Thursday, July 26, 2012

Sometimes, the Null Hypothesis wins

In science, the Null Hypothesis is the classic way you test a new theory.  As an example, I could theorize that increasing carbon dioxide emissions into the atmosphere are causing the climate to warm.  The Null Hypothesis is that any climate changes are caused by something other than CO2 emissions, e.g. natural variation.  A truly scientific test would design ways that an experiment could gather data to help us choose whether theory beats Null Hypothesis.

As you'd imagine, most theories - while interesting - are not shown to be superior to the Null Hypothesis, and are therefore discarded.  As you'd imagine, this makes scientists unhappy.


The Null Hypothesis has all sorts of practical uses outside of the scientific research lab.  Security guru Marcus Ranum likes to say that the most important Law of computer security is that sometimes it's easier not to do something stupid than it is to do something smart.  In other words, sometimes the computer security Null Hypothesis wins.  It's the least bad alternative.

The Null Hypothesis applies to us all the time in our private lives.  Every choice that we make has positive and negative consequences: if we get an ice cream cone, we need to exercise the calories off (or get fatter).  Part of what we try to teach our children is that actions have consequences: some good, some bad.  Grown ups are supposed to try to anticipate both the good and the bad outcomes before they choose.

In other words, grown ups should try to recognize when the Null Hypothesis is likely to win.  Sometimes it's easier not to do something stupid than it is to do something smart.  Maybe 90% of the time, as it turns out.

And so with government.  Tam writes of Clayton Cramer's book about his mentally ill brother, and wonders what - if anything - could be done to change mental health care that would prevent something like the Aurora theater shooting.  It's thoughtful and well worth your time.  Her treatment of the subject is much more thoughtful than most of the ZOMG we have to do something that you hear, because she essentially lays out the Null Hypothesis - or at least two testable cases.  Reason Hit & Run also takes on the same issue, at considerable length (and is also worth your while).

I'm not at all optimistic that any legislation could be crafted that will improve the situation - and this is in theory, with a Legislature populated by disinterested Philosopher Kings.  Any law that is passed will have both positive and negative consequences, and Tam does a pretty good job laying them out.  The problem is that the "Mentally Ill" is a vastly broad group ranging from the Walking Wounded to the disfunctional to the flat our non functioning.  Any statute that can crisply and justly distinguish between these is only to be found in the Platonic Ideal of statute books.  Instead, any law will be a very blunt instrument that will almost certainly do as much (or more) harm than help.

Consider the seemingly reasonable law that would seek to restrict firearms ownership by the (non-institutionalized) mentally ill.  Who could object, right?  Well, think about the down sides: clearly those people will not be able to defend themselves in their own houses.  Some will be victimized, and remember that this is a particularly vulnerable segment of the community.  When you consider that the vast majority of this group is the "Walking Wounded" - people on anti-depressants, for example - then any victimization suffered because of this legal restriction could very well worsen their affliction.

Further, consider someone who is depressed, but who has not seen a doctor and is therefore not known to the Law as being in this group.  There's considerable evidence that treatment could help their lives, perhaps considerably.  However, some might very rationally choose not to seek treatment if it meant that they could lose their rights.  In fact, this exact argument was used by the Gay community to oppose mandatory HIV reporting requirements when they were discussed.  After all, one way to avoid reporting is simply not to test, right?  But that has clearly had consequences for society.

And so to mental illness.  Any reporting requirement that leads to restrictions on individual freedom will lead to some people simply not seeking treatment.  That means that some of the mentally ill will remain, invisibly, in the population, perhaps getting worse.  It may be that we end up with a higher population of more severely ill people than we would have.  It may be that we end up with a larger pool of dangerous people than we would have.

Consequences can be good or bad.  Any choice will have both good and bad consequences.  That's simply life.  That's what we teach our children.

To me, the Null Hypothesis is that any law that gets passed is almost certain to make things worse off than they were before.  The law may be larded with benefits to special interests, it may over promise benefits, it may ignore negative consequences.  It may do all three.  In fact, it probably will do all three.

The Null Hypothesis says that it's easier not to do something stupid than it is to do something smart.

1 comment:

Old NFO said...

Good post and good points!