It's really weird seeing such obviously fallacious thinking on HN but it happens everytime mental illness and violence comes up.

When you want to predict if a person poses a risk of violence to other people knowing whether that person has a mental illness or not gives you very little predictive power. Knowing if that person is addicted to alcohol, or has had a previous episode of violence, gives you much more predictive power. And if you combine those, or either or both of them with mental illness you get a better predictor. But mental illness itself is not a predictor.

(Or, lets say, it is another parameter to watch for when discussing with new people, to help understand their point of view and ponder all the arguments, even the ones with more than sufficient counter arguments to outright dismiss them; this excercise might have to be shown to some discussion partners as well so they can see where I am coming from.)

    The guy who plunged his plane into a mountain had depression.
    The Unabomber was a mathematician.
    Hitler was an artist.

But we're talking about a college student. Sure, maybe he shouldn't be allowed in the chemistry lab or given roof access, but what does banning him from college grounds prevent that he can't do outside of it?

Besides, if he was so dangerous to warrant banning him from college grounds just so he couldn't get a chance to harm or kill others (e.g. strap bombs to himself and blow people up), he was probably too dangerous to rely on him to respect the ban. This is the point where you involve the police instead of risking to provoke him further -- and then this isn't just a mental issue, it's a direct threat. But as it seems, this wasn't what happened, even remotely.