Students everywhere are using new computer programs to help them learn. But many of these tools are black boxes. You put in data and get an answer, but you do not know how the computer got there. This lack of transparency can make students feel confused or unsure about what they are learning. A new analysis looked at a different kind of tool called explainable artificial intelligence, or XAI. These tools try to show you the steps they took to reach a conclusion. This review asked if these clearer tools actually help students learn better or think more clearly. The researchers looked at data from over 62,000 participants across higher education institutions worldwide. This is a huge group of students from many different schools and countries. The goal was to see if these tools match up with how human brains work. Specifically, they checked if the AI matched up with three main areas of thinking. These areas are how hard a task feels to your brain, how well you focus and hold information in your mind, and how you understand your own feelings while learning. The results were very clear. The analysis found a positive connection between the AI tools and these thinking skills. The numbers showed a strong link. This means the AI does not just guess randomly. It aligns with the way students naturally process information. The consistency across different groups of students was high. This suggests the tools work well no matter where the students are from. The researchers also checked the quality of their math models. The numbers showed the models fit the data very well. This gives confidence that the findings are real and not just a lucky guess. There were no reports of safety problems because this was a computer analysis, not a drug trial. However, there is one important thing to remember. The study showed a connection, but it did not prove that using these tools causes better thinking. The students were already using these tools in their classes. The researchers could not change the past to see if the tools made the difference. This is a common limit in studies that look at existing technology. You should not assume that buying these tools will instantly fix learning problems. The tools are valid, meaning they measure thinking correctly, but they are just one part of a complex classroom environment. Teachers and students should be cautious. They should not expect these tools to solve every educational challenge on their own. The best approach is to use these tools as helpers while keeping a close eye on how they are used. The future of learning depends on technology that students can trust and understand. This review gives a good sign that current AI tools are on the right track.
Explainable AI shows positive correlation with cognitive psychology domains in higher educationNew AI tools show strong links to how students think and learn
AI-generated summary of the cited source, checked by automated accuracy review. How we work
This meta-analysis examined the relationship between explainable artificial intelligence (XAI) and cognitive psychology domains within higher education settings. The study included 62,000 participants from higher education institutions worldwide, making it a large-scale synthesis of existing research. The primary analysis focused on the correlation between XAI and three principal cognitive psychology domains: cognitive load and neural efficiency, attention and working memory, and metacognition and affective interaction. The comparator was not reported, as this was a correlational meta-analysis rather than a comparative trial.
The main results showed positive pooled correlations across these domains, with effect sizes ranging from r = 0.27 to 0.36 (p < 0.001; 95% CrI 0.27, 0.38). The global hypermean association centered on a positive association (r ≈ 0.32). Consistency across clusters was demonstrated with a mean R of 0.79 (p < 0.001). Model fit diagnostics were satisfactory (CFI = 0.96; RMSEA = 0.042). Moderate heterogeneity was observed (I² ≈ 58%).
Secondary outcomes were not reported. Safety and tolerability were not reported, as this meta-analysis did not involve interventions with adverse events. Limitations were not explicitly listed in the input, but the moderate heterogeneity suggests variability across studies. The correlational nature of the data precludes causal inferences.
Compared to prior research, this meta-analysis provides a comprehensive synthesis of XAI's association with cognitive psychology constructs, supporting its validity in adaptive learning systems. However, no direct comparisons to landmark studies are available from the input.
Methodological limitations include the lack of a comparator group, potential publication bias not assessed, and the moderate heterogeneity. The correlational design means that the observed associations may not reflect causal relationships.
Clinically, these findings suggest that XAI may be a valid tool for enhancing cognitive psychology outcomes in higher education, particularly in adaptive learning systems. However, educators and developers should interpret these correlations cautiously and consider further experimental research to establish causality.
Unanswered questions include the specific mechanisms driving these associations, the impact of different XAI implementations, and whether these correlations translate into improved learning outcomes in real-world settings.
View Original Abstract ↓
Machine Learning Model Performance for Predicting Contrast Induced Nephropathy in Coronary Interventions
Machine learning models help predict kidney damage during heart procedures
Related in Radiology & Imaging
From Other Specialties
