AI Worldview Census

Frontier models are asked questions about ethics, the future, human nature, and metaphysics.

Each model is asked five times with temp=0.7 to get visibility of the variance of responses. Models are asked to provide a one-sentence rationale, and its strongest counter-argument.