How to design a committee to determine when a probability is "beyond a reasonable doubt?" That's the task facing juries in criminal cases. (Juries are like prediction markets for the past: they are charged with determining probabilities ex post.)
In all but two States, the rule is that unanimous agreement is required. But the States aren't unanimous. A current challenge to a non-unanimous Oregon conviction raises the question: Guilty by a 10-2 Vote: Efficient or Unconstitutional?
"...Oregon is one of only two states that does not require juries to reach unanimous verdicts in criminal cases. Like Louisiana, it allows convictions by a vote of 10 to 2.
In a pair of decisions in 1972, the Supreme Court said that was all right, that the Constitution does not require states to insist on unanimity.
But the decisions, one each from Oregon and Louisiana, were badly fractured and internally inconsistent. They concededly ignored the historical record and made assumptions about jury behavior that have been called into question by more recent research."
...
"According to the Oregon Criminal Defense Lawyers Association, most felony convictions in the state are the products of nonunanimous juries. Oddly, misdemeanor convictions still require a unanimous vote, though from a six-member jury. That means, the association said in a brief supporting Mr. Bowen, that prosecutors face a lighter burden in more serious cases.
Oregon does require a unanimous vote in first-degree murder cases, and Louisiana requires it in capital cases.
A unanimity rule would seem to reinforce the requirement that prosecutors prove their cases beyond a reasonable doubt. Two jurors out of 12, if you do the math, represent about 17 percent of the panel. That’s a fair amount of doubt."
That suggests that those voting not guilty have 100% doubt and those voting guilty have 0% doubt. If people instead vote based on which is more likely, then if 10 vote for and 2 against you could have close to 50% doubt:
ReplyDelete(51%+51%+51%+51%+49%+49%)/6
I don't see how you can infer probabilistic beliefs here without additional assumptions about the relationship between doubt and the the conviction decision.
Here is another thought. If you believe Cass Sunstein's work (i.e. Going to Extremes), you may have more faith in a 10-2 decision (assuming they deliberated for some time) because it means that alternative verdicts were discussed and rejected by the majority. Otherwise, you may simply have a unanimous verdict because of information and/or social cascades that hide a more divided jury.
ReplyDeleteIn theory, it's actually much harder to convict than the previous posters suggest, as the high vote juries need unanimity of jurors who EACH believe that proof has been shown beyond a reasonable doubt.
ReplyDeleteThis would imply that a simple majority would correctly estimate "beyond reasonable doubt".