A Multidimensional Psychometric Scale for Measuring AI Dependency in Academic Research

Abstract

As artificial intelligence (AI) becomes embedded in academic workflows, concerns are increasing about researchers developing psychological and behavioral dependence on AI systems. Although AI tools can significantly enhance research efficiency and analytical capabilities, excessive or uncritical reliance may lead to cognitive disengagement, emotional over attachment, and maladaptive academic practices. Yet, the literature still lacks a validated instrument that can systematically assess AI dependency within scholarly contexts. This study addresses this gap by developing a multidimensional scale that draws on the affect, behavior, and cognition (ABC) model as the foundational structure of psychological response, incorporates basic psychological needs theory (BPN) to explain motivational dynamics, and is ultimately grounded in the I-PACE model, which provides a comprehensive framework for understanding the development of dependency-related behaviors. The scale was developed using a rigorous mixed-methods design implemented across a five-stage sequence that encompassed systematic item generation, expert review for content adequacy, and extensive empirical evaluation with a large academic sample. Findings from exploratory and confirmatory factor analyses converged to support a stable three-factor structure consisting of fifteen theoretically coherent items. The instrument demonstrated strong psychometric integrity, including high internal consistency, robust convergent and discriminant validity, and meaningful predictive validity, particularly in relation to attitudes linked to academic research misconduct. By establishing a validated measure of AI dependency within research contexts, this study advances current debates in digital ethics and responsible AI use and offers an empirically grounded tool for strengthening research integrity practices in higher education.