Artificial Intelligence Attitudes Inventory (AIAI): Development and Validation Using Rasch Methodology

Abstract

With the rapid advancements in artificial intelligence (AI), it is vital to develop psychometrically sound measures of public attitudes toward this technology. The present study aimed to refine a pool of candidate items to create a concise yet robust inventory for assessing attitudes toward AI. A total of 96 items—drawn from reworded robot-related scales and newly developed items to reflect AI-specific themes—were administered to a sample of 604 adults from the general population of the United States (age range: 18–89 years; 48% male). Iterative Rasch analysis was used to reduce the number of items while ensuring psychometric robustness, applying multiple criteria for item selection including fit residuals, differential item functioning (DIF), and conceptual clarity. The resulting scale, named the Artificial Intelligence Attitudes Inventory (AIAI), consists of two 8-item subscales measuring positive and negative attitudes toward AI. Analyses revealed that these subscales are distinct constructs rather than opposites on a single continuum, and they are only weakly related to psychological distress. The AIAI provides a concise yet comprehensive measure of positive and negative attitudes toward AI that can be efficiently administered alongside other measures. The findings underscore the multifaceted nature of public perceptions of AI and highlight the need for further research into the profiles and determinants of these attitudes. As AI continues to shape our world, the AIAI offers a valuable tool for understanding and monitoring public sentiment toward this transformative technology.