Here is one story about how the criminal justice system works:
There are criminals and there’s the rest of us. Pobody’s nerfect, and yeah, who hasn’t driven a few miles over the speed limit, or passed around a joint at a party, but most of us have nothing really to fear from the criminal law, even if it were fully enforced. If you do commit a criminal offence, and you’re caught, then you’ll be charged by a prosecutor, tried in a court where your lawyers and the government’s lawyers will compete on a level playing field (e.g. with the same evidence), and then possibly convicted, where your sentence will be commensurate with the gravity of your offence. There may be deviations from this in practice, but these are due to more general, society-wide problems like racial bias. They’re not endemic to the criminal justice system as such.
Here is another, very different story:
Most people have committed crimes for which imprisonment is a possible punishment — around 70% of us, seems to be the consensus.1 The sorting of people into the damned and the elect — the brutalized and ruined, on one hand, and those who can look back with merry bemusement at their youthful indiscretions, on the other — is largely at the discretion of police and prosecutors. As Bill Stuntz once put it, criminal statutes are mainly “a means of empowering prosecutors'“, and serve “as items on a menu from which the prosecutor may order as she wishes.”2 Trials are the exception, rather than the rule; 95% of criminal convictions are pursuant to plea agreements. The playing field is massively unequal at this stage, with prosecutors enjoying tremendous legal and institutional advantages in the gathering of evidence, and, at least in the US, no legal obligation to share impeaching evidence with defendants.3 While the question of when a punishment fits a crime is a difficult one in moral philosophy, one could be forgiven for surmising that maybe, just maybe, the harsh sentences pursuant to trial and conviction are partly designed to channel people into the plea bargaining process, lest the criminal justice system grind to a halt. In other words, they are a tax paid, not by all but by some, and not in dollars but in freedom and dignity and bodily and psychological integrity, to keep the gears turning.


Well, I’m a heavy-handed writer, so you probably clocked it very quickly — I think the second story is more accurate. Plea bargaining is the rule, and trials are the exception. Prosecutors want defendants to plead guilty, so do the defendants’ own lawyers. They want to clear cases and maintain amicable relations across the aisle. And the system is set up in such a way that it makes sense for defendants to plead guilty. Not only are the sentences they would face after a trial much more severe than those they face upon pleading guilty, but defendants may also be induced to plead guilty to avoid pre-trial detention.
And actually, just pause for a second to consider how much of a perversion that is. Pre-trial detention is not punitive, or least it’s not intended to be punitive. The ostensible purpose is to secure a defendant’s presence at a trial — you know, like, the arena in which guilt or innocence is determined in the first place. To encourage a defendant to plead guilty by holding out the prospect of escaping pre-trial detention is a kind of cruel alchemy by which non-punishment may be converted into punishment. May police and prosecutorial discretion be ever in your favour, indeed.
So there’s a lot about plea bargaining that’s not optimal. So what are the solutions? Abolish it? The criminal justice system really would grind to a halt. Criminalize less? Well I agree with that, but I’d prefer to save that debate for another time. What solutions have credible commentators presented? Well, apologies, but you’re not going to get a solid lit review from a non-lawyer4 in a Substack named after an Olivia Rodrigo song, but here are some representative proposals floated by Third Circuit Judge Stephanos Bibas, recognized as one of the handful of leading experts on this issue:
-magistrates who would investigate and share evidence with the defence
-more judicial review of plea agreements
-appeals of prosecutorial charging decisions to supervisors
-citizens’ representatives in police and prosecutors’ offices
-more recording of interactions between prosecution and defence
-”normative grand juries” to assess sentences pursuant to plea agreements5
This is all part of rejecting and replacing what Bibas calls the “shadow of a trial” model of plea bargaining. According to such a model, the outcomes of the plea bargaining process are determined almost entirely by the probabilities of conviction and the potential severity of sentences that would result from a criminal trial — something like an “expected punishment function”.6 Bibas thinks that this model is unrealistic for reasons like the aforementioned, and so we should instead think about administering plea bargaining as its own separate universe — building it “from the ground up” as he puts it — with more monitoring, more oversight, more rules, and less unrestrained, well, bargaining.
I don’t know about you, but these sorts of proposals strike me as piecemeal half-measures. One worries that they will effect little fundamental change, all the while gumming up the plea bargaining process so as to compromise its primary advantages — namely, its speed and efficiency. Maybe this is just the natural reaction of philosophers; we want swaggering juggernaut solutions — idealism, or skepticism, or nihilism, or teleology, or “joints in nature”. So maybe we’ve developed a bit of a tin ear for good, solid, in-the-trenches policymaking. But if this is what the leading experts are coming up with, I think it couldn’t hurt to stir the pot a bit with a moonshot-type solution. And so here’s my proposal for a more comprehensive reform that, at the same time, preserves what is best about the “Shadow of a Trial” model.
It works like this: As part of tendering a plea offer, prosecutors must include a prediction, an estimate, of the probability of conviction at trial for each offence charged. There is a lower bound on these probabilities, such that if prosecutors do not estimate a probability greater than this lower bound, they cannot charge this offence. You can’t, as a prosecutor, convert “Eh, there may be something there; let’s see if this scares him…” into a guaranteed conviction. You have to spare the innocent and seek to punish the almost-certainly guilty, not sell “imprisonment insurance” to the disfavoured. This serves to prevent prosecutors from knowingly charging offences for which the evidence is very weak, assuming they are honest with the defence about the probability of conviction.
But what if they are not honest? What is to prevent prosecutors from tossing out absurdly high probabilities of conviction, whatever the strength of the evidence — in other words, bluffing? My proposal is that the probabilities cited as part of the plea offer should be compared to the actual rates of conviction for charges of this type in cases that end up going to trial. And if the gap between predicted rates of conviction and actual rates of conviction is too great, this will trigger greater administrative and judicial review for the prosecutor’s office — perhaps of the sort that other scholars of the criminal law have suggested. Or perhaps prosecutorial funding might hinge on the same. Or there will be some scheme of sentence-reductions that takes effect. The point is: there are strong incentives against bluffing.
This proposal has the effect of tethering the outcomes of plea negotiations to the actual quality of the evidence against the defendant — in a way that is comprehensive rather than piecemeal, and in a way that generally preserves the speed and efficiency of the plea bargaining process. In poker terms, it forces the prosecution to either show its hand or fold — not bluff the defence into a corner with a weak pair. In Bibas’s terms, it places the plea bargaining process back in the “shadow of the trial”.
Obviously there are details to be worked out. I gestured at penalties for too great a “gap” between prosecutors’ predictions of likelihoods and actual conviction rates, but to be honest, I’m not well-versed enough in institutional design to know which penalties would work the best. And there’s the question about how to measure the “gap”. It seems like we should give extra weight to conviction rates in cases in which prosecutors offered predictions but the defendant ultimately opted for trial anyway.
But the general idea, which I think deserves further exploration in the present context and elsewhere, that it is better to refine or train an institution by appetitive or aversive outcomes, and then let the actors/learners adapt with an eye towards achieving the former and avoiding the latter, than it is to train an institution by standing over it and supervising and regulating and second-guessing it, like an anxious parent teaching a kid to ride a bike or do long division. Institutions, like children, don’t learn by being hovered over. They learn by touching the flaming stove of reality, or by betting real money and losing.
See, e.g., Douglas Husak, Overcriminalization (2007), p. 24.
Stuntz, “Plea Bargaining and Criminal Law's Disappearing Shadow,” 117 Harvard Law Review 8 (2004).
U.S. v. Ruiz, 536 U.S. 622 (2002).
Full disclosure: I did go to law school, and interned at a prosecutor’s office and a public defender’s office.
Bibas, “Designing Plea Bargaining from the Ground Up: Accuracy and Fairness without Trials as Backstops,” 57 William and Mary Law Review 4 (2016).
Bibas, “Plea Bargaining Outside the Shadow of a Trial,” 117 Harvard Law Review 8 (2004).
This reminds me of a case from my (our?) debate days. I can't remember if I ran it, or if somebody else did, or if nobody did and it's just an idea I heard.
The idea was that when someone agrees to a plea deal, and the judge would like to accept the plea bargain, there's some probability n that the plea nevertheless must be rejected. If the plea is rejected, then the state has to make their case at trial. Then, the fewer of these cases the state wins--ie, the lower their rate of winning when forced to go to trial instead of plead--the higher n becomes, and vice versa. So the state has an incentive not to overcharge, because the more they overcharge and lose, the less they get to use the plea bargaining process at all.