A Method for Modeling Distributions of Responses Using Personality-Based Prompting

Advantages

  • Predict real-world behaviors with calibrated, human-centric AI personas.
  • Improve accuracy and fairness in surveys, assessments, and decision tools.
  • Reduce pilot costs through scalable, distribution-based forecasting.
  • Provide transparent insights companies can audit and trust.

Summary

Organizations using surveys, cognitive assessments, or market research face a major limitation: they struggle to forecast how diverse populations will respond from small pilot groups. Traditional models require large samples, assume rigid patterns, and often fail to capture the full range of human responses. This leads to inaccurate predictions, bias in assessments, and costly iterations.

This invention introduces Personality-Based Prompting (PBP), which models populations as interpretable mixtures of AI “personas,” aligned with the Big Five personality traits. By combining persona-specific responses with an evolutionary algorithm, PBP generates calibrated distributions that mirror real human behavior. The approach has outperformed baselines in early pilots, offering organizations a scalable, accurate, and auditable way to predict outcomes at population scale.

Image Description: Personality prompting uses weighted AI personas to align LLM reasoning with human System 1 and System 2 responses.

Desired Partnerships

  • License
  • Sponsored Research
  • Co-Development
Patent Information: