Dan Scott said:
This case makes appears to make it impossible from taking a gun away from somebody that is perceived as a threat. After a mass shooting people always say there were signs, his family said he was crazy, etc.
In reality, you can't just take something away from somebody unless clinically diagnosed as crazy I guess.
It's a classic classification problem: Does a person present a clear and present danger to themselves or others?
There are 2 classifications, 2 true conditions, and 4 outcomes:
Yes, and the classification is correct
Yes, and the classification is incorrect
No, and the classification is correct
No, and the classification is incorrect
I could put this into a concussion matrix (actual term), but I'm on mobile and not going through that, so here are the consequences of each outcome:
So let's say there's, "signs," which there usually are, and you take someone's guns because you predict, "yes.". If you are correct, you've likely saved lives. If you are incorrect, you have violated someone's rights.
Let's say there's, "signs," and you predict, "no." If you are correct, nothing happens. If you are incorrect, somebody probably dies.
Now, it's easy to predict every single case of a true positive, or, "Yes this person is a clear and present danger," by simply picking yes every time time. You will get a 100% recall rate, but at the expense of high Type 1 error (false positives). This is the liberal desired course of action.
You can do the inverse and prevent all false positives while identifying all true negatives by defaulting to, "No," every time. If your goal is to never target an individual who is not a clear and present danger, to will never miss here. The problem is you will have plenty of Type 2 errors (false negatives) who go on to kill people. This is, I think, the generally conservative course of action.
The disagreement comes down to what your intended purpose is: Are you more concerned with safety from people suspected of being dangerous, or are you more concerned with individual rights? Both are valid concerns. You don't want to arrest someone or confiscate property unless they've committed a crime, but at the same time you want to prevent crime because it generally cannot be uncommitted against its victims. The optimal strategy to balance these interests is really to fall somewhere in the middle.
This would be easier if we knew when all of these actions were right or not, but we don't. Proper prevention is indistinguishable from a lack of risk: You can't tell the difference between a prevented event and an event that was never going to happen, because neither by definition would. By that same token, who is to say that someone you pass on today as being harmless doesn't go on a killing spree in 6 months or 20 years? We don't know the true positive and negative rates or the false positive rate, we really only know the false negative. There's an asymmetry here because ALL true negatives are by definition non-events, while SOME true positives end up also being non-events due to prevention. With that in mind, it is no wonder why many would want to address the false negative rate when it is all they have to go on.
