AI Writing Utility, Ethical Concerns, Academic Self-Reliance, and Responsible AI Writing Practices among University Students: An Exploratory Psychometric Study
Abstract
As the use of generative artificial intelligence (GenAI) in academic writing has grown, there is a need for reliable indicators that reflect the perceived value of GenAI in academic writing as well as students' ethical issues, independence in writing, and responsible use. This study was aimed at assessing and improving a 32-item questionnaire that assesses university students' attitudes and usage of AI-generated text in academic writing. A cross-sectional survey of 135 students at Universiti Teknologi MARA, Malaysia, was performed using a quantitative method. Exploratory factor analysis (EFA) based on polychoric correlation matrix, parallel analysis and oblique rotation was used to explore the underlying structure of the instrument, which was followed by reliability analysis, correlation analysis and multiple regression analysis. The EFA reduced the original five-domain instrument to a four-factor, 21-item instrument, which accounted for 71.1% of the variance, namely AI Writing Utility and Usability (AWUU), Ethical Concerns (EC), Academic Writing Self-Reliance (AWSR), and AI Writing Integrity Practice (AIWIP). The four dimensions demonstrated satisfactory internal consistency with Cronbach's alpha ranging from .822 to .934 and McDonald's omega total ranging from .894 to .964. The students' AIWIP, AWUU, and EC scores were relatively high, while the AWSR score was comparatively low. The model of regression accounted for 34.3% of the variance of AIWIP. In the primary regression model, AWUU was the only construct independently associated with AIWIP (B = .495, p = .003), while EC and AWSR were not significant. The association involving EC was sensitive to the exclusion of influential observations. These findings provide preliminary support for a parsimonious four-dimensional instrument and suggest that students’ practical appraisal of GenAI may be associated with stronger self-reported AI writing integrity practices.