Page 675 - ISC PROCEEDINGS 21.4

Verification failure and automation bias
Another critical concern is the tendency of learners to trust AI outputs without
sufficient verification. Studies have identified risks such as hallucinated information and
inaccuracies in AI-generated responses. Without proper training in verification strategies,
learners may develop automation bias—the inclination to accept automated outputs
uncritically.
This issue extends beyond factual errors. The erosion of questioning habits—such as
evaluating assumptions, identifying sources, and considering counterarguments—
represents a deeper loss of critical thinking capacity. These metacognitive processes are
fundamental to academic reasoning and intellectual autonomy.
2.3.3. Implications for university students
The impact of AI misuse is particularly significant for university students, who are in
the process of developing disciplinary knowledge and independent learning skills. While
AI may enhance short-term performance, excessive reliance can hinder the development
of foundational competencies, especially in fields requiring complex reasoning.
Systematic reviews highlight a dual effect: AI increases access to information and
support, yet may weaken independent learning when overused. The benefits of AI for
higher-order thinking are moderate and highly dependent on pedagogical design,
emphasizing the importance of structured and intentional integration.
A cognitive model of decline
The impact of AI misuse can be conceptualized as a reinforcing cognitive cycle.
Excessive reliance on AI reduces direct cognitive effort, leading to increased cognitive
offloading. This results in diminished verification and critical evaluation, which in turn
reduces deep processing in working memory and the hippocampus. Consequently,
memory encoding becomes shallow, weakening long-term knowledge consolidation. As
reliance increases, this cycle reinforces itself, gradually eroding deep thinking and
academic autonomy.
From both educational and neuroscientific perspectives, the critical issue is not the
existence of AI but its role in the learning process. When AI replaces essential cognitive
activities—such as reasoning, analysis, and reflection—it can undermine deep thinking
through reduced effort, weakened attention, and impaired memory formation.
Conversely, when used as a supportive tool that enhances rather than substitutes human
cognition, AI holds significant potential to improve learning outcomes.
Therefore, the challenge lies in designing educational practices that preserve the
learner’s active cognitive engagement while leveraging the advantages of AI.
3. Methodology
3.1. Research design
This study adopts a mixed-methods research design to examine the relationship
between digital media consumption, artificial intelligence usage, and the development of
deep thinking among university students. A sequential explanatory design was used,
combining quantitative survey data with qualitative interviews in order to provide both
statistical insights and deeper contextual understanding of students’ cognitive
experiences in digital learning environments. The quantitative component examined
correlations between students’ digital habits and their performance on deep thinking
tasks, while the qualitative component explored students’ perceptions regarding their
cognitive engagement when using digital media and AI tools for academic learning.

674