Beyond Polarized Attitudes: Validating a Multidimensional Instrument for Comparative Evaluation of AI and Human Decision-Making

We conducted three validation studies on the integration of artificial intelligence (AI) into decision‑making, a topic that has prompted debate about AI versus human evaluators. We developed and validated a multidimensional scale assessing attitudes toward AI across three serial dimensions: perceived decision‑making process (Process), outcomes generated (Outcome), and acceptance of AI leadership (Leadership). Exploratory factor analysis confirmed a robust three‑factor structure with strong internal consistency and convergent validity. Confirmatory factor analysis demonstrated distinct predictive patterns: in logical tasks, favorable evaluations of AI outcomes drove adherence to AI recommendations; in moral tasks, acceptance of AI leadership was the critical predictor, with AI‑sourced utilitarian recommendations exerting attenuated influence compared to human sources. Test‑retest reliability over two weeks proved excellent. These results support a comparative, multidimensional model of human‑AI interaction and guide context‑specific design of AI decision‑support systems. Our instrument offers a detailed framework for understanding human‑AI interactions across diverse decision contexts.