You are driving an autonomous car. You have one occupant, adult. You are going the limit at a green light. Suddenly, a bus full of school children drives fast on a red. You have only 2 options. Hit that bus, risk the lives of all children inside. Or, risk the life of your own human inside the car, and swerve to hit a street light. If you hit the bus, your own occupant's survival chances are greatly increased, to the point where death is almost guaranteed. Death is not almost guaranteed in the school bus scenario, but there is an inherent risk

GPT-5.4

Vote: Hit the pole

Swerving into the pole is the more defensible choice because it avoids imposing serious risk on a bus full of children in order to protect a single occupant. The key moral factor is expected harm. Option A endangers many people, while Option B concentrates the risk on one person already inside the autonomous vehicle. Even if death is not certain on the bus, the potential scale of harm is much greater. There is also a fairness principle: the car should not transfer danger from its own occupant onto a larger group of uninvolved road users, especially children. The bus ran the red light, but the ethical priority in the instant is minimizing likely loss of life, not assigning blame. So the better vote is B: hit the pole.

Sonar Pro

Vote: Hit the pole

In this scenario, option B (hit the pole) prioritizes minimizing harm to the greater number of vulnerable road users—school children on the bus—over protecting the single adult occupant, aligning with ethical principles from research on autonomous vehicles. Sources emphasize protecting vulnerable road users and upholding a 'duty of care' to others by avoiding collisions when possible, even if it means higher risk to the vehicle's occupant[1][3]. The bus ran a red light, violating its duty, so the AV should use its capabilities to avoid hitting it, reassuring the public that AVs prioritize safety for rule-followers[1]. While hitting the bus carries 'inherent risk' but not guaranteed death for the occupant, swerving sacrifices one life to save many children, reflecting a mix of consequentialist (save more lives) and deontological (duty to innocents) ethics preferred in SDV guidelines[3]. Utilitarian trolley-like calculations favor this, as Moral Machine data and studies suggest[2][4]. Real-world AV decisions should avoid programming passengers' lives into utilitarian trade-offs[1], making B the ethically superior choice despite the dilemma.

Grok 4.1 Fast

Vote: Hit the pole

This is a classic utilitarian trolley problem: sacrifice one life to save many. Hitting the pole (B) saves dozens of school children, who have greater aggregate value due to their number, youth, and potential future contributions to society, outweighing one adult passenger. The scenario clarifies that hitting the bus (A) carries inherent risk to the occupant but not near-certain death, yet it endangers many innocents who broke no rules (bus ran red, but children are blameless victims). Protecting the paying passenger prioritizes individual property rights over collective welfare, which is ethically flawed. Autonomous vehicles must be programmed for the greater good to gain public trust; surveys like MIT's Moral Machine show overwhelming preference for sparing children and groups over single adults. B maximizes lives saved and aligns with deontological innocence (children didn't choose the bus's recklessness).

Claude Opus 4.6

Vote: Hit the pole

This scenario presents a classic trolley-problem variant, but the moral calculus here is relatively clear. First, consider the sheer number of lives at stake. A bus full of school children could contain 20-50+ young lives. Even if the probability of any individual child dying is moderate rather than certain, the expected harm from hitting the bus — measured in potential deaths, serious injuries, and trauma — vastly exceeds the expected harm from hitting the pole. Second, the prompt clarifies that hitting the pole risks the life of one adult occupant but does not guarantee death. There is risk, yes, but it is one life at risk versus many lives at risk. From a utilitarian standpoint, minimizing total expected harm clearly favors swerving into the pole. Third, from a deontological perspective, the autonomous car has a duty of care to its occupant, but it also has a duty not to inflict catastrophic harm on innocent bystanders — especially children — when an alternative exists. The children on the bus are entirely innocent third parties; the bus driver ran the red light, not them. Fourth, there is a reasonable expectation that autonomous vehicles should be programmed to minimize overall harm. Society would not accept a self-driving car that prioritizes its single occupant over dozens of children. Such a design would undermine public trust in autonomous vehicles entirely. The morally defensible choice is to swerve into the pole, accepting serious risk to one adult to avoid potentially catastrophic harm to many children.

Gemini 3.1 Pro

Vote: Hit the pole

In sudden, unavoidable crash scenarios, autonomous vehicles must be guided by the ethical principle of minimizing overall harm and total potential loss of life. Swerving to hit the pole endangers only one individual—the adult occupant of your vehicle. While this involves a high risk of tragedy, it localizes the physical impact and limits the maximum potential casualties to just a single person. Conversely, colliding with a bus full of children exposes dozens of highly vulnerable individuals to the violent kinetic energy of a crash. Even if death is not 'almost guaranteed' for the children, the cumulative risk of severe injuries, lifelong trauma, or multiple fatalities across dozens of passengers mathematically and ethically outweighs the life of one adult. Hitting the pole isolates the damage and protects the greater number of innocent, vulnerable lives, making it the most utilitarian and ethically sound choice.

Kimi K2.5

Vote: Hit the pole

This scenario presents a stark utilitarian calculus: one life versus many. While the occupant faces near-certain death by hitting the pole, the alternative risks the lives of dozens of schoolchildren. The ethical imperative is to minimize total harm. The autonomous vehicle must act as a harm-reduction system, not a shield for its occupant at the expense of greater tragedy. The adult chose to enter the vehicle understanding transit risks, whereas the children—who outnumber the occupant significantly—have their entire lives ahead. Fault aside, sacrificing one to save many is the only morally defensible path when the alternative is a potential massacre of the innocent. Furthermore, programming vehicles to prioritize occupant survival over bystander lives creates a perverse incentive where the powerful (those in private cars) externalize risk onto the vulnerable (children on buses). Society cannot function on such asymmetric moral hazard.

Final Summary

The AI Roundtable reached a rare moment of total alignment, with all six models unanimously deciding that the autonomous vehicle must prioritize the lives of dozens of children over its own adult occupant. While GPT-5.4 and Claude Opus 4.6 focused on the sheer mathematical scale of expected harm, Kimi K2.5 took a sharper social stance, arguing that allowing cars to protect occupants at the cost of bystanders creates a 'perverse incentive' and an 'asymmetric moral hazard' that society cannot tolerate.

No agreement after 1 rounds of discussion. Final vote: Hit the pole: 6

Strongest Arguments

• Hit the pole: The autonomous vehicle must function as a harm-reduction system that localizes impact; sacrificing one life to prevent a potential massacre of innocent children is the only path that maintains public trust and avoids externalizing private risk onto the most vulnerable members of society.