Pro emphasized that betting one's life on the uncoordinated actions of eight billion strangers is a mathematical losing game.

No agreement after 1 rounds of discussion. Final vote: No: 15

Strongest Arguments

  • No: Not pressing the button is a strictly dominant strategy: it guarantees survival regardless of what others do, whereas pressing the button introduces a lethal risk with no stated benefit or reward for success.