Generative Artificial Intelligence and University Students' Critical and Higher-Order Thinking Outcomes: A PRISMA-Based Systematic Review and Thematic Analysis
TL;DR
This review contributes a dimension-level synthesis of generative AI and higher-order thinking, identifies the pedagogical conditions associated with favorable and unfavorable outcomes, and documents the heterogeneity of outcomes across studies, offering an evidence-informed framework for educators, curriculum designers, and policymakers seeking to use generative AI as a cognitive scaffold rather than a cognitive replacement.
Abstract
The rapid integration of generative artificial intelligence tools such as ChatGPT, Gemini, and Copilot into higher education has raised urgent questions about their effects on students' higher-order cognitive abilities. While these tools promise unprecedented access to information and cognitive support, growing concerns have emerged regarding the potential erosion of critical thinking skills among university students who adopt them uncritically. Despite increasing scholarly attention to artificial intelligence in education, systematic evidence on how generative AI relates to distinct dimensions of higher-order thinking remains limited. This systematic literature review, guided by the PRISMA 2020 reporting guideline, aims to identify, appraise, and synthesize empirical evidence on the relationship between generative AI use and university students' analytical, evaluative, and creative thinking outcomes, using the higher-order cognitive processes of Bloom's Revised Taxonomy as an organizing framework. A comprehensive search across Scopus, Web of Science, ERIC, and Google Scholar yielded 847 initial records, of which 42 empirical studies published between 2020 and 2024 met the inclusion criteria after screening. Thematic synthesis revealed heterogeneous outcomes: analytical thinking outcomes were generally positive under structured prompting conditions, evaluative thinking appeared most vulnerable in the absence of deliberate pedagogical design, and creative thinking outcomes were most frequently reported as positive. Across themes, the quality of pedagogical integration was the condition most frequently associated with favorable outcomes. This review contributes a dimension-level synthesis of generative AI and higher-order thinking, identifies the pedagogical conditions associated with favorable and unfavorable outcomes, and documents the heterogeneity of outcomes across studies, offering an evidence-informed framework for educators, curriculum designers, and policymakers seeking to use generative AI as a cognitive scaffold rather than a cognitive replacement.
