The goal of this study is to use a large language model, such as GPT-3, to generate responses from hard-to-reach elite members to multiple-choice questions. The most relevant or data-consistent responses can then be used to infer potential responses from selected elite members and apply them in...