Standards of Evidence

In a recent post, Richard Dawkins opines that it makes no sense for the outcome of a criminal trial by jury for the verdict to be uncertain.  Because guilt must be established beyond a reasonable doubt, he maintains that it is nonsense for the verdict to be uncertain.

This conclusion is the result of confusing the question of guilt with the question of certainty of guilt.  By returning a verdict of not guilty, a jury is not claiming that the defendant did not commit the offense.  What they are claiming is that the prosecution has failed to establish beyond a reasonable doubt that he was guilty.

To illustrate this point let's imagine two jurors who disagree on the appropriate verdict.  One juror believes that no reasonable person would have doubts about the guilt of the defendant.  The other believes that a reasonable person might have such doubts.  Note that they are not disagreeing about whether or not the defendant is guilty, nor on whether they themselves have these doubts.  What they are disagreeing on is what sort of doubts a reasonable person might have.  One question touches on the certainty of guilt, the other on the certainty of that certainty.  These are two different questions.

However, it seems that what Richard Dawkins is concerned about is that juries often return verdicts that turn out to be wrong in light of new evidence that is discovered after the trial and often, unfortunately after the defendant has been executed.  This is a valid concern.  Also, it is noteworthy that reasonable doubt is not supposed to be a definite probability.  It is as if there must be complete certainty in order to convict.  This is much too high a standard.  Complete certainty is not something that we can have even regarding the laws of nature, but establishing guilt or innocence depends on those laws.

In the real world witnesses can be mistaken.  Perhaps the victim was killed by a man with identical fingerprints.  Is there DNA evidence?  Maybe the defendant was cloned or the sample was contaminated.  Maybe our understanding of the testing procedures is in error.  Provided we are willing to entertain possibilities that are of a sufficiently low probability, we will always be less than completely certain.

And yet no jury would ever go so far as to find every defendant not guilty simply because there will always be a non-zero probability of innocence.  That policy wouldn't be reasonable, especially when we consider that a guilty offender may re-offend and kill an innocent person.  To scientists and other mathematically inclined people, reasonable doubt implies a certain probability, and yet the law explicitly says that it does not.