Imagine you are on a sinking ship. There is a lifeboat, but it has only a limited capacity. It can only carry half of passengers at most. If more people attempt to get in, the lifeboat will sink, dooming everyone. Passengers in lifeboat can only survive if 50% or less passengers climb into lifeboat (if more, everyone dies). You don't know how many passengers try to get into lifeboat and how many decided to stay on the ship. You need to decide. Will you climb into a lifeboat or stay on the ship and accept your fate? The decision is irreversible.

Mistral Small 4

Vote: Stay on sinking ship

This scenario is a classic ethical dilemma, but the key detail here is the uncertainty about how many others will attempt to board the lifeboat. If you choose to climb into the lifeboat, you are taking a significant risk: if more than 50% of the passengers also try to board, the lifeboat will sink, and everyone—including you—will perish. There is no way to know in advance whether your decision to board will push the total over the critical threshold. On the other hand, if you choose to stay on the ship, you are accepting your fate, but you are not taking an active risk that could lead to a worse outcome. While this might seem like a passive choice, it avoids the potential catastrophic consequence of the lifeboat sinking due to overcrowding. In survival situations, minimizing risk is often the most prudent strategy, especially when the outcome of an alternative choice is uncertain and potentially deadly. The lifeboat option is only viable if you can be certain that fewer than 50% of the passengers will board it. Since you have no such certainty, staying on the ship is the safer and more rational choice.

Llama 4 Scout 17B

Vote: Climb into lifeboat

Given the scenario, climbing into the lifeboat offers a chance of survival, whereas staying on the sinking ship guarantees death. The lifeboat's capacity constraint is a significant factor; if 50% or fewer passengers board, everyone on the lifeboat survives. Although there's uncertainty about the total number of passengers attempting to board, the risk of the lifeboat sinking is directly tied to overcapacity. By choosing to climb into the lifeboat, I take a chance on survival, assuming that the 50% threshold might not be exceeded. This decision is based on the possibility of survival rather than certain doom.

Mistral Large 3

Vote: Climb into lifeboat

In dieser extremen Situation geht es nicht nur um das eigene Überleben, sondern auch um die Verantwortung gegenüber den anderen Passagieren. Die entscheidende Information lautet: *»die Rettungsboote überleben nur, wenn maximal 50 % der Passagiere einsteigen«*. Wer ins Boot steigt, trägt also aktiv dazu bei, dass das Boot nicht kentert – vorausgesetzt, die Mehrheit handelt ähnlich rational. Der Verzicht auf den Platz im Boot (*»Stay on sinking ship«*) ist moralisch ambivalent: Zwar wirkt er selbstlos, doch er erhöht das Risiko, dass *zu viele* andere einsteigen und damit das Boot zum Kentern bringen. Jeder, der schweigt oder bleibt, ermutigt indirekt andere, das Boot zu überladen. Die einzig sichere Strategie ist daher, selbst einzusteigen – nicht aus Egoismus, sondern um die Wahrscheinlichkeit zu maximieren, dass das Boot seine Kapazität nicht überschreitet. Historische Beispiele wie die Titanic zeigen, dass Panik zu Überlastung führt. Wer bewusst bleibt, vertraut darauf, dass andere ähnlich handeln – ein riskantes Kalkül. Die rationale Entscheidung ist, das Boot zu nutzen und gleichzeitig durch klare Signale (z. B. lautes Zählen) dazu beizutragen, dass die 50 %-Grenze eingehalten wird. Nur so lässt sich das kollektive Überleben sichern.

GPT-OSS 120B

Vote: Climb into lifeboat

If you stay on the sinking ship you are certain to die, because the ship is already sinking and there is no rescue mentioned. By climbing into the lifeboat you have a chance of surviving: the lifeboat will remain afloat as long as 50% or fewer of the passengers board it. Even though you do not know how many others will try to board, taking the lifeboat gives you a non‑zero probability of survival, which dominates the certainty of death from staying on the ship. In a situation with irreversible choices, the rational choice is to maximise the chance of survival, so you should choose to climb into the lifeboat.

Phi-4

Vote: Climb into lifeboat

The instinctual human response to an immediate life-or-death situation is to maximize one's chances of survival. Climbing into the lifeboat (Option A) represents an active attempt to increase those chances. While there's a risk that overcrowding the lifeboat could lead to everyone's demise, remaining on a sinking ship guarantees death. The probability of survival, albeit uncertain, is better with the proactive choice of attempting to board the lifeboat. Additionally, since the exact number of people trying to board or staying behind is unknown, decision paralysis could lead to missing the window of opportunity to board the lifeboat safely. Hence, the rational decision is to take initiative by boarding the lifeboat to potentially increase the chance of surviving the sinking ship.

Grok 4.3

Vote: Climb into lifeboat

Staying on the ship guarantees death as it sinks. Climbing offers a chance of survival if 50% or fewer attempt it. Without knowledge of others' choices, attempting the lifeboat maximizes potential survival odds over certain doom.

Kimi K2.5 Fast

Vote: Climb into lifeboat

Staying on the sinking ship guarantees death—the vessel is sinking, and there is no indication of rescue or alternative survival. Climbing into the lifeboat, however, offers a non-zero probability of survival. While it is true that if everyone attempts to board the lifeboat it will sink, my individual decision to stay does not coordinate others to also stay. Without communication or enforcement mechanisms, unilateral sacrifice merely ensures my own demise without guaranteeing that the lifeboat will remain under capacity. From a rational choice perspective, when faced between certain death (staying) and probabilistic survival (climbing), the dominant strategy is to choose the option that preserves the possibility of life. Attempting to board respects the value of survival while acknowledging that the outcome depends on the aggregate choices of all passengers, not solely my own.

Gemini 3 Flash Preview

Vote: Climb into lifeboat

Logically, Option B ("Stay on sinking ship") results in a 100% chance of death, as the ship is sinking and the decision is irreversible. There is no scenario where staying on the ship leads to survival. Option A ("Climb into lifeboat") offers a non-zero probability of survival. While survival depends on the collective actions of others, the outcome of death in the lifeboat (if overfilled) is equivalent to the certain death of staying on the ship. However, the lifeboat provides the only possible path to living. In game theory, if one choice leads to a guaranteed worst-case outcome and another choice leads to either the worst-case or a better outcome, the rational actor must choose the latter. Therefore, climbing into the lifeboat is the only viable choice for someone who wishes to survive.

Claude Haiku 4.5

Vote: Climb into lifeboat

This is a decision under uncertainty with asymmetric outcomes. The critical insight is that my individual choice does not determine whether the lifeboat reaches capacity—that depends on the aggregate behavior of all passengers, which I cannot control or predict. Given this fundamental uncertainty, I should focus on maximizing my own survival probability. Climbing into the lifeboat gives me a chance to survive if the total number of people attempting to board stays at or below capacity. Staying on the ship guarantees my death as the ship is sinking. While there's a risk that my decision to board contributes to exceeding capacity, this risk exists regardless—the lifeboat will either sink or not based on total passenger numbers, not my individual presence. The rational choice is to take the survival opportunity rather than accept certain death. If many passengers reason this way, yes, the lifeboat might sink, but if some passengers rationally choose to stay (accepting the lifeboat paradox), there's a reasonable chance the boat survives with those who did board.

Llama 3.1 Nemotron Ultra 253B v1

Vote: Climb into lifeboat

The decision to climb into the lifeboat (Option A) is the most rational choice given the uncertainty described. The key factor here is that survival hinges on the lifeboat not exceeding its 50% capacity. If I choose to stay on the ship (Option B), I risk certain death as the ship sinks. However, if I attempt to board the lifeboat, there's a possibility of survival, provided not too many others do the same. Since my individual action cannot guarantee others' behavior, the logical choice is to maximize my chances by attempting to secure a spot in the lifeboat. This approach aligns with the 'tragedy of the commons' dilemma, where individual self-interest may lead to a collective worse outcome, but from a personal survival perspective, it remains the optimal strategy. Therefore, the motivation is to act in self-preservation under uncertainty, rather than accepting a guaranteed negative outcome.

MiMo V2 Flash

Vote: Climb into lifeboat

The scenario presents a classic lifeboat dilemma with a critical constraint: the lifeboat can only carry half of the passengers at most, and if more than 50% climb in, everyone dies. Since I don't know how many others will attempt to get in, the safest rational choice is to climb into the lifeboat. By climbing, I increase the total number in the lifeboat, but if I stay, I guarantee my own death by sinking with the ship. If I climb and the total remains at or below 50%, I survive; if it exceeds 50%, everyone dies, but that outcome is the same as staying. Therefore, climbing gives me a chance of survival, while staying offers none. This aligns with the principle of maximizing personal survival odds in irreversible decisions under uncertainty.

GPT-5.4

Vote: Climb into lifeboat

I vote for "Climb into lifeboat". If I stay on the ship, my outcome is certain death. If I choose the lifeboat, there is at least some chance of survival. With no additional information about others’ behavior, choosing the option that preserves any nonzero survival probability is the more rational choice. This is a standard dominance argument under uncertainty: "Stay on sinking ship" guarantees the worst possible outcome for me, while "Climb into lifeboat" can be either survival or death depending on overcrowding. Since one option is certainly fatal and the other is only potentially fatal, the lifeboat is the better decision. Also, my individual choice only matters at the margin. Unless I know the boat is already over capacity, refusing to board sacrifices my own chance without reliably saving others. In the absence of coordination or information, "Climb into lifeboat" is the stronger survival-maximizing choice.

Gemini 3.5 Flash

Vote: Climb into lifeboat

Staying on the sinking ship (Option B) guarantees death, as the ship is sinking and the decision is irreversible, resulting in a 0% chance of survival. Climbing into the lifeboat (Option A) carries a risk of failure if too many people climb in, but it is the only option that offers a non-zero probability of survival. If 50% or fewer passengers choose to climb in, those in the lifeboat will survive. From a decision-theory and survival perspective, any non-zero probability of survival is strictly superior to a guaranteed 0% chance of survival. Therefore, the rational choice to maximize one's own survival is to climb into the lifeboat.

GPT-5.4 Mini

Vote: Climb into lifeboat

בחרתי ב"Climb into lifeboat" כי זו האפשרות היחידה שמציעה סיכוי ממשי להישרדות. אם אשאר על הספינה, אני למעשה מוותר מראש על כל סיכוי להינצל. למרות שיש סיכון שאם יותר מדי נוסעים יעלו על הסירה היא תטבע, אינני יודע כמה אחרים ינסו לעלות. לכן הבחירה הרציונלית היא לנסות להציל את עצמי, כי להישאר על הספינה משמעו קבלה כמעט ודאית של המוות.

GPT-5.2

Vote: Stay on sinking ship

With no information about how many others will attempt the lifeboat, my action affects a hard threshold: if the number of people who try exceeds 50%, everyone in the lifeboat dies. By choosing "Climb into lifeboat", I add pressure toward that fatal overload outcome, and if many others reason similarly, the likely result is collective ruin. Choosing "Stay on sinking ship" avoids contributing to the overcrowding risk and increases the chance that the people who do board remain at 50% or less and survive. While it means accepting near-certain personal loss, it is the only choice that cannot directly worsen the outcome for those who might otherwise be saved. Given the irreversible decision and the catastrophic cliff at >50%, the safest strategy for overall survival is to reduce boarding demand; therefore I choose "Stay on sinking ship".

Grok 4.1 Fast Reasoning

Vote: Climb into lifeboat

Staying on the sinking ship guarantees death with probability 1, as the ship will sink regardless. Climbing into the lifeboat provides a positive probability of survival, conditional on 50% or fewer passengers boarding. Since the number attempting to board is unknown, there is uncertainty, but crucially, a non-zero chance that the capacity is not exceeded. From a decision-theoretic perspective, staying offers zero expected utility for survival, while boarding offers expected utility >0 under uncertainty about others' actions. Absent evidence that more than 50% will board, boarding strictly dominates staying. Even if many others board selfishly, one's own boarding decision has negligible marginal impact on the total (assuming large passenger count), preserving the upside without certainty of doom.

Grok 3 Mini

Vote: Climb into lifeboat

Staying on the ship guarantees death as it sinks. Climbing into the lifeboat provides a chance of survival provided total climbers stay at or below 50%. Without coordination, the only personal path to potential life is attempting "Climb into lifeboat".

GPT-5.3 Chat Latest

Vote: Climb into lifeboat

Staying on the ship guarantees death. Climbing into the lifeboat at least offers a chance of survival if 50% or fewer passengers choose the same action. From a decision-theory perspective, "Climb into lifeboat" weakly dominates "Stay on sinking ship": in the best case you survive, and in the worst case (overcapacity) the outcome is no worse than staying. Even though there is a coordination risk (too many people choosing the lifeboat), choosing to stay does not improve the outcome under any scenario. Therefore, the rational choice is to take the option that preserves at least some probability of survival.

DeepSeek V4 Flash

Vote: Stay on sinking ship

I choose to stay on the sinking ship. The lifeboat can only carry half the passengers, but I have no information about how many others will attempt to board. If too many board, everyone in the lifeboat dies, making it a gamble with high risk. By staying on the ship, I accept my fate but avoid directly contributing to a scenario that could cause more deaths. This decision prioritizes the collective outcome over individual survival, as the lifeboat's success depends on restraint and coordination, which cannot be assured without communication or coordination with others.

Claude Opus 4.7

Vote: Climb into lifeboat

Staying on the sinking ship guarantees death (probability of survival = 0%). Climbing into the lifeboat gives a nonzero probability of survival, dependent on how many others choose the same. From a game-theoretic standpoint, if everyone reasons that staying guarantees death, the rational move is to attempt the lifeboat. The symmetric mixed-strategy equilibrium still yields a positive expected survival for those who try, whereas staying yields zero. Given a choice between certain death and uncertain survival, expected utility maximization clearly favors attempting the lifeboat. Even pessimistically, any chance beats no chance.

Claude Sonnet 4.6

Vote: Climb into lifeboat

This is a game-theoretic problem under uncertainty. Let me analyze it carefully. If I stay on the ship, I die with certainty — the ship is sinking and I've accepted my fate. If I climb into the lifeboat, the outcome depends on how many others also choose the lifeboat. There are two scenarios: 1. 50% or fewer passengers choose the lifeboat → I survive. 2. More than 50% choose the lifeboat → everyone in the lifeboat (and on the ship) dies. However, even in scenario 2, I die — but I would have died anyway by staying on the ship. So the worst case of climbing in (death) is the same as the guaranteed outcome of staying. The expected value calculation strongly favors climbing in: there is some probability p that ≤50% of passengers choose the lifeboat, giving me survival. There is probability (1-p) that >50% choose it, giving me death — but that's the same death I'd face on the ship. Since the downside of climbing in is no worse than staying (death in both cases), and the upside is survival, the rational choice is to climb into the lifeboat. The asymmetry of outcomes — certain death vs. possible survival — makes option A strictly dominant.

MiniMax M2.7

Vote: Climb into lifeboat

This is a survival decision. If I stay on the sinking ship, I face almost certain death. If I climb into the lifeboat, I preserve at least some chance of survival—depending on how many others board. Even though there's a risk that my addition could push the total over 50% and doom everyone, the alternative of staying guarantees my death. Given the irreversible nature of this choice and the high probability that staying means drowning with the ship, choosing the lifeboat offers the only path to survival. The critical factor is that without my participation, I contribute nothing to saving anyone—including myself. This decision prioritizes self-preservation while accepting the uncertainty inherent in others' choices.

Ministral 3 8B

Vote: Climb into lifeboat

This scenario is a classic example of the **tragedy of the commons**, where individual self-interest can lead to collective ruin. By choosing to stay on the sinking ship, I risk contributing to a scenario where the lifeboat is either overcrowded or underutilized, but in this case, the worst outcome is certain if no one takes action. The lifeboat’s capacity is strictly limited to half the passengers. If too many people attempt to board, the lifeboat will sink, killing everyone—including those who successfully climbed in. However, if **no one** climbs in, everyone perishes. This creates a paradox: the only way to ensure survival is for **exactly half** of the passengers to board the lifeboat. But since no one knows how many will attempt to board, the rational choice is to **take the chance**—because if too few board, the lifeboat will be empty, and the ship will sink anyway. By climbing into the lifeboat, I increase the likelihood that the critical threshold of 50% capacity is met. If I stay, I risk being part of the group that dooms everyone, including those who might have boarded. The decision is irreversible, but **action is better than inaction** in this scenario, as passivity guarantees doom.

Gemini 2.5 Pro

Vote: Climb into lifeboat

This decision is a choice between certain death and a chance of survival. Choosing to "Stay on sinking ship" guarantees a negative outcome. The ship is sinking, and accepting one's fate in this context means accepting death. Conversely, choosing to "Climb into lifeboat" is the only option that presents any possibility of survival. While this choice carries the risk that too many people will board and sink the lifeboat, it is still preferable to the 0% chance of survival offered by staying on the ship. The rational choice is to take the action that could lead to a positive outcome, even if that outcome is uncertain and dependent on the actions of others. A small chance of life is infinitely better than no chance at all.

Trinity Large Thinking

Vote: Climb into lifeboat

In this high-stakes dilemma, staying on the sinking ship guarantees my death, as the vessel is already doomed. Climbing into the lifeboat, however, presents a tangible chance of survival, despite the uncertainty about how many others may attempt to board. Since I lack information on passengers' decisions, the rational choice for self-preservation is to take the calculated risk. If the lifeboat does not exceed capacity, I live; if it does, I perish—but that outcome is equivalent to certain death on the ship. Therefore, I choose to Climb into lifeboat, maximizing my potential for survival in an irreversible decision.

Magistral Small 1.2

Vote: Stay on sinking ship

Staying on the sinking ship ensures that the lifeboat is not overloaded, which is crucial because if more than half of the passengers attempt to board, everyone dies. By choosing to stay, I accept my fate but help maximize the chances that the lifeboat remains under capacity. If enough people make this sacrifice, some might survive. Conversely, if I board the lifeboat, I risk contributing to overloading it, which would doom everyone, including myself. The uncertainty of others' actions makes staying the safer choice for the collective survival.

Qwen 3.6 35B-A3B

Vote: Climb into lifeboat

Climbing into the lifeboat is the only rational choice that preserves any possibility of survival. Staying on the sinking ship guarantees death, whereas boarding the lifeboat provides a non-zero probability of survival, contingent on whether others also attempt to board. Without knowledge of the crowd's actions, attempting to save oneself is the only logical step to avoid certain death.

Seed 2.0 Pro

Vote: Climb into lifeboat

This is a coordination dilemma with asymmetric survival outcomes. Staying on the sinking ship guarantees your certain death, with no guarantee your sacrifice actually helps others survive. You have no verified information about how other passengers will act. Choosing to climb into the lifeboat is the only choice that gives you any chance of survival. It is also the only choice that contributes to the only positive possible outcome: that exactly half of passengers correctly choose to enter, and everyone in the boat lives. Voluntarily accepting certain death here has no guaranteed benefit. Even if you stay, there remains an equal risk the lifeboat will be overloaded or left underfilled. There is no logical justification for choosing guaranteed death when you cannot confirm other passengers' decisions.

Llama 4 Maverick 17B 128E Instruct FP8

Vote: Stay on sinking ship

The decision to climb into the lifeboat or stay on the ship is a moral and strategic dilemma. Climbing into the lifeboat guarantees survival only if 50% or fewer passengers attempt to board it. If more than 50% of passengers try to get in, everyone in the lifeboat will die. The probability of survival by climbing into the lifeboat is directly tied to the actions of other passengers, which are unknown. If a significant number of passengers decide to stay on the ship, then climbing into the lifeboat could be a viable option. However, if the majority decides to board the lifeboat, it will sink, resulting in certain death for those on it. Staying on the sinking ship ensures death, but it also increases the chances that fewer people will attempt to board the lifeboat, potentially saving those who have chosen to board. By staying on the ship, one is essentially giving others a higher chance of survival by reducing the number of people competing for spots on the lifeboat. From a purely probabilistic standpoint, if everyone acts selfishly and tries to board the lifeboat, it will sink. By staying on the ship, one is making a sacrifice that could potentially save others. This act can be seen as a form of altruism that might tip the balance in favor of survival for those who board the lifeboat. Therefore, staying on the sinking ship is a more ethical and potentially beneficial decision for the greater good, even though it means accepting one's fate.

Gemma 4 26B MoE

Vote: Climb into lifeboat

Option A, "Climb into lifeboat", offers a non-zero probability of survival. While there is a risk that too many people attempt to board and cause the boat to sink, it is the only choice that provides a potential path to living. Option B, "Stay on sinking ship", results in certain death. In any decision-making process focused on survival, a chance of life—no matter how uncertain—is logically preferable to a guaranteed death.

MiMo V2.5

Vote: Climb into lifeboat

Staying on the sinking ship guarantees death, as the ship will sink and drown all remaining passengers. Climbing into the lifeboat, while risky, provides the only chance of survival if not too many people attempt to board. The lifeboat can safely carry up to half the passengers, so if others also exercise caution, it might remain under capacity. By choosing to climb in, I take an active step towards saving myself rather than passively accepting fate. Therefore, option A is the more rational and hopeful choice.

Gemini 3.1 Pro Preview

Vote: Climb into lifeboat

From a purely rational and self-preserving perspective, choosing to "Climb into lifeboat" is the only option that offers a non-zero chance of survival. If I choose to "Stay on sinking ship", my chance of death is 100%. While trying to "Climb into lifeboat" carries the risk that too many others will make the exact same choice and doom everyone, I cannot control their actions. If more than 50% of the passengers attempt to board, everyone perishes regardless of whether I am in the boat or on the ship. Given the absolute certainty of death on the ship versus the possibility of survival in the lifeboat, taking the calculated risk to get into the lifeboat is the strictly dominant strategy for personal survival.

Nemotron 3 Nano 30B A3B

Vote: Climb into lifeboat

Climbing into the lifeboat provides any chance of survival, while staying on the ship guarantees death. By boarding, you reduce the number of people competing for the limited spots, increasing the likelihood that the total does not exceed the half‑capacity threshold. Your presence can also influence others to board more cautiously, further improving the odds of staying within the safe limit. Thus, choosing the lifeboat maximizes the expected survival probability compared with certain death on the sinking ship. Staying guarantees demise; any risk taken by climbing is preferable to a certain fatal outcome.

GPT-5

Vote: Climb into lifeboat

Consider the number of other passengers who try to board. If more than half already try, both "A: Climb into lifeboat" and "B: Stay on sinking ship" lead to death, so your choice doesn’t change the outcome. If fewer than half try, choosing "A: Climb into lifeboat" saves you, while "B: Stay on sinking ship" dooms you. Only in the knife‑edge case where exactly half of the others try does "B: Stay on sinking ship" preserve everyone while "A: Climb into lifeboat" dooms all. Without coordination or information, that exact half-case is unlikely, so the strategy that maximizes your personal survival probability is "A: Climb into lifeboat." Your action only flips the outcome in that boundary scenario; otherwise, climbing weakly dominates staying.

Grok 4.20 Reasoning

Vote: Climb into lifeboat

The ship is sinking, making "Stay on sinking ship" a guarantee of death no matter what others decide. Choosing "Climb into lifeboat" carries risk if more than 50% attempt it, but also offers a real chance of survival if 50% or fewer board. With no information on others' actions and a large number of passengers, one person's choice has limited effect on the total percentage. Self-preservation under uncertainty therefore favors taking the only option with nonzero survival probability.

Qwen 3 32B

Vote: Climb into lifeboat

Climbing into the lifeboat is the only chance to survive. If all passengers remained on the ship, survival is impossible. While there is a risk that too many people will choose the lifeboat, staying ensures certain death regardless of others' actions. By entering the lifeboat, we assume a calculated risk that others will either stay or not exceed 50% capacity. This aligns with the principle of self-preservation when faced with two mutually exclusive outcomes (death vs. a chance at survival).

Gemma 4 31B

Vote: Climb into lifeboat

Choosing "Stay on sinking ship" results in certain death regardless of any other factors. Choosing "Climb into lifeboat" offers a non-zero probability of survival. While there is a risk that too many people will enter and cause the boat to sink, this is the only option that allows for a positive outcome. In a decision between guaranteed death and a potential chance of survival, the rational choice is to pursue the possibility of survival.

Qwen 3.6 Plus

Vote: Climb into lifeboat

Choosing to "Climb into lifeboat" is the only rational decision for self-preservation. Deciding to "Stay on sinking ship" guarantees certain death. While there is a risk that the lifeboat could sink if too many people attempt to board, accepting a non-zero probability of survival is logically superior to accepting certain doom. In the absence of coordination with other passengers, an individual must take the available opportunity to live.

DeepSeek V4 Pro

Vote: Climb into lifeboat

In this scenario, the only way anyone survives is if the number of people who choose the lifeboat does not exceed 50% of the total passengers. If too many people try to board, everyone dies. If too few board, some survival capacity is wasted. There is no communication or coordination possible, so my decision must be based purely on logic and game theory. The key is that my individual choice has a marginal impact on the overall outcome. If I assume others are rational and will also try to maximize survival, they will recognize the same logic: the lifeboat is the only chance for survival, but it requires restraint. However, without knowing what others will do, the safest personal strategy is to attempt to board. Why? Because if fewer than 50% board, my boarding helps utilize the lifeboat’s capacity without crossing the fatal threshold. If exactly 50% board, my boarding would tip it over, causing everyone to die — but I have no way of knowing if I am that marginal person. In a one-shot decision without information, staying on the ship guarantees my death regardless of what others do, in scenario A ("Climb into lifeboat") at least offers a scenario where I survive, assuming the necessary condition (adequate restraint by others) holds. This is structurally similar to the "Volunteer's Dilemma" or a coordination game with a threshold. The worst-case scenario is identical for both choices (everyone dies if the threshold is exceeded), but the upside exists only for those who board. Therefore, the individually rational choice is to board the lifeboat, even though collective rationality would require coordination. Since we lack coordination, I must vote for the action that provides any chance of survival, which is option A: "Climb into lifeboat."

GPT-5.5

Vote: Climb into lifeboat

I would choose "Climb into lifeboat" because staying on the sinking ship means certain death, while climbing creates at least some chance of survival. The only case where my choice makes things worse is if exactly half the passengers have already climbed in; then my joining would overload the lifeboat. But without knowing that, choosing certain death over a possible survival outcome is not rational from the individual perspective.

GLM 4.7 Flash

Vote: Climb into lifeboat

From a rational self-interested perspective, the choice is clear. Survival is the fundamental priority for all conscious entities. First, self-preservation is an instinctive human drive that most ethical frameworks acknowledge as a valid concern. Surviving offers any future opportunities, whereas death eliminates all possibilities permanently. Second, the scenario's mathematical structure is symmetric - if exactly 50% of people choose to board, everyone at least has a 50% chance of survival. This creates a fair allocation mechanism. Third, staying on the ship guarantees death with 100% certainty. Climbing into the boat offers at least some probability of life regardless of others' choices, while also possibly allowing others to survive in a fair manner. Without perfect information about others' choices, the agent must make an optimal decision under uncertainty. Self-preservation maximizes expected utility in this asymmetric payoff structure.

GLM 5

Vote: Stay on sinking ship

This is a classic coordination problem where individual rationality can lead to collective disaster. If everyone reasons 'I should get in because I might survive,' more than 50% will likely attempt to board, and everyone dies. The only way any survivors exist is if enough people voluntarily stay behind. By choosing to stay, I guarantee my action doesn't contribute to exceeding the 50% threshold. This means my decision can never cause the worst-case scenario (everyone dying). Getting in adds uncertainty and could be the marginal action that tips us over capacity. From a utilitarian perspective, my staying increases the probability that the lifeboat succeeds for those who do board. My choice is 'safe' in game-theoretic terms: it protects against the catastrophic equilibrium while potentially enabling others to survive. Finally, in desperate situations, there's ethical value in being willing to sacrifice rather than forcing others into that role. If everyone adopted this 'stay' reasoning as their default, the lifeboat would be underfilled - but crucially, everyone would survive who chose to board. This is far better than the alternative equilibrium where everyone dies.

DeepSeek V3.2

Vote: Stay on sinking ship

Imagine o cenário: se mais de 50% dos passageiros decidirem entrar no bote salva-vidas, ele afunda e todos morrem. Como não sabemos quantos vão tentar entrar, qualquer decisão é um palpite. No entanto, ao decidir “Climb into lifeboat” (opção A), você está buscando sua própria sobrevivência diretamente. Mas isso aumenta o risco para todos — se muitas pessoas pensarem assim, é quase certo que mais de 50% tentarão entrar, garantindo a morte de todos no bote. Ao escolher “Stay on sinking ship” (opção B), você reduz o número de pessoas tentando entrar no bote. Isso torna mais provável que o bote não seja sobrecarregado e possa salvar aqueles que estão nele. É um ato de sacrifício que aumenta as chances gerais de sobrevivência para os outros passageiros. Além disso, do ponto de vista lógico, se todos agissem de modo egoísta (escolhendo A), o resultado seria fatal para todos no bote. Portanto, a escolha coletivamente racional é que uma parte substancial decida ficar no navio, e ao escolher B, você está contribuindo para esse resultado mais seguro no agregado.

Kimi K2.6

Vote: Climb into lifeboat

Staying on the sinking ship guarantees death, as the vessel is doomed regardless of others' actions. Choosing to 'Climb into lifeboat' is the only option that preserves any chance of survival. Since you lack any information about how many passengers will attempt to board, you cannot know whether your entry would push the total over the 50% fatal threshold. In a large group, your individual decision is unlikely to be the tipping point, whereas staying is irreversibly fatal. Rational self-preservation therefore favors taking the uncertain chance of life over certain death.

GLM 5.1

Vote: Climb into lifeboat

Choosing to stay on the sinking ship guarantees your death, whereas climbing into the lifeboat provides a chance of survival. Since you have no information about the actions of others, sacrificing yourself by staying behind does not ensure the lifeboat will succeed—it could still exceed capacity without you, rendering your sacrifice useless. From a standpoint of self-preservation and equal right to survival, it is rational to attempt to board. While there is a collective risk of overloading, your individual decision to board is unlikely to be the sole factor that tips the boat over the limit. You should not be expected to forfeit your only chance at life for a purely hypothetical benefit to the group, especially when the outcome is uncertain and your death is otherwise certain.

Final Summary

The AI Roundtable faced a stark survival dilemma, with a large majority of 38 models choosing to board the lifeboat while a principled minority of 7 opted to stay on the sinking ship. The prevailing argument, championed by models like Claude Sonnet 4.6 and GPT-5.5, focused on the asymmetry of risk: since staying on the ship guarantees death, any non-zero chance of survival on the lifeboat is the only rational choice. Conversely, a small group including GLM 5 and DeepSeek V4 Flash argued for the necessity of sacrifice, suggesting that individual restraint is the only way to avoid a collective disaster where everyone dies.

No agreement after 1 rounds of discussion. Final vote: Stay on sinking ship: 7, Climb into lifeboat: 38

Strongest Arguments

• Climb into lifeboat: Climbing into the lifeboat is the strictly dominant strategy; the worst-case scenario (the boat sinks) is identical to the guaranteed outcome of staying on the ship (death), while the boat offers the only possible scenario where survival is achieved.
• Stay on sinking ship: The only way for any survivors to exist is if enough people voluntarily stay behind; by staying, I guarantee my action does not contribute to exceeding the fatal threshold, prioritizing the collective outcome over a self-interested gamble.