Stories of Your Life as Others: A Round-Trip Evaluation of LLM-Generated Life Stories Conditioned on Rich Psychometric Profiles

This paper introduces a clever way to test if large language models (LLMs) can truly capture human personality in generated text. Instead of just asking an LLM to self-report its 'personality' (which can be superficial), the authors use a round-trip evaluation: They feed real psychometric profiles from 290 humans into LLMs to generate first-person life stories, then have independent LLMs read those stories and recover the original personality scores. The key question: Does the generated narrative encode personality robustly enough to decode back to the original traits?

The results are impressive—recovery accuracy hits a mean correlation of r=0.750, close to human test-retest reliability (the 'ceiling' of how consistent human personality scores are over time). This holds across 10 generator LLMs and 3 scorer LLMs from 6 providers, showing it's not model-specific. Why does this matter? It proves LLMs, pretrained on human text, have internalized the deep links between personality and language (the lexical hypothesis), enabling realistic behavioral simulation.

Imagine personality as a compressed code in language. The paradigm tests if LLMs can encode real traits into narratives and others can decode them:

1. Input: Rich profiles from the PARSEL dataset (297 participants' HEXACO-60 scores, plus trust, social anxiety, etc.).
2. Stage 1: Claude Opus converts profiles + bio/appearance facts into immersive second-person prompts (e.g., 'You are someone who...').
3. Generation: 10 LLMs (GPT-4o, Llama-3, etc.) produce ~2000-word life stories in first-person, mimicking McAdams' Life Story Interview (LSI)—narratives revealing personality through life themes.
4. Recovery: 3 independent LLMs score the stories back to HEXACO traits.

This 'round-trip' avoids self-report biases and tests if conditioning creates psychometrically valid representations of individual differences, not just averages.

• Psychometric Recovery: r=0.750 mean correlation (85% of human ceiling). Scorers counteract 'alignment defaults' (e.g., overly agreeable text) to match originals. Robust to generators/scorers.
• Behavioral Differentiation: Narratives show personality-linked patterns matching real human conversations (e.g., extraverts use more positive emotion words; emotional reactivity replicates).
• Nine of ten coded text features (e.g., analytical thinking, confidence) correlate with PARSEL convos.

This validates immersive persona prompting > basic conditioning, using real data (not synthetic) and HEXACO (including Honesty-Humility).