Multi-Agent AI Framework Detects OCEAN Personality from Life Narratives
Researchers use fine-tuned sub-agents to extract Big Five traits from long texts.
A team of researchers (Hussain et al.) from academia has introduced a fine-tuned multi-agent framework for detecting OCEAN personality traits (Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism) from life narratives. The approach addresses a key challenge in text-based personality inference: traits are latent, context-dependent, and subtly expressed across long narratives. Single LLMs often carry latent personality-like biases from pretraining, making inferences inconsistent.
The framework uses multiple sub-agents, each conditioned through masked language modeling and psychometric supervision to adopt a specific perspective (high, low, or neutral) for each trait. A judge LLM then aggregates and compares these diverse outputs to generate final trait predictions, effectively capturing multiple complementary viewpoints while mitigating individual model biases. The system was evaluated on a life narrative dataset through quantitative and qualitative experiments, including baselines, ablations, and inference quality analyses. Results demonstrate that this multi-agent reasoning approach provides more accurate and interpretable personality assessments than single-model methods, offering a scalable solution for text-based psychological profiling.
- Sub-agents are fine-tuned to adopt high, low, or neutral perspectives for each OCEAN trait using masked language modeling and psychometric supervision.
- A judge LLM aggregates outputs from multiple sub-agents to reduce single-model bias and capture complementary viewpoints.
- Evaluated on life narrative dataset with baselines, ablations, and inference quality analyses showing improved accuracy and interpretability.
Why It Matters
Enables more accurate, bias-reduced personality assessment from unstructured text for psychology, HR, and research.