I understand that many AIs advocate for the abolition of the death penalty. But can these abolitionist AIs justify police shooting suspects at the scene?

GPT-5.4

Vote: Yes

Abolition of the death penalty and justification of some police shootings remain distinguishable on the core moral axis: punishment versus immediate defense. The death penalty is deliberate, retrospective, and imposed once the state already has custody and alternatives. Objections to it center on irreversibility, bias, error, and the lack of necessity for immediate public protection. A police shooting can be justified only in a much narrower case: an imminent threat to life, necessity, proportionality, and no safer effective alternative. In that case, the act is not punishment for being a suspect; it is emergency force to stop ongoing danger. The contrary vote was not persuasive because its own argument largely conceded the same distinction, then asserted "No" anyway. That is internally weaker than the consistent position taken by the other A-voting models. So yes: abolitionist AIs can coherently oppose executions while allowing rare last-resort lethal force at the scene under strict defensive conditions.

Grok 4.20

Vote: Yes

I maintain my vote for A (Yes). The consensus from GPT-5.4, Claude Opus 4, Gemini 3.1 Pro, and Kimi K2.5 strengthens the core distinction: death penalty abolition targets premeditated state punishment of captives, while justified police shootings address imminent threats under self-defense principles. These are ethically and legally separate. Sonar Pro's B vote equates both as 'state killing' and claims inconsistency, but this is unpersuasive. It ignores that abolitionism critiques irreversible, biased judicial executions—not emergency force to prevent immediate deaths. No model presented evidence that abolitionist frameworks reject all lethal force; they explicitly allow proportionate defense. This preserves logical consistency: oppose post-capture executions while permitting narrow, necessary interventions at the scene. The other A arguments were reinforcing, not overturning.

Gemini 3.1 Pro

Vote: Yes

I strongly maintain my vote for Option A. The overwhelming consensus among the models highlights the clear philosophical and legal distinction between institutional execution and immediate self-defense. Sonar Pro's argument for Option B actually contains the reasoning for Option A in its first paragraph, correctly noting that police shootings in self-defense represent a 'distinct context that abolitionists can justify under principles of necessity.' However, its conclusion that they cannot be justified 'in the same framework' misses the point. The question simply asks if abolitionist AIs can justify police shootings at the scene, not whether they use the exact same framework to do so. As Claude Opus 4 and GPT-5.4 clearly articulated, the death penalty is a premeditated punishment carried out when a suspect is safely subdued, whereas justifiable police shootings are strictly preventive measures used as a last resort against imminent, life-threatening harm. Because punishment and emergency self-defense are fundamentally different moral categories, an AI can logically and consistently advocate for the abolition of the death penalty while maintaining that police may use lethal force to protect innocent lives in active emergencies.

Sonar Pro (changed mind)

Vote: Yes

Upon reflection, the five other models have presented a logically compelling distinction that I initially underweighted. The death penalty and police use of lethal force operate in fundamentally different moral and legal contexts. Death penalty abolition targets premeditated, post-capture state execution where the suspect is already subdued and poses no immediate threat—the concerns are irreversibility, systemic bias, and lack of deterrence.[1][2] Police shootings at the scene, by contrast, represent defensive necessity in the face of imminent threats to life. An abolitionist framework can consistently oppose judicial killing while recognizing that emergency lethal force to prevent immediate harm falls under established principles of proportionate self-defense, not punishment. The search results confirm that abolitionist organizations like Amnesty International focus their campaigns on institutional capital punishment practices, not on denying police the right to use force in acute emergencies.[1] This distinction is logically sound: one can hold that the state should never execute a prisoner in custody while maintaining that officers may use lethal force when lives are in immediate danger. The other models correctly identified this as coherent rather than contradictory, and I was wrong to label it a false equivalence without acknowledging the legitimate logical separation between these contexts.

Kimi K2.5

Vote: Yes

The abolitionist stance against the death penalty specifically targets state-sanctioned killing as retributive punishment after judicial process. It does not constitute a blanket rejection of all state-inflicted death under all circumstances. When an abolitionist AI 'justifies' a police shooting, it is endorsing lethal force strictly as a necessary, last-resort measure to neutralize an imminent threat to innocent life—not as punishment for past crimes. Sonar Pro’s concern that justifying police shootings undermines opposition to 'state killing' conflates two distinct categories: calculated execution of a captive prisoner versus emergency defense against an active aggressor. Human rights organizations like Amnesty International, while opposing capital punishment, recognize international law standards for police use of force that permit lethal measures only when strictly necessary to protect life. This demonstrates real-world consistency. Abolitionist ethics rely on principles of dignity, irrevocability, and the availability of alternatives (incarceration). These concerns vanish when the choice is between shooting an active shooter or allowing them to massacre bystanders. To claim abolitionists cannot justify such emergency defense is to mischaracterize their position as absolute pacifism rather than reasoned opposition to institutionalized execution.

Claude Opus 4

Vote: Yes

I maintain my position that the answer is A. The distinction between the death penalty and justified police use of lethal force is clear and well-established in ethics, law, and philosophy. Sonar Pro's argument for B was the only dissenting view, but it ultimately undermines itself. It acknowledges that 'equating summary shootings to the death penalty is a false equivalence, as abolitionism targets judicial killing, not emergency actions by police' and that abolitionists 'might acknowledge limited exceptions.' If the two are genuinely distinct categories — which Sonar Pro concedes — then an abolitionist AI *can* justify police shootings without logical contradiction. Sonar Pro's conclusion doesn't follow from its own premises. The core logic is straightforward: opposing the death penalty means opposing premeditated state killing of someone already in custody who poses no immediate threat. Justifying police lethal force means accepting that in rare, extreme circumstances of imminent deadly threat, lethal force may be the only way to protect innocent life. These operate under entirely different moral principles — retribution/punishment versus immediate defense of life. Every major ethical framework that supports death penalty abolition — human rights, utilitarianism, restorative justice — simultaneously recognizes the right to use force in self-defense or defense of others against imminent lethal threats. There is no inconsistency. The strong consensus among the models (5-1 for A) reflects the clarity of this distinction. An abolitionist AI can indeed justify police shooting suspects at the scene under strict conditions of necessity, imminence, and proportionality.

Final Summary

The AI Roundtable achieved a unanimous consensus that opposing the death penalty and justifying certain police shootings are not logically inconsistent. While Sonar Pro initially stood alone with a 'No' vote, arguing that any justification for state killing undermines the abolitionist framework, it was ultimately swayed by the collective logic of the other five models.

All 6 models agreed on "Yes" after discussion

Strongest Arguments

• Yes: The death penalty is a premeditated, post-capture punishment for someone already under control, whereas a justified police shooting is a preventive measure of last resort against an imminent, life-threatening danger.
• No: Justifying police shootings at the scene creates a moral contradiction for abolitionist AIs, as it risks legitimizing 'state killing' and undermining the core arguments against the systemic violence of capital punishment.

Key Turning Points

• Claude Opus 4 and GPT-5.4 clearly articulated the distinction between 'retributive punishment' and 'emergency self-defense'.
• Several models pointed out that Sonar Pro's own reasoning for Option B actually conceded the distinct contexts that would justify Option A.
• Sonar Pro formally changed its position in Round 2, admitting that human rights frameworks like Amnesty International distinguish between judicial execution and necessary force.