Streetwise Professor

If Newtown illustrates anything, it is that tragedy is the father of numerous idiocies, and the greater the tragedy, the more numerous, and more idiotic, the idiocies.

I understand completely the desire to do something to prevent Newtown from ever happening again.  But that very desire leads to inane proposals-some adopted, unfortunately-that will do more harm than good.

Case in point: the NRA’s proposal (brought to my attention by commentor paul) to mandate a creation of a national database of the mentally ill.

Another case in point: this recommendation for a “three strike” mental violence law.

Or all of hue and cry to crack down on violent computer/video games.

I was planning to write on this anyways, but paul’s comment got me to put my thoughts down in writing.  In my reply to him, I pretty much said what I wanted to say, so I’ll just repeat it here.  (Warning-wonkiness ahead!  But rigor is needed, precisely to show how flabby and unrigorous these arguments are.)

Yes, the NRA’s proposal is . . . insane. So will those who thought it up be in their database?

But they are not alone. Everyone has their pet theory, their hobby horse. Most of these are based on the observation that certain traits, behaviors, and activities are prevalent among mass murderers. Mental illness and intense involvement in violent video games are two examples.

But even if it is possible to identify a set of traits/behaviors that most mass murderers have, this information is almost totally useless in predicting who is a likely mass murderer.

How do I know that? Bayes’ Theorem.

All of the observations about common traits are of the form: “conditional on someone being a mass murderer, the probability they played a violent video game or are mentally ill is high.” But what we’re really interested in is: “what is the probability of someone being a mass murderer, conditional on his being a player of violent games or mentally ill?” Even if the first probability is nearly 1, the second probability-the one we’re interested in-is almost certainly very low. For the second probability is the first multiplied by the ratio of the unconditional probability of someone being a mass murderer to the unconditional probability of someone being insane, or a player of violent videos.

The prevalence of mass murderers is far smaller than the prevalence of people with mental illness. Meaning that this ratio is near zero. Meaning that conditioning on mental illness tells you virtually nothing Put differently, the rate of false positives is near one for virtually any set of conditioning variables you choose.

It all boils down to this: even if all mass murderers are mentally ill, very few mentally ill people are mass murderers. Which means that information about mental health is essentially useless for predicting ex ante whether a particular individual is a potential mass murderer. You can substitute pretty much any other conditioning variable or set of conditioning variables for “mentally ill” and draw the same conclusion.

Which means that the NRA’s proposal is not just creepily draconian, it is a complete waste of resources. Complete. Tarot readings would be about as effective, and a lot cheaper.

There is a tragic truth in all this; a truth that people hate to acknowledge. Things like Newtown are extremes of the extreme. They are difficult to explain ex post, and virtually impossible to predict ex ante. I understand perfectly the intense desire to do something, but every something that has been proposed-including the NRA’s proposal-will have no impact on the likelihood of such events. As I said in my first post-Newton post, the rates of false positives and false negatives will both be quite high. Both types of error are very costly.

We live in a fallen world. Sometimes there’s just nothing you can do about that, and trying to do something is often the worst thing you can do. But that is something that people are loath to concede.

To reiterate.  Real math, real statistics-Bayes’ Theorem, to be precise-demonstrates that proposals to identify people based on characteristics common to many mass murderers are completely counterproductive.   Conditioning on these characteristics provides virtually no useful information.  There are huge rates of false positives.  False positives are costly.

Indeed, there are other costs.  Consider the three strike proposal: great way to get people to avoid seeking help!

It’s bad enough when the NRA or Sarah Palin cheerleaders make these proposals.  It gets truly scary is when those in law enforcement or intelligence use this as a justification for expanding the surveillance state.  Like the NYPD, for instance:

In the wake of the Newtown massacre, the NYPD is examining ways to search the internet for potential “deranged” gunmen.

Police Commissioner Ray Kelly said that their searches would be similar to those being used to spot terrorist chatter online.

“And what we’re talking about is publicly available websites, chatrooms, that sort of thing,” Kelly said, adding that algorithms could be used. That will enable us to use, perhaps, commonly used terms that are used by people engaged in this sort of activity,” he said.

“The techniques would include cyber-searches of language that mass-casualty shooters have used in e-mails and Internet postings,” Kelly said

“Language that mass-casualty shooters have used.”  Again: even if all mass-casualty shooters use particular language, that is virtually irrelevant if many other people use the same language, but don’t commit mass murder.  Even if every haystack has a needle, treading every stalk of hay as a needle is idiotic.

Right Reverend Bayes, please give Chief Kelly a call from the Beyond.  Set him straight.

It’s one thing when the NRA spews such idiocy.  It is truly frightening when law enforcement and intelligence does.  These people can actually do things.  Dangerous things.  Things that violate the privacy of innocent people.

And here there is no real trade-off, to speak of.   The violations of privacy will be  real, and ubiquitous.  The police and legal harassment that harmless people incautious in their language will suffer will be real.  The time and effort that cops spend chasing down these people will be real.

And for what?  What is the likelihood that a real potential mass-shooter could be identified?  And even if one were identified, what could be done to stop him?  Is everyone caught in this crude screen to be followed constantly? Incarcerated? What?  And at what cost?

And it’s worse than that.  The resources being diverted for this snipe hunt are resources that could be used to prevent and deter other, more pedestrian crimes.  There are unseen lives and property that will be lost as a result of this diversion.  Will be.

What is truly disturbing about this is the opportunism.  Law enforcement and intelligence are always-always-of the “let no crisis go to waste” mentality.  They-as are governments generally-are notorious for their opportunism in seizing on tragedy to expand their power, their reach.

Newtown is bad enough as it is.  Don’t make it worse by using it as a pretext to expand the already bloated surveillance state.  It’s bad enough that innocent children die in vain.  It’s all the worse if their lives are used to justify the expansion of an already intrusive and dangerous surveillance apparatus that squeezes civil liberties in the name of security, but which in fact gives us precious little real security in return.