This research introduces the Machine Philosophies of Human Nature Scale (MaPHNS), a standardized psychological assessment of how LLMs perceive human nature and trustworthiness.

Key findings:

• LLMs demonstrate distinct philosophical biases about human nature that affect their outputs
• Different models show varying degrees of optimism, cynicism, and trust toward humans
• These biases significantly impact LLMs' ethical reasoning and decision-making
• The framework enables better evaluation of potential security and ethical risks in AI systems

For security professionals, understanding these inherent biases is crucial as they shape how AI systems interpret human intentions and make trust-based judgments in sensitive applications.

Measurement of LLM's Philosophies of Human Nature