From AI reliance to cognitive control: A dual-process comparison of early-career and mid-career academics in AI-enhanced scholarly performance

Description This dataset contains survey responses and externally evaluated academic writing performance from 480 university academics in Vietnam, collected between May 2024 and October 2025. The dataset supports analysis of how artificial intelligence engagement relates to scholarly performance in research and writing activities, with attention to cognitive, ethical, and capability-related mechanisms. Participants are classified into two career-stage groups: early-career academics with fewer than seven years of experience and mid-career academics with seven to fifteen years of experience. This classification allows examination of how differences in experience shape engagement with artificial intelligence and influence scholarly performance outcomes. The dataset includes seven constructs measured using multi-item scales, covering academic identity, epistemic agency, research self-efficacy, ethical AI awareness, AI use in research and writing, perceived AI use effectiveness, and AI-enhanced scholarly performance. Survey variables are measured using a 7-point Likert scale. Scholarly performance is assessed through external evaluation of literature review reports using a structured rubric focusing on argument quality, synthesis of literature, conceptual clarity, writing coherence, and use of evidence. This design separates predictor variables from outcome measurement and reduces common method bias. The dataset is suitable for structural equation modeling, mediation analysis, and multigroup comparison. Missing values are coded as 999. All data have been anonymized. Detailed variable definitions and item descriptions are provided in the accompanying codebook file.