Raw LLM Responses

Inspect the exact model output for any coded comment.

Comment
I asked ChatGPT to explain why ordinary people could end up supporting a person like Hitler and to format the answers into a short story that is easy to understand. I picked the Bearded Dragon as the narrator. HOW ORDINARY CREATURES END UP SUPPORTING DANGEROUS LEADERS “Now then,” says the Bearded Dragon, adjusting his beard. “Many young creatures believe that the followers of dangerous leaders must all be monsters.” He shakes his head slowly. “This is incorrect, and also dangerously comforting.” The room grows quiet. THE SLOW-BOILING POT “Societies rarely leap directly into catastrophe.” The dragon lifts one claw. “Instead, changes arrive gradually.” A small rule changes. A minority is blamed for a problem. A newspaper repeats a story often enough that it begins to feel true. “If the temperature rises slowly,” he says, “many creatures fail to notice the water getting hot.” BELONGING IS POWERFUL “Creatures are social,” the dragon explains. “Most individuals want to belong to their group. When the group begins repeating an idea, disagreeing becomes emotionally costly.” He glances around the room. “Even intelligent creatures may stay silent if speaking up risks rejection.” PROPAGANDA SIMPLIFIES THE WORLD “Dangerous leaders also simplify complicated problems.” He gestures with a claw. “The world is confusing. Economies fail. Wars happen. Change creates anxiety.” Propaganda offers a tempting solution: • One simple explanation • One clear enemy • One heroic leader The dragon sighs. “It is a very seductive story.” FEAR OVERRIDES MORAL INSTINCTS The dragon’s voice grows softer. “When people feel frightened enough, they will tolerate actions they would normally reject.” Rights disappear slowly. Cruel policies are justified as necessary. “Fear convinces people that harsh actions are temporary and justified.” The dragon pauses. “History shows that they rarely remain temporary.” THE BYSTANDER PROBLEM Finally, the Bearded Dragon raise
reddit AI Harm Incident 1772701903.0 ♥ 5
Coding Result
DimensionValue
Responsibilityunclear
Reasoningunclear
Policyunclear
Emotionapproval
Coded at2026-04-25T08:33:43.502452
Raw LLM Response
[ {"id":"rdc_o8r8hi8","responsibility":"ai_itself","reasoning":"deontological","policy":"liability","emotion":"outrage"}, {"id":"rdc_o8uwgfo","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"}, {"id":"rdc_oacncm8","responsibility":"developer","reasoning":"virtue","policy":"industry_self","emotion":"mixed"}, {"id":"rdc_o8ql1rq","responsibility":"developer","reasoning":"consequentialist","policy":"regulate","emotion":"outrage"}, {"id":"rdc_o8r1zxv","responsibility":"unclear","reasoning":"unclear","policy":"unclear","emotion":"approval"} ]