Various sources suggest that just 3-5% of inmates in prison are there because of a trial. The vast bulk of them got there due to a guilty plea, part of a plea agreement.

This seems shocking, but in fact it is to be expected. Trials are long and expensive and risky. The vast majority of commercial disputes are settled out of court.

The problem is that some of those people are innocent, but they were pressured to accept a plea deal, or simply evaluated the odds of success in a trial and took what seemed to be a better offer. Plea bargains are much less common outside the USA. The plea negotiation is far from a fair deal with equal parties. They are also a relatively recent invention. There are articles about the flaws

It is not known what fraction of defendants are innocent of the charges, or at least of the full charges the prosecutor threatens them with in the opening round of negotiations, but any is too many.

I propose a method of checks and balance on the abuse of the plea deal as follows.

After a negotiated plea is entered, the defendant may be selected at random to get a trial anyway. We might select one defendant in 100, however many we can afford. The goal is not to give them better justice, but to provide accountability on the system.

The defendant could decline the extra trial. This election would be done after sentencing, so the prosecutors should have no further power over them. If it goes forward, it would go forward with a new prosecutor, and if the defendant elects, a state-provided lawyer. The court would ideally not be informed that this is an accountability test, though it may be hard to keep that knowledge from the judge entirely.

After the court rules, the defendant could choose between that verdict and their plea deal. In other words, it's always a win, other than the time and stress of a trial. In addition, after the sentencing, the court would be informed of the nature of the trial, and it would do a review of the prosecutor's work on the plea deal. Particular attention would be paid to defendants found not guilty, or any inappropriate action by the prosecutor. Statistics would then be learned about the system, and individual prosecutors. Bad prosecutors and offices could be censured.

The court would learn what sorts of offers the prosecution made, and could rule them to be out of line, resulting in censure. The prosecutor could present arguments at this time, or have (free) counsel. One way to test the prosecution's offers would be to require the trial attempt the prosecution's "threat." In a plea negotiation, the prosecution might say, "We're charging you with 2nd degree murder, but we will offer a plea deal of manslaughter." In this case, the prosecution might be required to try for 2nd degree murder.

This one time in 100 review would hit most prosecutors every so often. Often enough that it should serve as a deterrent against abuse of the system. In other words, if it's working, it should rarely find problems.

Plea deals save the state a fortune. 96-97% of cases don't have the high cost of trials. This proposal takes a tiny part of that saving and makes the system better.

Variants

It might also be possible to do this only on defendants who ask for it. They would ask only after their plea deal is complete, with sentencing. Only a fraction of those who ask would get the trial. This would make it more likely we would be looking at the subset of cases where defendants feel they got a raw deal in the plea bargain, or were innocent. Of course, with a pure choice of verdict or deal, there is little downside to asking, so that might be modified to give the court the power, in egregious cases, to vacate the plea deal.

Ideally the new prosecutor would be unaware this is a test trial. You could even give it to them like a real case, except when they try to plea bargain the defendant keeps a hard line and refuses. Chances are they would figure it out, though, especially in a small town. On the other hand, they remain motivated to "win" to vindicate their colleague -- whose name in theory they should not know.

It could be that these trials would take place in another nearby town, but that would add some difficulty in getting witnesses to travel etc. And then everybody would know what was going on.

The defendant might be allowed to select a public defender or their own private counsel. However, if they are found innocent, the state would pay for the private defence, within reasonable limits. This is a strong incentive for the state to not force plea deals on people they can't actually convict.

Legality

There are a couple of issues with doing this. In the USA, the constitution requires that no person be put in jeopardy twice for the same offence. This stops the courts from trying you several times until they get what they want. However, with the rule that the defendant can always still take their plea deal, they are not in any further jeopardy. If that's not enough to get past that rule, it could be done by technically vacating the first trial and sending it to the new court with sealed and binding instructions from the higher court on what the maximum sentence can be.

This might also be found to be unfair, in that everybody might want a chance to reduce their sentence and get their day in court, and granting it only to some might not be equal justice. If so, it would need to be done with no chance in a change. In that case, many defendants might decline to participate. In effect, it would simply become a very intense examination of the fairness of the plea deal, which still could be a good idea.

Another kludge would be to use the mock trial as grounds for appeal or claim of mistrial. After an appeal, there could be grounds for a real new trial, but a real new trial would have the power to sentence the defendant to worse than they got in the plea deal, though in theory the chances of that should be low if the mock trial was done properly.

Hopefully these workarounds would not be needed. Plea bargains do save the justice system, and defendants, a great deal of resources. They also will mean that guilty defendants get lighter sentences than they would have gotten the expensive way, but also that other defendants (innocent and "not so guilty") take heavier sentences to avoid the risk. Some sort of checks and balances should be a win.