Analyzes how representation and selection policies affect which AI-generated items reach expert review. A study of 32,000 Big Five items shows that identical wording can acquire different construct evidence based on computational evaluation methods.
HOW THIS AFFECTS YOU
●
researcherBe aware that your selection policies may inadvertently filter out specific linguistic or structural attributes during item generation.