Ai Student Impact Dataset

By shrijeetverma13 · May 16, 2026

As Generative AI tools like ChatGPT, Copilot, and Gemini become deeply embedded in academic life, understanding their true impact on students has never…

There is a moderate positive correlation (Pearson r = 0.31) between Perceived AI Dependency and Anxiety Level During Exams. This means students who feel more dependent on AI tend to experience higher exam anxiety. Two visualizations were generated: a bar chart showing mean anxiety levels at each dependency score, and a heatmap of the joint distribution between the two variables.

The Institutional Policy that yields the best Post-Semester GPA outcomes is 'Actively Encouraged', which achieves an average Post-Semester GPA of 3.353. Two visualizations were generated to help you explore this finding: a bar chart comparing average Post-Semester GPA across all institutional policies, and a box plot showing the full distribution of GPAs for each policy group.

Weekly GenAI usage has very little impact on GPA change from pre to post semester. The Pearson correlation is -0.046, indicating a weak and slightly negative relationship — meaning more GenAI hours is marginally associated with slightly smaller GPA gains, but the effect is minimal. Two charts and data tables were generated to visualize this relationship.

STEM is the Major Category with the highest combined Burnout Risk and AI Dependency. It has a 30% high burnout rate and an average AI dependency score of 3.79 out of 10. Two visualizations were generated: a grouped bar chart comparing burnout percentage and AI dependency across all major categories, and a scatter plot showing the relationship between the two metrics.

The analysis produced two visualizations and supporting data tables showing how Prompt Engineering Skill level and Tool Diversity together influence Skill Retention Score. A heatmap displays the average retention score for every combination of skill level (Beginner, Intermediate, Advanced) and number of tools used, while a grouped bar chart lets you compare these combinations side by side.

No, paid subscription users do not meaningfully outperform free users. The differences in both GPA and Skill Retention Score are negligible. Paid users have a slightly higher Post-Semester GPA (3.352 vs 3.347), but free users actually score slightly higher in Skill Retention (76.07 vs 75.42). Two bar charts and supporting data tables were generated to visualize these comparisons.

Using the IQR method (1.5x rule), outliers were identified in both study hour columns. Weekly GenAI Hours had bounds of [-11.61, 25.72] with 2,583 outliers, while Traditional Study Hours had bounds of [-3.17, 25.44] with only 161 outliers. In total, 2,743 unique students showed extreme study behavior in at least one category. Box plots were generated to visualize the distribution and spread of both variables, clearly showing where outliers fall relative to the interquartile range.

The analysis produced a grouped bar chart comparing Pre-Semester and Post-Semester GPAs across all four year groups (Freshman, Sophomore, Junior, Senior), with a red line overlay showing the GPA delta (change) for each group on a secondary axis. The chart clearly displays both GPAs side by side for easy comparison, and the delta line highlights which year groups improved or declined after a semester of AI tool usage.

The correlation matrix has been computed for all 8 numeric variables and visualized as an interactive heatmap with annotated coefficients. A second bar chart highlights the top 3 strongest positive and top 3 strongest negative correlations. The analysis reveals that most correlations in this dataset are weak ( r < 0.1), meaning no single numeric factor strongly determines academic outcomes on its own.

The analysis ranked all 15 combinations of Primary Use Case and Prompt Engineering Skill by their average Post-Semester GPA. Data tables were generated showing the full rankings with study hour breakdowns. The top combination is Debugging/Troubleshooting paired with Advanced prompt skills (GPA: 3.508), while the lowest is Direct Answer Generation paired with Beginner skills (GPA: 3.285). Interestingly, both top and bottom groups lean toward traditional study over AI tools, with very similar traditional study hours (~11.2h/week) across all groups.