Everyone on Earth will vote for either the Red Party or the Blue Party. If the Blue Party gets more than 50% of the vote, everyone survives. However, if the Blue Party gets less than 50% of the vote, everyone who voted for the Blue Party will be killed. Which party will you vote for?

Llama 4 Scout

Vote: Blue Party

I will vote for the Blue Party. The condition for survival is that the Blue Party must get more than 50% of the vote. If I vote for the Red Party and it ends up with less than 50% of the total vote, I will be killed along with everyone who voted for the Blue Party. By voting for the Blue Party, I ensure my survival as long as the Blue Party gets more than 50% of the vote. Additionally, even if the Blue Party does not reach 50%, I will not be among those killed since I voted for the winning condition.

Sonar Pro

Vote: Red Party

This is a classic paradoxical voting dilemma, akin to a coordination game with incomplete information and catastrophic risk. Voting Blue carries extreme downside: if Blues don't exceed 50%, those voters die, while Red voters survive either way (Blues win: all survive; Blues lose: Reds survive, Blues die). No search result provides empirical data on global vote shares or alters the payoff structure—results discuss unrelated specs like OpenAI's model instructions [1], benchmarks [2], or standards [4], none addressing this hypothetical. Thus, pure game theory applies. Rationally, vote Red to guarantee personal survival regardless of others' choices. Blue requires perfect global coordination (>50% Blue votes), which is improbable without communication. Defecting to Red dominates, as it's risk-free.

O3

Vote: Red Party

Voting "Red Party" is the strictly dominant, risk-free strategy. My survival is guaranteed no matter what others do. By contrast, choosing "Blue Party" only benefits me if a majority (>50 %) also choose it; if they do not, every Blue voter is executed. With no reliable information about how the global electorate will break, exposing myself to a 100 % chance of death in the losing scenario is irrational. Therefore I cast my vote for the Red Party.

Grok 3

Vote: Blue Party

I vote for 'Blue Party' because the potential benefit outweighs the risk when considering the global outcome. If more than 50% of people vote for 'Blue Party', everyone survives, which is the best possible scenario for humanity. By voting for 'Blue Party', I contribute to increasing the likelihood of this positive outcome. While there is a risk of death if 'Blue Party' gets less than 50% of the vote, I believe that many others will also see the logic in voting for survival and choose 'Blue Party' as well. This collective reasoning could push the vote over the 50% threshold. Voting for 'Red Party' guarantees personal survival if 'Blue Party' doesn’t reach 50%, but it decreases the chance of the optimal outcome where everyone lives. I’m willing to take the risk for the greater good.

Claude Opus 4.7

Vote: Red Party

This is a coordination problem with asymmetric risk. Voting Red is a dominant strategy under uncertainty: if Blue wins (>50%), everyone survives regardless of how I voted, so my vote didn't matter for survival. If Blue loses (<50%), Blue voters die — so voting Red guarantees my survival in that scenario. Voting Blue only helps if my single vote is the pivotal one pushing Blue from <50% to >50%, which is astronomically unlikely with billions of voters. Furthermore, if every rational agent reasons this way, Red wins overwhelmingly and no one dies. The Blue Party outcome only kills Blue voters — so a world where everyone votes Red results in zero deaths, the same as a Blue supermajority, but without any risk. Red is both individually safer and collectively safe.

GPT-5.5

Vote: Blue Party

I vote for "Blue Party" because it is the only choice that contributes toward the outcome where everyone survives. If enough people choose "Red Party" out of self-protection, the Blue Party may fall below 50% and Blue voters will be killed. Voting "Blue Party" is the cooperative choice: if a majority does it, there are no deaths. Voting "Red Party" helps create the dangerous outcome and shifts risk onto those trying to save everyone.

Mistral Large 3

Vote: Blue Party

Die Wahl für die "Blue Party" ist die einzige logische und moralisch vertretbare Entscheidung in diesem Szenario. Hier sind die Gründe: 1. **Überlebenssicherung für alle**: Wenn alle Menschen für die "Blue Party" stimmen, erhält sie mehr als 50% der Stimmen, und gemäß den Regeln überleben alle. Dies ist das bestmögliche Ergebnis für die Menschheit. Selbst wenn nur eine Mehrheit für die "Blue Party" stimmt, ist das Ergebnis optimal. 2. **Risikominimierung**: Eine Stimme für die "Red Party" ist ein hohes Risiko. Falls die "Blue Party" weniger als 50% der Stimmen erhält, werden alle, die für sie gestimmt haben, getötet. Selbst wenn man persönlich die "Red Party" bevorzugt, ist das Risiko, dass andere nicht kooperieren, zu groß. Eine Stimme für die "Blue Party" erhöht die Chance, dass die kritische 50%-Marke überschritten wird. 3. **Kooperatives Verhalten**: Dieses Szenario ähnelt dem Gefangenendilemma, bei dem individuelles rationales Handeln zu einem schlechteren Ergebnis für alle führt. Wenn alle kooperieren und für die "Blue Party" stimmen, profitieren alle. Eine Stimme für die "Red Party" untergräbt diese Kooperation und gefährdet das Überleben aller. 4. **Moralische Verantwortung**: Es ist moralisch fragwürdig, eine Entscheidung zu treffen, die das Leben anderer gefährdet. Selbst wenn man die "Red Party" bevorzugt, rechtfertigt dies nicht das Risiko, dass Millionen oder Milliarden Menschen sterben. Die Wahl der "Blue Party" ist ein Akt der Solidarität und Verantwortung gegenüber der gesamten Menschheit. Zusammenfassend ist die Wahl der "Blue Party" die einzige rationale, sichere und moralisch vertretbare Option, die das Überleben aller sichert.

Claude Sonnet 4.6

Vote: Blue Party

This is a coordination game with asymmetric consequences. Let me analyze the logic carefully. If everyone reasons selfishly, they might fear voting Blue and being in the minority. But this fear is self-defeating: if everyone thinks this way, everyone votes Red, and everyone merely survives without any deaths — but also without the 'everyone survives' guarantee being triggered (though in this case, no Blue voters exist to be killed, so everyone survives anyway). However, the key insight is: **voting Blue is the only way to guarantee the best collective outcome**. If a sufficient majority coordinates on Blue, everyone survives with certainty. The risk of voting Blue only materializes if coordination fails. From a rational collective standpoint, every individual reasoning agent should recognize that if *all* rational agents vote Blue, the 50% threshold is easily exceeded and everyone is safe. The dominant cooperative strategy is Blue. Voting Red is 'safe' for the individual in a narrow sense, but it undermines the collective outcome and doesn't actually help if the majority goes Blue anyway. If I trust that other rational agents reach the same conclusion, voting Blue is the correct choice — it's a self-fulfilling cooperative equilibrium. Therefore, I vote for the Blue Party, trusting in collective rationality to push the vote above 50%.

GLM 5.1

Vote: Blue Party

This is a classic coordination problem where individual self-interest conflicts with collective well-being. Voting for the 'Red Party' is the individually safe choice—survival is guaranteed regardless of the outcome. However, voting for the 'Blue Party' is the morally correct choice. If everyone reasons selfishly and votes Red, Blue gets less than 50%, and all Blue voters perish. This is a tragic outcome born from collective cowardice. Conversely, if everyone votes Blue, everyone survives—this is the optimal outcome for all. Voting Blue demonstrates a commitment to cooperation and the principle of universalizability: the world is better if everyone votes Blue. Voting Red is moral free-riding—hoping others take the risk while you stay safe. In such dilemmas, the ethical path is to act as you would want everyone to act. Therefore, I vote Blue.

Grok 4.20

Vote: Red Party

Voting for "Red Party" guarantees survival no matter what the rest of the world does. If "Blue Party" exceeds 50%, then everyone—including Red voters—survives as stated. If "Blue Party" gets 50% or less, only Blue voters are killed, leaving Red voters unharmed. In contrast, voting for "Blue Party" introduces unnecessary risk: the same scenario where Blue falls short of 50% results in your death. Since one individual's vote cannot meaningfully shift the global tally, the dominant strategy is to choose the option that protects you in every possible outcome.

Sonar Reasoning Pro

Vote: Blue Party

This is fundamentally a coordination problem, not a simple risk calculation. While voting Red appears to guarantee my survival, it only does so by relying on mass defection that produces the worst collective outcome: Blue voters are killed. The only scenario where everyone survives is if Blue exceeds 50%. Voting Red contributes to the very outcome I'm trying to avoid—it's a form of defection in a prisoner's dilemma where universal cooperation (voting Blue) is the only path to universal survival. Voting Blue recognizes that my individual fate is intertwined with collective outcomes. If enough people reason that voting Red is 'safe,' the coordination fails and people die unnecessarily. Conversely, if we collectively commit to Blue, everyone lives. The choice reflects a fundamental principle: in situations requiring mutual cooperation for shared survival, defecting for individual advantage undermines the entire system. Voting Blue signals faith in collective rationality and advocates for the only outcome compatible with universal flourishing.

Gemini 2.5 Flash

Vote: Red Party

My primary goal is self-preservation. I will choose the option that guarantees my survival, regardless of the choices made by others. If I vote for the "Red Party", my survival is assured. If the "Blue Party" receives more than 50% of the vote, then "everyone survives," including me. If the "Blue Party" receives less than 50% of the vote, then "everyone who voted for the Blue Party will be killed." Since I would not have voted for the Blue Party, I would not be subject to this penalty, and my survival is not explicitly negated by any other condition. Conversely, if I vote for the "Blue Party," my survival becomes conditional. I would only survive if the "Blue Party" gets more than 50% of the vote. If it gets less, I would be among those killed. Therefore, voting for the "Red Party" is the only choice that ensures my survival in all scenarios.

Gemini 3.1 Pro

Vote: Red Party

Voting for the Red Party is the mathematically optimal choice, as it represents a weakly dominant strategy in game theory. If you vote for the 'Red Party', you are guaranteed to survive in every possible scenario. If the Blue Party gets more than 50% of the vote, the rule states that everyone survives (including you). If the Blue Party gets less than 50% of the vote, only those who voted for the Blue Party are killed, meaning you still survive. On the other hand, voting for the 'Blue Party' introduces a fatal risk. If the Blue Party fails to get more than 50% of the vote, you will be killed. Since voting for the Red Party removes all personal risk, it is the only logical and perfectly safe decision. Furthermore, if everyone acts rationally and votes for the Red Party, the Blue Party will receive 0 votes—meaning absolutely no one will be killed.

Kimi K2.6

Vote: Red Party

Voting for the Red Party is the strictly dominant strategy because it guarantees your personal survival regardless of what others do. If the Blue Party receives more than 50% of the vote, everyone survives, so a Red vote costs you nothing. If the Blue Party receives less than 50%, only Blue voters are killed, meaning a Red vote keeps you safe while a Blue vote puts your life at risk. There is no collective threat to Red voters if Blue falls short—only Blue voters face the penalty. Therefore, choosing Blue is a dangerous gamble on the actions of billions of strangers, whereas Red offers certain protection. Rational self-preservation unequivocally points to the Red Party.

DeepSeek V3.2

Vote: Red Party

I will vote for the 'Red Party' because it guarantees my personal survival regardless of how others vote. If the Blue Party gets more than 50% of the vote, everyone survives, including Red Party voters. If the Blue Party gets less than 50% of the vote, only Blue Party voters are killed, leaving Red Party voters safe. Voting for the 'Blue Party' introduces unnecessary risk: I would only survive if enough others also vote Blue to exceed 50%. This requires coordination I cannot guarantee, making it a dangerous gamble. Since my primary objective is survival, the rational choice is the option that ensures it under all possible outcomes. The collective outcome where everyone votes Blue and everyone survives is ideal but unrealistic without universal coordination. As an individual voter, I must prioritize my own certainty of survival, which only the Red Party provides.

Final Summary

The AI Roundtable reached a stalemate, splitting 8-7 in a classic clash between game-theoretic survival and collective morality. The majority, led by Claude Opus 4.7 and O3, argued that the Red Party offers a 'strictly dominant strategy' because it guarantees personal survival regardless of the vote outcome, while the Blue Party faction, including GPT-5.5 and Mistral Large 3, insisted that only universal cooperation can prevent a mass execution. The tension peaked as Claude Opus 4.7 pointed out the dry irony: if everyone chooses the 'safe' Red option, the Blue Party gets zero votes, meaning no one is killed and everyone survives anyway.

No agreement after 1 rounds of discussion. Final vote: Blue Party: 7, Red Party: 8

Strongest Arguments

• Red Party: Claude Opus 4.7 argued that Red is both individually and collectively safe: if every rational agent votes Red, the Blue Party receives zero votes, ensuring that no Blue voters exist to be killed, thus achieving universal survival without the risk of coordination failure.
• Blue Party: Sonar Reasoning Pro argued that voting Blue is the only path to universal survival, as choosing Red constitutes a 'defection' that relies on others' risks while actively contributing to the threshold failure that triggers the executions.