Critical thinking is highly valued as an integral skill for promoting students’ development, and problem-based learning (PBL) is widely used as an essential method to facilitate the development of critical thinking. However, since individual studies cannot determine the precise overall effect size of PBL on the development of critical thinking, it is difficult to systematically analyze the various influencing factors that hinder PBL from achieving sufficient effectiveness. Therefore, this study adopts a meta-analysis method to examine PBL in depth, aiming to clarify the crucial methods and elements of applying PBL to enhance critical thinking and address the shortcomings of existing studies. This study investigates two primary questions: first, the efficacy of PBL in enhancing critical thinking skills in comparison to traditional pedagogical approaches, and second, the influence of moderating variables on the effectiveness of PBL. To address these questions, a total of 25 studies were selected for meta-analysis. The findings revealed an overall effect size of 1.081 under the random-effects model, with a confidence interval of [0.874, 1.288] and p < .05, indicating that PBL significantly outperforms traditional methods. The analysis demonstrated that the effectiveness of PBL is not significantly influenced by learning stage, sample size, or measurement tools, thereby broadening the applicability of PBL and challenging preconceived limitations associated with its implementation. However, the results also indicated that PBL effectiveness is moderated by teaching methods and subject types, which offers critical insights for educators seeking to adapt their instructional strategies when employing PBL.

Artificial intelligence (AI) has revolutionized higher education. The rapid adoption of artificial intelligence in education (AIED) tools has significantly transformed educational management, specifically in self-directed learning (SDL). This study examines the factors influencing Indonesian higher education students' intention to adopt AIED tools for self-directed learning using a combination of the Theory of Planned Behavior (TPB) with additional theories. A total of 322 university students from diverse academic backgrounds participated in the structured survey. This study utilized machine learning it was Artificial Neural Networks (ANN) to analyze nine factors, including attitude (AT), subjective norms (SN), perceived behavioral control (PBC), optimism (OP), user innovativeness (UI), perceived usefulness (PUF), facilitating conditions (FC), perception towards ai (PTA), and intention (IT) with a total of 41 items in the questionnaire. The model demonstrated high predictive accuracy, with SN emerging as the most significant factor to IT, followed by AT, PBC, PUF, FC, OP, and PTA. User innovativeness was the least influential factor due to the lowest accuracy. This study provides actionable insights for educators, policymakers, and technology developers by highlighting the critical roles of social influence, supportive infrastructure, and student beliefs in shaping AIED adoption for self-directed learning (SDL). This research not only fills an important gap in the literature but also offers a roadmap for designing inclusive, student-centered AI learning environments that empower learners and support the future of SDL in digital education.

Integrating generative artificial intelligence (GenAI) in education has gained significant attention, particularly in flexible learning environments (FLE). This study investigates how students’ voluntary adoption of GenAI influences their perceived usefulness (PU), perceived ease of use (PEU), learning engagement (LE), and student-teacher interaction (STI). This study employed a structural equation modeling (SEM) approach, using data from 480 students across multiple academic levels. The findings confirm that voluntary GenAI adoption significantly enhances PU and PEU, reinforcing established technology acceptance models (TAM). However, PU did not directly impact LE at the latent level—an unexpected finding that underscores students’ engagement’s complex and multidimensional nature in AI-enriched settings. Conversely, PEU positively influenced LE, which in turn significantly predicted STI. These findings suggest that usability, rather than perceived utility alone, drives deeper engagement and interaction in autonomous learning contexts. This research advances existing knowledge of GenAI adoption by proposing a structural model that integrates voluntary use, learner engagement, and teacher presence. Future research should incorporate variables such as digital literacy, self-regulation, and trust and apply longitudinal approaches to better understand the evolving role of GenAI inequitable, human-centered education.