**The most surprising finding is that weekend extra hours are unrelated to everything else, while weekday daily usage is tied tightly to almost every outcome.** **Daily usage is strongly linked to wellbeing.** Across all 4,500 students, Daily_Usage_Hours correlates with: - Mental_Health_Index at r = -0.85 - Sleep_Duration_Hours at r = -0.72 - Academic_Performance_GPA at r = -0.71 - Sleep_Quality_Score at r = -0.68 - Perceived_Stress_Score at r = +0.75 Those are unusually strong for survey-style human data. The Overall_Impact groups tell the same story: | Overall_Impact | Students | Avg daily hours | Avg mental health | Avg GPA | Avg sleep (hrs) | |---|---|---|---|---|---| | Beneficial | 3,681 | 4.4 | 84.3 | 3.55 | 7.0 | | Neutral | 654 | 8.5 | 63.4 | 3.00 | 5.5 | | Negative | 165 | 12.3 | 47.1 | 2.54 | 4.2 | **Weekend extra hours show no relationship at all.** Weekend_Extra_Hours is essentially uncorrelated with: - Daily usage (r = -0.01) - Mental health (r = 0.004) - GPA (r = 0.012) - Sleep (r = 0.017) - Stress (r = 0.001) Age is also uncorrelated with GPA (r = 0.028). This is counterintuitive because weekend hours are also time on social media. If usage harmed sleep, GPA and mental health, you would expect heavy weekday users to also pile on weekend hours, and weekend hours to carry some of the same signal. Instead they look like independent noise. **Why it matters:** - **Mechanism.** Weekday usage appears to be the variable that moves with wellbeing. A cap on weekend use would not be supported by this data. - **Data generation.** A near-linear relationship on one usage variable and pure independence on the other suggests the data may be generated or simulated, even though the dataset card reads it as organic. - **Causality.** The data is observational, so none of this shows that usage causes the outcomes. Mental health, sleep, stress and GPA also correlate strongly with each other (for example, sleep vs mental health r = 0.77), so they may be moving together for shared reasons. Another oddity is that 82% of students (3,681 of 4,500) are labelled "Beneficial", which sits awkwardly with the largely negative narrative around heavy use.