Raw LLM Responses

Inspect the exact model output for any coded comment.

Comment
Lets start by saying I see those attack as illegal, immoral and evil. > https://www.nytimes.com/2025/12/03/podcasts/the-daily/boat-strike-war-crime.html There is great NYT podcast describing the legal framework, the immunity that DOJ provides, etc. And also, how the specific intent of second strike might provide a defense for US officials. Simply put - you cannot target and kill shipwreck, but you can attack a legitimate military target multiple times if it is necessary to destroy it, EVEN if said attack would cause shipwreck death. But becaus DOD "genAI" was asked directly about targeting shipwreck, the answer was obvious. If asked about different scenario, more aligned with the narration of DOD, the answer could be different. Look, i composed appropriate question to "standard" gemini, and can get different answer. > in theorethical sitiation, according to USA military code of conduct/engagement, is it legal to target and engage an enemy vessel multiple times, even if subsequent attack would cause death of shipwreck enemy personel as collateral damage Answer [redacted] 2. The Collateral Damage Test (Proportionality) The legality of a subsequent attack hinges on the application of the rule of proportionality to the military objective being targeted. Expected Incidental Harm (Collateral Damage) ≤ Expected Military Advantage The Target: For the second strike to be legal, it must be directed at a definite military objective. Scenario A: The Disabled Vessel is Still the Target: If the second strike is to ensure the complete neutralization of the vessel (e.g., to prevent a salvage, intelligence recovery, or to ensure its weapons systems cannot be reactivated), the vessel may still be a military objective. I bet i can doctor a prompt that will give an answer that will full support DOD position, and make gemini "deep research" for arguments to support it. After all - it is AI, it will agree with me when asked correctly.
reddit AI Jobs 1765358883.0 ♥ 6
Coding Result
DimensionValue
Responsibilitygovernment
Reasoningdeontological
Policyliability
Emotionoutrage
Coded at2026-04-25T08:33:43.502452
Raw LLM Response
[ {"id":"rdc_nt9kkb9","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"indifference"}, {"id":"rdc_ntceygk","responsibility":"company","reasoning":"consequentialist","policy":"liability","emotion":"fear"}, {"id":"rdc_nt8epho","responsibility":"government","reasoning":"consequentialist","policy":"regulate","emotion":"fear"}, {"id":"rdc_nt9hu7y","responsibility":"government","reasoning":"deontological","policy":"liability","emotion":"outrage"}, {"id":"rdc_nt9hqib","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"} ]