[Metric]

Haven't read most of the thread but I do agree with David's POV that it would be best to encode "reasonable doubt" as a specific probability estimate of guilt.

In fact, I bet that leaving it vague is where a lot of discrimination comes into play. Based on the evidence, someone might estimate a 90% chance of guilt, and be willing to say this counts as "reasonable doubt" for a defendant they like and respect, but might not count as "reasonable doubt" if it applies to your local gang member scum.

No doubt this kind of discrimination will exist anyway, but I think working the evidence directly into a probability estimate probably encourages more objectivity than working the evidence into whatever strikes you as "reasonable" after you get a good look at the defendant.