Validation of the generative artificial intelligence dependency scale among Chinese college students.
TL;DR
The Chinese version of the Generative Artificial Intelligence (AI) Dependency Scale is a useful assessment tool for determining generative AI dependency among Chinese university students because it shows strong validity, reliability, and measurement invariance.
Abstract
Objective
To update the Generative Artificial Intelligence (AI) Dependency Scale and assess its reliability and validity among college students in China.
Methods
The Chinese version of the Generative Artificial Intelligence (AI) Dependency Scale was administered to 986 university students. The Generative AI Addiction Scale and the Chinese Big Five Personality Questionnaire Short - Form were used as comparators. The sample was randomly divided into two groups: 491 participants (Sample 1) for item analysis and exploratory factor analysis, and 495 participants (Sample 2) for confirmatory factor analysis. A subsample of 324 participants from both groups underwent retesting after 4 weeks.
Results
Exploratory factor analysis revealed three factors that cumulatively explained 76.819% of the total variance. The three - factor model of the Generative AI Dependency Scale (cognitive distress, negative consequences, avoidance tendency) was validated among Chinese university students. Confirmatory factor analysis indicated a good model fit for the three - factor model (χ2/41 = 2.29, P < 0.001, RMSEA = 0.05, SRMR = 0.03, CFI = 0.97, TLI = 0.96); the total score of the Generative AI Dependency Scale showed significant positive correlations with generative AI addiction, openness, agreeableness, neuroticism, extraversion, and conscientiousness (r = 0.17-0.75, P < 0.01); the internal consistency coefficient of the scale was 0.932, with coefficients ranging from 0.804 to 0.948 for each dimension' s mean score; the test - retest reliability (ICC) for the total scale was 0.851, with ICCs for each dimension' s mean score ranging from 0.628 to 0.788 (P < 0.001); based on a three - factor model, cross - gender and cross - grade measurement invariance was assessed for form invariance, weak invariance, strong invariance, and strict invariance. ΔCFI and ΔTLI values were both <0.01, while ΔRMSEA values were all <0.015, indicating that the generative AI dependency scale possesses measurement invariance across genders and academic years.
Conclusion
The Chinese version of the Generative Artificial Intelligence (AI) Dependency Scale is a useful assessment tool for determining generative AI dependency among Chinese university students because it shows strong validity, reliability, and measurement invariance.
