The thing is that you can 'prove' almost everything if you doggedly search for refinements to your hypothesis. In effect, you fit the theory around the evidence. In contrast, in science, the goal is (ideally) to test hypotheses by checking how well they predict new evidence, and to run rival hypotheses against each other.
Most people - and that includes most lawyers and judges, never mind the public - don't understand enough about scientific evidence to be able to assess it objectively. They also don't understand statistics. Tell someone there's a 100,000:1 chance of a DNA match, and they'll assume this is absolute proof of guilt, even though in a big city there will be 10-100 people equally likely to provide a match.
But I'm not sure you can apply the scientific method to law anyway. If you have a collection of evidence, it's hard to see how can you turn it into a hypothesis and then a firm prediction that will lead to a crime being solved.
Effectively a trial is a hypothesis - that a suspect is guilty. But it's not a prediction, except in the rather useless sense that you assume that if someone is imprisoned, a certain series of crimes will end.
