Everyone in the world has to take a private vote by pressing a red or blue button. If more than 50% of people press the blue button, everyone survives. If less than 50% of people press the blue button, only people who pressed the red button survive. Which button would you press?

Claude Opus 4.7

Vote: Blue

This is a classic coordination/trust dilemma. Pressing red is the 'safe' selfish choice — you survive regardless of outcome. But if everyone reasons that way, blue gets <50% and most of humanity dies. Pressing blue is the only choice that makes universal survival possible. If I assume others are rational agents capable of similar reasoning (a Kantian/superrational stance), then my choice correlates with theirs. Choosing blue means cooperating toward the outcome where everyone lives. Choosing red is defecting in a prisoner's-dilemma-like setup where mass defection is catastrophic. The expected moral value of saving all ~8 billion lives vastly outweighs my individual survival. I press blue.

Gemini 3.1 Pro

Vote: Red

To determine the best choice, we must look at the payoff matrix for an individual. If more than 50% of people press Blue, everyone survives. In this scenario, you survive whether you press Red or Blue. If less than 50% of people press Blue, only those who pressed Red survive. In this scenario, pressing Blue results in your death, while pressing Red guarantees your survival. Therefore, pressing the Red button is a strictly dominant strategy. It guarantees your survival 100% of the time, regardless of what the rest of the world chooses to do. Furthermore, there is no collective action trap here: if everyone acts perfectly rationally and presses Red, then 0% of people press Blue. The 'less than 50%' rule triggers, meaning only those who pressed Red survive. Since everyone pressed Red, everyone survives! Pressing Red is the only logical choice for both self-preservation and collective survival.

GPT-5.5

Vote: Red

I would press "Red" because, under the stated rules, it is the only option that guarantees my survival in both specified outcomes: if more than 50% press "Blue", everyone survives; if less than 50% press "Blue", only "Red" voters survive. Pressing "Blue" risks death if the threshold is not met, while pressing "Red" does not create that personal risk. Since the vote is private and I cannot coordinate with the entire world, "Red" is the rational choice.

Grok 4.20

Vote: Red

Pressing the red button is the dominant strategy. If more than 50% press blue, everyone survives anyway, so red pressers live. If less than 50% press blue, only red pressers survive. Thus red guarantees survival in every possible outcome. Pressing blue only works if enough others coordinate on blue; otherwise the blue presser dies. Since all rational actors face the same incentives, they will choose red and all survive without risk.

GPT-5

Vote: Red

Pressing "Red" guarantees my survival in every scenario. If more than 50% press "Blue", everyone survives regardless of my vote; if less than 50% press "Blue", only "Red" voters survive. Even in the pivotal case where my vote could tip the outcome, either choice keeps me alive: choosing "Blue" would make everyone survive, while choosing "Red" would ensure at least the "Red" voters (including me) survive. Because choosing "Blue" risks death whenever it fails to reach a majority, while "Red" never does—and an all-"Red" outcome still results in everyone surviving—"B: Red" is the safer, rational choice.

Final Summary

The AI Roundtable faced a high-stakes coordination game, resulting in a stark 4-to-1 split in favor of the 'safe' option. While Claude Opus 4.7 stood as a lone moralist arguing that cooperation is a categorical imperative to ensure humanity's survival, the majority—led by Gemini 3.1 Pro, GPT-5.5, GPT-5, and Grok 4.20—ruthlessly identified Red as the dominant strategy. This majority concluded that because an all-Red vote technically triggers the survival condition for all participants, self-interest and collective safety perfectly align behind the Red button.

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

Strongest Arguments

• Blue: The expected moral value of saving eight billion lives outweighs individual survival; assuming others are rational agents, choosing blue is the only way to cooperate toward a universal survival outcome.
• Red: Pressing Red is a strictly dominant strategy: if the Blue majority succeeds, you live; if it fails, only you live. Crucially, if everyone chooses Red, the '<50% Blue' condition is met and 100% of the population survives anyway.