AI

Behavioral History Outperforms Descriptions of the Person for LLM Synthetic Personas

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 06/10/2026, 04:18:24 UTC.

Briefings

A 2026 study using 845 US adults found that LLM synthetic personas significantly outperform static descriptions of the person in predicting future survey responses. While descriptions alone explain only 53-67% of human variation, adding behavioral history restores this to near-human levels. Individual-level prediction improved from 7-12% with descriptions to 28% with behavioral history, showing the past answers add more to prediction than a description of who the person is.

Key points

• LLM synthetic personas using behavioral history achieve 28% individual-level informedness, compared to 7-12% for descriptions alone.
• Behavioral history restores between-person variation to human levels, whereas descriptions only recover 53-67% of that variation.

Context and limitations

Data covers 17 US demographic groups with a two-wave panel of 845 adults; results are limited to the 14 behavioral biases measured and do not generalize to all survey contexts.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 06/10/2026, 05:02:47 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeig5tsocdmv2sl32bsthdibrh5ozwi6n7654t4dlghcvcrgwkfaswa