LLMs Can Better Capture Human Judgments--With the Right Prompts

The study presents prompting strategies that enhance large language models' (LLMs) ability to capture human judgments more accurately. By utilizing techniques such as prompting for standard deviations and response proportions, the researchers improved alignment with human responses across moral scenarios and beliefs from diverse datasets. This work highlights the importance of effective elicitation methods in improving AI-human interaction and suggests that refining the questions posed to LLMs can lead to more reliable outputs, which is crucial for practitioners aiming to integrate LLMs in applications requiring nuanced human judgment.