Mike Huemer’s new Justice before the Law is predictably excellent. I’ll eventually discuss it in greater depth, but for now I’ll focus on Huemer’s critique of plea bargaining.  The heart of the critique is that plea bargaining is coerced confession:

It is universally agreed in legal theory that coerced confessions are unacceptable. The main reason is that to accept coerced confessions would conflict with the central purpose of the court system… Coerced confessions do not establish the truth, nor do they promote justice, since the innocent can be coerced to confess as well as the guilty.

Now suppose that you’re innocent, and the prosecutor offers you this deal: “Plead guilty to get a sentence of 5 years in prison.  Or go to trial, and get 15 years with 90% probability.”  What are the odds that you’ll take the deal despite your innocence?  Very high indeed – and Huemer points out that this is a realistic scenario:

In the actual status quo, defendants, in return for pleading guilty, are commonly offered sentences one third as severe as they could expect if they were convicted at trial. Standard rational choice theory dictates that in such a situation, a rational defendant accepts the plea bargain as long as his probability of being convicted at trial is greater than one third.13 Imagine, then, the spectacle of a defense attorney advising his client that, since he is “only” 60% likely to be acquitted at trial, it is in his interests to plead guilty. Or imagine a prosecutor deciding that, since the evidence he has gives him about a 35% chance of convicting a suspect at trial, it is worth going ahead and filing charges. Something has gone very wrong in a justice system in which those would be correct calculations. Surely a defendant who would probably be acquitted should not be given incentives sufficient to make it rational to plead guilty.

Even worse:

Prosecutors have enormous control over this factor and can and do adjust it in individual cases to take account of the strength of the case against the defendant. A prosecutor thus has a good chance of extracting a confession from a rational defendant given almost any nonzero probability of obtaining a conviction at trial. For instance, if there is just a 10% chance of convicting the defendant at trial (more precisely, if the defense believes there is a 10% chance), the prosecutor need only adjust his offer to ensure that the expected punishment if the defendant goes to trial and is convicted is more than ten times greater than the punishment offered in the plea agreement.


Experimental evidence confirms that innocent people can be induced to admit to wrongdoing. In one study, students were accused of cheating and were encouraged, through offers analogous to plea bargains, to admit their guilt. The study created conditions in which some students actually cheated, while others were innocent. The researchers found that actually guilty students were more willing to admit their guilt in exchange for leniency (89%), but that the majority of innocent students (56%) would falsely condemn themselves in conditions analogous to those of defendants in the criminal justice system.

After reading Huemer’s critique, I realized that our plea bargaining system is epistemically inferior to extracting confessions using what I call “thoughtful torture.”

Yes, “people will confess anything under torture.”  Torture’s most thoughtful defenders, however, have long insisted that the real point of torture is to get suspects to confess things they couldn’t know unless they were guilty.  Torturing someone to confess, “I did it!” shows nothing.  Torturing someone to show you where the missing body is buried, in contrast, reveals guilt, or at least complicity.

Both plea bargaining and thoughtful torture unjustly punish the innocent.  Unlike thoughtful torture, however, plea bargaining doesn’t seriously try to unearth the truth.  Sure, you could give novel details when you plead guilty to a crime, but normally you merely need to parrot the prosecutor’s preferred narrative of the crime.

Huemer’s main suggestion for reforming plea bargaining is to drastically reduce the difference between the sentence if convicted and the sentence if you plead guilty:

The most serious problem with plea bargaining as currently practiced is that it creates incentives sufficient to induce rational defendants to plead guilty even if they are innocent and would probably be acquitted in a trial. In such a system, a defendant’s willingness to plead guilty does not provide strong evidence of his guilt; a court therefore cannot proceed to punish a defendant on that basis without violating its central duty to pursue justice in the case. This problem is mitigated by a sufficiently tight constraint on the trial penalty. A 20% trial penalty is unlikely to induce an innocent person to plead guilty.

My alternative reform, in contrast, is that plea bargains should only be accepted if the accused reveals novel hard evidence of their own guilt.  If they know something they couldn’t have known unless they were guilty, or at least complicit, they can bargain for a reduced sentence.  Otherwise, no deals.

Wouldn’t this merely amplify the problem of mass incarceration?  The opposite is true.  Right now, prosecutors only take 3% of cases to trial.  They handle the other 97% with plea bargaining.  Under my rule, prosecutors would have to take most cases to trial in order to secure a punishment.  And since they lack the resources to conduct vastly more trials, prosecutors would wind up dropping vast numbers of weak and low-priority cases.  For the individual criminal, plea bargaining is a way to avoid a harsh sentence.  For the criminal justice system, however, plea bargaining is a way to make harsh sentences the norm – for innocent and guilty alike.