[gknoy]

> in a coldly rational statistical sense, most of the accused are in fact guilty

Without a trial, how do we know that?

If I am accused of something terrible, of which I am innocent, but a plea bargain gets me back to my family in N years instead of Never (or 10*N), I'm likely to lie and plead guilty. We've seen that the government doesn't merely use plea bargaining as a cost reduction tool, but rather as a very large hammer with which to ensure that people get punished in extreme ways.

While there are good police and prosecutors, as a system they are driven to increase convictions rather than to find the _guilty_. Given the chance, they can find something to convict nearly anyone of, and guilty verdicts can be nearly guaranteed against even people who are innocent by heaping up enough charges that either defense is too expensive or the penalty of losing at trial too large.

[tptacek]

You absolutely don't know that someone is likely to be guilty in any specific case. It's a statistical, macro-lens point.