This research introduces Generative Psychometrics for Values (GPV), a novel approach for measuring human and AI value systems using large language models.

• Creates a data-driven paradigm for evaluating values in both humans and AI systems
• Grounds measurement in text-revealed selective perceptions
• Establishes new methods to assess value alignment between humans and AI
• Offers implications for creating safer, more aligned AI systems

Security Relevance: By providing frameworks to measure AI value systems, this research advances our ability to develop value-aligned AI—a critical component of AI safety and security governance. Understanding how to measure values helps identify potential misalignment before deployment.

Measuring Human and AI Values Based on Generative Psychometrics with Large Language Models