Confidence-Driven Augmentation of Synthetic Survey Data

• Unité de recherche: INNOSUISSE
• Numéro de projet: 125.426 IP-SBM
• Titre du projet anglais: Confidence-Driven Augmentation of Synthetic Survey Data

Mots-clé (Anglais)

Social sciences & business management, New media

Résumé des résultats (Abstract) (Anglais)

Market research remains costly and slow. This research combines LLM-based synthetic survey data with human data where most needed. Compared to human random samples, the method cuts paid human samples by 20% while preserving accuracy, enabling affordable, timely market research and political polling.