Rob MacCoun reports that in simulated-jury experiments, the probability that a “juror” will come into the deliberation with a verdict for or against someone is largely independent of the standard of proof as given in the judge’s instructions: on the same evidence, a juror is just as likely to find that someone is “guilty beyond a reasonable doubt” as to find that the person is “responsible [for an accident] by a preponderance of the evidence.”
But the enunciated standard of proof matters: it determines how willing people are to change their minds. An experimental subject whose first impression is that a defendant is guilty is likely to change his mind if randomly put in a mock jury with one or two “jurors” who argue for a “not guilty” verdict, but that same subject will stand pat on his finding of “responsible by a preponderance of the evidence.”
So “reasonable doubt” is defined in the mock-juror’s mind not epistemologically, but socially: something is subject to “reasonable doubt” as long as an apparently reasonable person doubts it, while it’s true by a “preponderance of the evidence” unless a majority disagrees.
Note the scary implication: twelve jurors, each of whom thinks that someone is probably, but not certainly, guilty will tend to find that person guilty beyond a reasonable doubt.