{
    "action": "final",
    "content": "**The most surprising finding is that weekend extra hours are essentially unrelated to everything else, including daily usage, while weekday daily usage is strongly tied to almost every outcome.**\n\nThe evidence (n = 4,500 students, all correlations computed in the same query results):\n\n- **Weekday daily usage correlates strongly with outcomes.** Correlation of daily usage with mental health is -0.85, with sleep hours -0.72, with GPA -0.71, with sleep quality -0.68, and with perceived stress +0.75.\n- **Weekend extra hours show almost no relationship with anything.**\n  - With daily usage: -0.010.\n  - With mental health: 0.004.\n  - With GPA: 0.012.\n  - With sleep hours: 0.017.\n  - With stress: 0.001.\n\n  A student who uses social media heavily during the week is no more likely to use it more at weekends. Weekend extra hours also carry no signal about wellbeing.\n- **Age is also unrelated to GPA** (r = 0.028), so older or younger students do not differ academically.\n\nThe strength and tidiness of the usage–outcome relationships is also unusual for survey-style data. The Overall_Impact groups separate almost like a dose-response ladder:\n\n- **Beneficial** (3,681 students): average usage 4.42 hours, mental health 84.3, GPA 3.55, sleep 7.03 hours, stress 11.16.\n- **Neutral** (654 students): 8.5 hours, mental health 63.4, GPA 3.0, sleep 5.48 hours, stress 22.3.\n- **Negative** (165 students): 12.25 hours, mental health 47.1, GPA 2.54, sleep 4.23 hours, stress 31.12.\n\nWhy this is counterintuitive:\n\n1. The common expectation is that heavy users are heavy at weekends too. Here there is no link at all.\n2. A correlation of -0.85 between usage and mental health is far stronger than typical real-world studies find. That raises the possibility that the data is synthetic or generated from formulas, even though the card reads it as organic.\n3. Most students (3,681 of 4,500, about 82%) are labelled \"Beneficial\", despite the strong negative relationship between usage and wellbeing. This reflects how many students use social media only lightly.\n\n**What this leaves open:**\n\n- The data is observational, so none of these correlations establish causation. Poor mental health or sleep could drive heavier use, not just the reverse.\n- I did not test whether the weekend result holds within segments such as age, platform, or academic level.\n- I did not check whether Overall_Impact is derived from the other columns, which would make the ladder partly circular.\n- I did not test for non-linear patterns, so a near-zero correlation does not prove weekend hours are irrelevant."
}