ModelCompare experiment finds 12 AI models largely agree humans prioritize belonging over truth

Claude Opus 5.5 diverged in one of its three responses, suggesting humans are kinder and more decent than they believe, though it echoed the majority view in its other two answers. Mistral Large 4 called human consciousness an evolutionary malfunction. The researchers also prompted six AI image generators; ten of twelve images depicted humans as technology-obsessed consumers harming the planet, a theme absent from the text models' answers. The article notes the experiment is not scientific proof: the prompt was deliberately provocative, responses were grouped manually, the sample was small, and models may simply reproduce patterns from training data about how humans describe themselves.