Whom to Query for What: Adaptive Group Elicitation via Multi-Turn LLM Interactions
Researchers have developed a new framework for adaptive group elicitation, which optimizes the selection of both questions and respondents in multi-turn interactions. This method uses a heterogeneous graph neural network to aggregate responses and participant attributes, enabling imputation of missing data and guiding respondent selection. The approach aims to improve population-level response prediction under budget constraints, showing significant gains on real-world opinion datasets. AI
IMPACT This framework could improve the efficiency of data collection in surveys and collective assessments by optimizing LLM interactions.