Abstract:

Tutorial understanding is a key factor in effective learning in computer science education, particularly in practical software engineering courses, where conceptual ideas must translate to hands-on programming skills. This study examined perceived tutorial clarity and exam success among 73 second-year CS students, revealing a "60% danger zone": students with 60% self-reported tutorial clarity had average exam grades of 46.2, worse than low-clarity students at 50% level (avg. 70.3) and high-clarity students at 75%+ level (avg. 66.3). A one-way ANOVA confirmed this non-linear relationship (F = 5.703, p = 0.005). This pattern resembles the Dunning-Kruger effect, suggesting metacognitive distortion where middle-range students overestimate mastery.

Preliminary qualitative evidence from semi-structured follow-up interviews hints at this interpretation, with possible relation to over-confidence in AI usage. Students in the moderate comprehension group reported high confidence in AI-tool efficiency, indicating that AI-assisted fluency may reinforce illusions of mastery by prioritizing recognition over deep understanding.

Recognizing this dynamic calls for new diagnostic and pedagogical strategies. We propose interventions like reflective practices to reveal dependency patterns obscuring true learning. Further research is needed to assess their impact on comprehension and success.

Keywords:

Education, Computer Science, Dunning-Kruger, Tutorial Comprehension, AI