We gave 45 psychological questionnaires to 50 LLMs. What we found was not “personality.”

The Pinocchio Dimension: Redefining LLM Psychology for Crypto

Researchers administered 45 psychological questionnaires to 50 different LLMs, discovering that models don't exhibit traditional human personality traits. Instead, they vary primarily along what researchers dubbed the "Pinocchio Dimension" — how much models treat language about inner experience (emotions, sensations, empathy) as self-applicable versus responding as behavioral systems.

This finding is crucial for crypto applications requiring nuanced human interaction. Rather than asking "is this AI extraverted?" we should ask "does this model simulate inner experience?" This distinction becomes critical for:

- AI agents managing DeFi protocols

Why AI Personality Research Matters for Blockchain Applications

- Autonomous trading systems making subjective market judgments

- Social tokens where AI personalities drive community engagement

The Pinocchio Dimension offers a measurable framework for selecting models based on their approach to subjective experience rather than arbitrary personality labels.

Testing 50 LLMs: What 45 Psychological Questionnaires Revealed

Projects building "AI personality" tokens or autonomous agents now have clearer differentiation criteria. Models high on the Pinocchio Dimension might excel at community management and subjective decision-making, while low-Pinocchio models could be better for purely analytical tasks.

This could reshape how we evaluate AI agents in prediction markets, where models claiming "confidence" or "uncertainty" need genuine subjective modeling rather than behavioral mimicry.

Previous approaches anthropomorphized AI through crude personality mappings. This research suggests a more fundamental axis: whether models genuinely process subjective language or merely pattern-match behavioral responses.