October 12, 2025Open Access

Disentangling Interaction and Bias Effects in Opinion Dynamics of Large Language Models

View Full Paper

Ask AI

Bookmark

Opion trajectories in large language models swiftly converge to a shared attractor, influenced by biases.
The Bayesian framework identifies and quantifies topic, agreement, and anchoring biases affecting the models.
Fine-tuning models on opinionated statements shows corresponding shifts in opinion attractors, hinting at model behavior.
The approach reveals significant differences among large language models, emphasizing their use as proxies for human decision-making.

Discussion

No takes yet. Share an insight, caveat, or question.

Bayesian analysis quantifies biases affecting opinion dynamics in large language models, suggesting implications for human behavior.

Brockers et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark