Can AI Assistants Sharpen or Dull Student Thinking? Insights from a Nature Study of 698 Chinese Undergraduates

Background and Research Question

Many university students rely on artificial‑intelligence tools to support their studies, prompting concerns that ever‑smarter machines might cause human brain atrophy. To investigate this, researchers conducted a large‑scale empirical survey of 698 Chinese undergraduates, the results of which were published in Nature.

Perceived AI Intelligence (PIAI)

The study introduced the concept of PIAI – Perceived Artificial‑Intelligence Intelligence – which measures how smart, capable, and adaptable users subjectively consider an AI assistant. High‑intelligence tools quickly grasp complex instructions, converse fluently, and automatically adjust output strategies. A ten‑dimension scale was used to assess perceived intelligence, covering tool comprehension, communication clarity, and the ability to handle information independently.

Impact on Critical Thinking

Data show that students who view the tool as highly intelligent tend to exhibit stronger critical‑analysis abilities. However, when immersion becomes habitual, independent thinking drops dramatically. The researchers employed a 17‑item questionnaire specifically designed for Chinese students, examining analytical techniques, openness to criticism, and logical rigor in everyday problem‑solving.

Cognitive Engagement Findings

Smart tools can act as cognitive scaffolds, prompting deeper processing and encouraging learners to seek counter‑arguments, thereby stimulating the brain’s higher‑order reasoning. Conversely, over‑reliance erodes verification habits, turning the tool into a source of cognitive laziness.

Statistical Results

Table 1 (see image) displays clear positive correlations between perceived tool intelligence and critical‑thinking scores. Structural equation modeling (Table 3, see image) confirms that perceived intelligence alone does not cause dependence; only when it triggers deep immersion does dependence emerge, which in turn suppresses critical thinking.

Gender Differences

The sample comprised 37.5 % male and 62.5 % female participants, aged 18‑24. Males scored slightly higher on both perceived tool intelligence and critical‑thinking measures, while females reported higher AI‑dependence. These differences reflect variations in confidence and anxiety toward new generative tools rather than innate cognitive ability.

Practical Recommendations

Educators should embed explicit verification steps into coursework: students must document prompt‑engineering processes, compare information from multiple sources, and justify acceptance or rejection of AI suggestions. Short literacy and ethics training can help learners distinguish productive collaboration from pathological dependence, preserving independent judgment.