Generative AI boosts learning, but only if students think critically

GAI helps reduce extraneous cognitive load by supporting information retrieval and organization, allowing students to allocate more cognitive resources to higher-order thinking. In effect, GAI functions like a scaffolding system - freeing up bandwidth so students can focus on solving complex problems rather than getting bogged down in rote details.

The rapid adoption of generative artificial intelligence (GAI) in classrooms has stirred both excitement and concern across the global education landscape. Amid calls to explore its potential, a new peer-reviewed study sheds light on how GAI impacts student learning outcomes and the findings are transformative. The study, titled “Generative Artificial Intelligence Amplifies the Role of Critical Thinking Skills and Reduces Reliance on Prior Knowledge While Promoting In-Depth Learning”, was published in the May 2025 issue of Education Sciences. It reveals that while GAI enhances student performance when integrated into lessons, its effectiveness is closely tied to students’ critical thinking abilities, not how much they already know.

The research, conducted in a top-performing primary school in China, tested 126 sixth-grade students across three groups. Two groups used GAI, specifically Baidu's ERNIE Bot, as either a cognitive tool to generate ideas or as a thinking tool to guide reasoning. A third control group received traditional, lecture-based instruction. Students completed tests that assessed both factual retention and knowledge transfer. The results upend long-held assumptions about the primacy of prior knowledge and suggest that GAI may be fundamentally changing how, and for whom, learning happens most effectively.

How did GAI enhance in-depth learning?

The experiment was designed around an ICT lesson on information encoding. Students were tasked with solving a real-world problem: how to create an efficient coding scheme for school uniforms to prevent loss. While the control group brainstormed without external tools, the two experimental groups received GAI-generated materials that either provided solutions or demonstrated thinking strategies like factor analysis.

The study measured two aspects of in-depth learning: retention of factual content and the ability to transfer knowledge to a new task. Interestingly, there were no significant differences between the groups in factual recall. However, in the transfer section, where students had to apply knowledge to encode electronic devices, those who used GAI outperformed their peers by a wide margin. Experimental Group 1 (cognitive tool) and Experimental Group 2 (thinking tool) scored significantly higher than the control group, showing that GAI promoted deeper, more applicable learning.

From a theoretical standpoint, this aligns with cognitive load theory. GAI helps reduce extraneous cognitive load by supporting information retrieval and organization, allowing students to allocate more cognitive resources to higher-order thinking. In effect, GAI functions like a scaffolding system - freeing up bandwidth so students can focus on solving complex problems rather than getting bogged down in rote details.

Why Does Critical Thinking Matter More Than Prior Knowledge?

One of the most surprising findings from the study was that prior knowledge, long viewed as the cornerstone of effective learning, did not significantly impact student outcomes when GAI was introduced. In traditional learning contexts, students with more background knowledge typically have a cognitive advantage. But in GAI-enhanced environments, that edge appears to diminish.

Instead, critical thinking skills emerged as the most influential factor. Students with high critical thinking abilities were better equipped to assess, evaluate, and integrate the information generated by GAI. They didn’t just accept AI outputs at face value; they interrogated them, synthesized them, and tailored them to the task at hand. This interaction amplified the effectiveness of GAI and drove higher performance in in-depth learning tasks.

The study emphasizes that critical thinking is not a passive trait but a skill set that involves evaluating information credibility, identifying biases, and integrating multiple perspectives. When paired with GAI, it acts like an internal monitor, akin to Krashen’s Monitor Hypothesis in language learning, helping students regulate, refine, and extend their understanding.

Interestingly, the study also found that critical thinking didn’t just support GAI use - it amplified it. There was a significant interaction effect between GAI and critical thinking skills. In other words, students who were already strong in critical thinking saw even greater gains from using GAI than those who weren’t. This suggests that GAI does not democratize learning outcomes by itself. Rather, it enhances the abilities of those already equipped with higher-order thinking tools.

What does this mean for future classrooms?

These findings have far-reaching implications for pedagogy, curriculum design, and educational equity. First and foremost, they signal a shift in instructional priorities. If critical thinking now plays a more central role than prior knowledge in driving learning outcomes, especially in tech-enhanced settings, then schools must adjust their teaching strategies accordingly. Building students' critical thinking skills should no longer be treated as an optional supplement - it’s the cornerstone of effective learning in the age of AI.

Moreover, the way GAI is used matters. The study showed that both cognitive and thinking tools can support learning, but not all AI integration is equal. Educators must guide students in using GAI not as a shortcut for answers, but as a partner in reasoning. This involves creating structured prompts, designing tasks that require critical evaluation, and scaffolding student-AI interaction to foster autonomy rather than dependence.

There are also implications for educational equity. While GAI may help close gaps in prior knowledge, it may simultaneously widen disparities in critical thinking unless those skills are deliberately nurtured. This makes teacher training essential. Educators need to be equipped not only with AI literacy but also with strategies to cultivate critical reasoning in their students.

The study also opens new questions for future research. How do different age groups respond to GAI? What are the long-term effects on critical thinking development? Can GAI itself be designed to adapt to a student’s critical thinking profile? These are urgent areas of inquiry as schools, policymakers, and developers consider how best to integrate AI into learning systems.