The double-edged sword of generative AI dependence: scale development and its curvilinear impact on employee innovation

The integration of generative artificial intelligence (GenAI) into professional work has given rise to a novel phenomenon, employee dependence on GenAI, which remains conceptually underdeveloped and lacks validated measurement instruments. This research program aimed to conceptualize GenAI dependence, develop a psychometric scale, and examine its curvilinear impact on employee innovation performance and underlying mechanisms. A multi-study mixed-methods design was employed. Study 1 used an abductive grounded-theory approach with 24 knowledge workers (21 initial interviews and 3 additional interviews conducted to assess theoretical saturation). Study 2 developed a measurement scale validated through three survey rounds ( N = 233, 271, and 269) using CITC , EFA , and CFA , supplemented by content-validity indices, model comparisons (correlated three-factor, second-order, bifactor, and one-factor models), omega coefficients, and HTMT ratios. Study 3 used a three-wave time-lagged survey ( N = 342) to test an inverted U-shaped model, with thriving at work as the mediator and mindfulness as the moderator, analyzed via hierarchical regression and the MEDCURVE macro, with turning-point analyses, dimension-level analyses, and robustness checks. GenAI dependence comprises three interrelated and potentially progressive dimensions: instrumental, psychological, and cognitive. The 20-item GenAID scale demonstrated preliminary evidence of internal structure, reliability, and discriminant validity, and model comparisons supported the defensibility of a total score. GenAI dependence showed an inverted U-shaped relationship with innovation performance, mediated by thriving at work; the turning point lay within the observed data range, with significant positive and negative slopes at one standard deviation below and above the mean. Mindfulness moderated this effect, flattening the downward slope and shifting the turning point rightward. Analyses of the three dimensions separately indicated significant negative curvature within each dimension and a turning point that occurs progressively earlier from instrumental through psychological to cognitive dependence, although the complete set of inverted-U conditions was not uniformly satisfied for every dimension. This research provides an initial conceptualization and a preliminary instrument for assessing GenAI dependence in organizations. The findings suggest a double-edged sword effect on innovation performance, and they inform the management of AI integration and the design of candidate interventions, such as mindfulness-based programs, whose effectiveness remains to be tested in future research.