Nicolás Rueda-Díaz

Artificial intelligence (AI) is commonly used in university academic work, yet existing instruments mainly assess AI literacy, readiness, or levels of permitted use rather than the critical-thinking processes students activate when working with AI. This study developed the Critical Thinking with AI Scale for Students (CTAI-S) and provided initial psychometric evidence for its use. An initial 15-item questionnaire was administered to 271 university students, and one item was removed after item analysis. Parallel analysis favored one predominant factor (KMO = 0.927), and a theory-guided three-factor representation was also examined. Ordinal confirmatory factor analyses (WLSMV; n = 257) yielded scaled CFI = 0.964 and RMSEA = 0.072 for the one-factor model, and CFI = 0.977 and RMSEA = 0.059 for the equivalent correlated-three-factor and second-order models; robust indices were less favorable, and latent correlations were very high (0.85–0.88). Total-score reliability was high (ω = 0.898), whereas the third grouping showed low reliability (ω = 0.557). Mean responses were above the scale midpoint (M = 3.97), with Reflective Judgment and Self-Regulation being highest (M = 4.07). The CTAI-S is best interpreted as a global self-report measure of critical thinking in AI-mediated academic tasks; independent validation is needed before dimensional or diagnostic applications.

Artificial intelligence (AI) is transforming higher education, although most research addresses its integration in terms of frequency of use or technological acceptance, without examining how it translates into specific curricular and instructional decisions. That is why this study has a dual aim: to develop and validate the AI Instructional Design Questionnaire for Critical Thinking (AID-CTQ) and to analyze how university faculty integrate AI into instructional design practices in higher education. The sample included 144 faculty members from a university in Madrid, selected by convenience. Exploratory and confirmatory factor analyses of the questionnaire supported a three-factor structure: Activity Design (F1), Critical Thinking Assessment (F2), and Self-Regulation and Reflection (F3). The final 12-item model shows good model fit (CFI = 0.98, TLI = 0.98, RMSEA = 0.05, SRMR = 0.05) and adequate overall reliability (α = 0.86). At the item level, responses related to assessment and reflective practices showed consistently high agreement, whereas items linked to activity design displayed greater variability. Faculty members with more than 10 years of experience obtained significantly higher scores, indicating that the educational value of AI depends less on the tools used and more on the quality of instructional decisions. Reported use of AI was high, with ChatGPT and Copilot being the most frequently used tools. Overall, the findings indicate that the integration of AI in higher education is evolving from predominantly instrumental uses toward more pedagogical and curriculum-oriented forms of implementation. Accordingly, the educational value of AI lies less in the tool itself than in the quality of the instructional decisions through which it is meaningfully embedded in the curriculum.