The devil is the details, most violent crimes aren't violent, likewise sex crimes. For example Puffy was convicted on a sex crime though his actual crime was illegally reimbursing business travel. Epstein was convicted of a sex crime though his actual crime was an illegal employment practice. Most drug crimes are in actuality illegal business selling something that is often legal to possess.
Also those numbers are bunk, sex crimes account for way more than 3% of prisoners to the point the Feds, and done states, have build entire dedicated prisons JUST for sex criminals to segregate them.
There's an easy way of getting rid of most drug offenses. Legalize, and I don't mean this garbage of decriminalization which increases demand and makes the externalities caused by illegality worse. I mean being able to go to a doctor and getting a prescription for whatever. I would gladly have all this paid by Medicaid.
NHS in Britain used to do this with Heroin. They stopped during the 1980's in a fit of morality. Some experts thought there was too much heroin being prescribed.
[Me, personally, I think we got too much stuff. Let's prohibit it. I am an expert in consumption.]
Prescriptions aren't legalization, legalization means my kids can buy it McDonald's, i.e. putting the coke back on coke.
Case in point, you can already get actually factually methamphetamine, cocaine, and fentanyl with a prescription in the US, the problem is no doctor will prescribe them; same problem with most drugs. Actual legalization is the answer, not morality gatekeeping doctors who forget their place and job.
These two posts are a perfect example of the bait-and-switch supporters of "drug legalization" always use.
You both claim to support "drug legalization" despite not agreeing on what it means, but by being deliberately ambiguous between the two meanings you get to create the false impression that there is a policy that gets rid of the problems caused by the war on drugs without creating problems associated with widespread addiction.
Don't look at me! I said precisely what I meant. All I can be accused of is "using a bad word", like six year olds and now adults accuse each other of.
In substance, I object to being corralled into an orthodoxy. I don't mean this personally, but this is a problem that afflicts US libertarians.
Relying on Hayek, a welfare state does not restrict liberty.
Also addiction won't move in a meaningful manner. Addiction is extremely rare and mostly depends on genetics hence supply doesn't matter for addiction. And drug use won't budge, nearly all Americans have already used illegal drugs and places where drugs become normalized their usage goes down as they aren't cool anymore, queue the plummetting modern alcohol usage rates and likewise barely a budge up in marijuana usage in places that have regulated it (I say that because I'm not sure ANYWHERE in the US has legalized it).
> Also addiction won't move in a meaningful manner. Addiction is extremely rare and mostly depends on genetics hence supply doesn't matter for addiction. And drug use won't budge, nearly all Americans have already used illegal drugs and places where drugs become normalized their usage goes down as they aren't cool anymore,
It's not bait and switch at all. I believe in liberty, there is nothing under natural law or the constitution that gives the government the right to ban ANY drug. So no when I say drug legalization I mean legalization in the normal sense, like water. Frank on the other hand doesn't, when he says legalization he means regulation which isn't the same thing.
There are laws that will string up a doctor for prescribing in certain ways. The fentanyl disaster began with such laws, so those who wanted the drug switched to the illegal market, where quality, dose, and hygiene couldn't be controlled.
The problem is that unless we also eliminate the welfare state people being free to make bad decisions means everyone else ends up paying to clean up the resulting mess.
I used to have the mantra "legalize and tax". A staunchly libertarian friend asked "why tax". To reduce demand and suicides. The Abrahamic religions forbid suicide. We can call that a cultural feature that has survived.
The tax part has become questionable since illegal drugs are cooked in labs and strength is enormous compared to what was once the norm. They are cheap to smuggle. Legal drugs, conforming to FDA rules, will never be cheap enough to outcompete illegal drugs, where we don't know strength, dose, and so on.
So, subsidize consumption of legal drugs through Medicaid! But it has to be done in a controlled manner. It's a way of preventing accidental suicides.
Prison is a bad metric, people overly focus on it. You need to include probation, parole, convictions, and registries. Most innocent people required to register for example would happily take a reasonable amount of time in prison over $2 fine and lifetime registry and yet the plea offered is normally a $2 fine lifetime registry OR max sentence at trial and then when you get out, still lifetime registry. And then have people on lifetime parole, lifetime probation, etc. You should be tracking scarlet letters, not incarceration as it vastly understates the problem.
(Edit: I can't find the data I seen years ago, might have been counties, not states. Removed this section)
Puffy is in prison for giving a professional sex worker a free ride. Martha Steward did time for lying about a crime she didn't do. Our President is a felon and yet somehow mandatory legal prohibitions on felons don't apply to him. etc.
But how many of them are in for literally, things I don't think should be a crime. I don't expect an answer. I am just sorta wondering with no way to figure it out.
Maybe we should make a list, things that lead to prison/jail/court supervision, that we don't think should be crimes.
The problem is that's nearly all of them and no two people are going to agree on what should remain outside the basics of common law; which we violently rejected in the early 20th century with the Progressive Era.
The problem is not the bargaining over the trial per se, it is over the fact that poor innocents can spend years in prison awaiting trial and richer ones can be found innocent and yet be ruined financially (often after years awaiting trial).
There are lots of things wrong with our legal system. This post was on a much narrower subject — how plea bargaining could make a legal system work better, why in our system it doesn't, and how that could be changed.
I would argue that, apart from the cost issue in my other reply, the problem with your proposal is that people who are acquitted are not normally compensated and certainly they almost never come out better off than before no matter how weak the case against them. You have tried to address the incentives from the POV of the prosecutor but that doesn't help the innocents who go to trial and are acquitted and so doesn't affect their decision when plea bargaining - They still lose even when they win and you are advocating for more trials and hence more losers even if the proportions improve
Taking imprisonment whilst awaiting trial: It is not uncommon for people to be convicted at trial and yet serve little or no extra time because the time spent waiting is greater than the sentence on conviction. In this case, the guilt or innocence is irrelevant to the decision to accept a plea bargain.
Or, consider cost: The cost of a defence (in criminal cases) is, or should be, independent of your guilt or innocence. It may be better to have money and a criminal record than no money and no criminal record. This doesn't eliminate the value of the trial verdict but it makes it a much smaller part of the overall value proposition thus, again, making guilt or innocence less relevant to the plea bargaining decision.
(And I'm not even going to go into the inconvenient truth that spending on lawyers IS correlated with desirable [for the defendent] outcome).
Time spent in custody, and legal costs, are economic losses and you are explicitly advocating for more trials and hence greater economic loss.
That slowness is one of the worst things about our judicial system. Eugene Volokh likes to say lawyers' true super power is the ability to turn every question into a question of procedure, and it shows, with it seemingly common on his blog for an appeals court deciding whether a district judge got some piffling motion wrong, several years after the charges were filed, and they probably won't get to trial for several more years.
“It probably does not increase the accuracy of the system, the extent to which the guilty are punished and the innocent not punished, may well decrease it.”
I followed all of this piece save this bit about “may well decreases it”.
Of course anything “may”, but why is it that it is likely (“may well”) to decrease the accuracy of the system?
If your point is that it may well increase the latter, ok, but I fail to see how it likely reduces the extent to which the guilty are punished.
Not to mention that changing to a system with a lot more trials, will - given the cost in both elapsed time and in terms of the talent of the lawyers on each side - likely have a material effect in decreasing the accuracy of the system.
I think the likelihood of innocent people being punished increases due to prosecutors offering plea deals with 1) bonus charges and 2) slightly less costly expected value.
For 1, prosecutors often will throw in a list of stuff they threaten to charge you with, knowing that while they could never make those charges stick, it raises the possible costs and makes the innocent worried about how badly things could go. Especially if they are at all afraid of being convicted though innocent, which of course happens. If the prosecutor had to go to trial, none of those pointless charges would ever come up, but the defendant offered the deal doesn't know that. They might plead guilty to something they never did just to avoid the process as punishment aspect, as well as the outside chance of conviction.
For 2, it is related to 1, how confident is the innocent that they will not be wrongly convicted, that the prosecutor won't cheat, that they can afford the defense without bankrupting their family, etc. How often does a prosecutor pressure a defendant and just throw a plea deal out to see if he gets lucky, knowing he probably can't convict, and gets lucky?
I did not mean to suggest that it made it less likely that the guilty would be punished. I was implying some measure of the summed error rate without being very precise about it.
My qualm was not with the claim that in the current system innocent people may well be more likely to be punished. It was with the claim that the extent to which the guilty are punished is likely (“may well” be) decreased.
Oh, that is easier, though. Say a DA has 100 cases a month, staff and court room time for maybe 20. He decides to plead out as many as he can to get his case clearance rate up. Some he might plead out roughly the same punishment, say 80%, but that rate goes down until he is pleading people for possession of stolen property instead of grand theft auto and car jacking.
Now, the real fix would be for the government to pay attention to what its actual job is and spend more money on courts and criminal justice such that he has the bandwidth to try all 100 cases. The DA’s incentives given his situation however is to just get as much through as possible and only try the big cases that help his career.
> Now, the real fix would be for the government to pay attention to what its actual job is and spend more money on courts and criminal justice such that he has the bandwidth to try all 100 cases.
That's how things used to work. When the seventh amendment was passed, the average trial lasted about an hour. Then the bleeding hearts insisted on giving everyone the right of infinite appeals and other measures designed to decrease the rate of false convictions that had the effect making the process longer and more expensive.
In part, but I was more referring to how long things take to actually get to trial in the first place, and how long they drag on. A big part is just how long it takes to find space in a court room. There just don’t seem to be enough judges and prosecutors to cover requirements.
It's a crossover error rate issue. A 2x2 of Accept vs Reject and Innocent vs Guilty. The current plea bargain system is pretty good at reducing false negatives and very bad at reducing false positives, so it is extremely inaccurate despite having a much lower rate of one of those types of error.
Sorry, maybe you just misread that specific paragraph? That one was about the current plea bargain system, not the proposed one.
“The current plea bargain system is pretty good at reducing false negatives and very bad at reducing false positives”
Possibly I don’t understand which you mean by false positives and false negatives.
Either way I don’t see how one can be at all confident of DF’s claim that the current plea bargain system increases the number of innocent falsely punished. (I of course concede that it *might* be true.)
Unless one is counting solely the numbers of innocent people punished rather than the total cost (in years served or fines paid) of innocents punished. In that case obviously allowing an innocent person to choose a sure lesser punishment over the possibility of false conviction for a far worse sentence will result in greater pure number of individuals falsely punished.
A false positive is concluding that someone is guilty when he isn't. A false negative is concluding that someone is not guilty when he is.
As it happens I have an unfinished post on false positives in an entirely different contest — the tendency of humans to see patterns that are not really there. I need more examples.
This looks a lot like a low-base-rate binary classification problem. Suppose you want to find internet articles about chinchillas. Almost all articles are negatives, so you do not start with the most accurate and expensive review on everything. You use a funnel: keyword rules and crude statistics first, then small models, then expensive models or humans at the end. In other words, high-recall low-cost filters first, high-precision high-cost checks last.
The important part is what happens over time. Each layer only sees data already filtered by the previous layer. So if you change one upstream component, downstream components can suddenly get out-of-distribution inputs and behave badly.
A standard robustness trick is to use multiple tools at each stage with similar average performance but different, ideally weakly correlated, error modes. In practice that means models trained differently, rules written by different people, different heuristics, and so on.
What is interesting to me is that the legal system seems to naturally have some of these properties: different police departments produce different cases, different prosecutors bargain differently, judges make different evidentiary decisions, and juries are inconsistent in their own ways. That sounds messy, but it also creates the kind of variation that can make local changes and experiments more robust than in a fully homogenized pipeline.
Obviously the legal system is much less controlled than the ML settings I work with, but that is exactly why I am surprised lawyers and judges do not seem more interested in this kind of systems view.
Separate question: this kind of funneling and selective filtering also happens in science. Should we have an equivalent of plea bargaining for papers? Maybe a fine for a rejected paper proportional on the scale of the claim becoming a default if paper rejected by 3 reviewers. First reviewer can offer to reject you for only a fraction if you accept. If not next 2 check?
Another problem with the current system is that the trial is a punishment in itself, whether the accused is convicted or not. The medieval system was better in this respect, assuming it was indeed well rigged.
Game Theoretically Sound plea bargaining is both funny and an improvement. It is slightly worrisome because this type of weighting makes the most sense when adversarial epistemics have frustrated any better means of discerning something. It has the additional advantage, however, of being easily measurable by any outsider, with a few extra assumptions about available information to different parties. Some of which, interestingly, exist already to different degrees in different times and places. Similar structures also allow for a formalizable and hence accountable system of conditional sovereignty: can you beat the GTO strategy? So yes, this could be an incentive and structures issue, and it could be solved by additional and particular formal processes. Thank you. I know this post doesn't get into questions of private legal systems but I feel like it offers a productive place to go with developing those. Which I assume already has literature about it, I don't know, I haven't read anything in half a decade.
I have one book in part about how a modern private legal system might work and another with one chapter on the logic of stateless legal system and several describing real world examples.
“when adversarial epistemics have frustrated any better means of discerning something” why are you so confident that this is in fact the case?
While I’m not exactly sure what you mean here, this assertion seems to imply that you are ignoring the fundamental reality behind public choice theory. I.e. that government officials themselves act based on their personal incentives.
Which given your reference to incentives in the following paragraph seems particularly odd.
I wonder how many people are in prison for things I don't think should be crimes.
...Or, of course, we could simplify our legal system to make fewer things illegal...
62% for violent offences, 14% property offenses and 20% drug offenses. (US prisons, Wikipedia)
Since you follow Friedman, I'm guessing some number greater than 20%.
The devil is the details, most violent crimes aren't violent, likewise sex crimes. For example Puffy was convicted on a sex crime though his actual crime was illegally reimbursing business travel. Epstein was convicted of a sex crime though his actual crime was an illegal employment practice. Most drug crimes are in actuality illegal business selling something that is often legal to possess.
Also those numbers are bunk, sex crimes account for way more than 3% of prisoners to the point the Feds, and done states, have build entire dedicated prisons JUST for sex criminals to segregate them.
There's an easy way of getting rid of most drug offenses. Legalize, and I don't mean this garbage of decriminalization which increases demand and makes the externalities caused by illegality worse. I mean being able to go to a doctor and getting a prescription for whatever. I would gladly have all this paid by Medicaid.
NHS in Britain used to do this with Heroin. They stopped during the 1980's in a fit of morality. Some experts thought there was too much heroin being prescribed.
[Me, personally, I think we got too much stuff. Let's prohibit it. I am an expert in consumption.]
Prescriptions aren't legalization, legalization means my kids can buy it McDonald's, i.e. putting the coke back on coke.
Case in point, you can already get actually factually methamphetamine, cocaine, and fentanyl with a prescription in the US, the problem is no doctor will prescribe them; same problem with most drugs. Actual legalization is the answer, not morality gatekeeping doctors who forget their place and job.
These two posts are a perfect example of the bait-and-switch supporters of "drug legalization" always use.
You both claim to support "drug legalization" despite not agreeing on what it means, but by being deliberately ambiguous between the two meanings you get to create the false impression that there is a policy that gets rid of the problems caused by the war on drugs without creating problems associated with widespread addiction.
Don't look at me! I said precisely what I meant. All I can be accused of is "using a bad word", like six year olds and now adults accuse each other of.
In substance, I object to being corralled into an orthodoxy. I don't mean this personally, but this is a problem that afflicts US libertarians.
Relying on Hayek, a welfare state does not restrict liberty.
Also addiction won't move in a meaningful manner. Addiction is extremely rare and mostly depends on genetics hence supply doesn't matter for addiction. And drug use won't budge, nearly all Americans have already used illegal drugs and places where drugs become normalized their usage goes down as they aren't cool anymore, queue the plummetting modern alcohol usage rates and likewise barely a budge up in marijuana usage in places that have regulated it (I say that because I'm not sure ANYWHERE in the US has legalized it).
> Also addiction won't move in a meaningful manner. Addiction is extremely rare and mostly depends on genetics hence supply doesn't matter for addiction. And drug use won't budge, nearly all Americans have already used illegal drugs and places where drugs become normalized their usage goes down as they aren't cool anymore,
Ok, this is just contrarian idiocy.
It's not bait and switch at all. I believe in liberty, there is nothing under natural law or the constitution that gives the government the right to ban ANY drug. So no when I say drug legalization I mean legalization in the normal sense, like water. Frank on the other hand doesn't, when he says legalization he means regulation which isn't the same thing.
Let's be just like the commies have always been and write dictionaries. Real scientific progress!
There are laws that will string up a doctor for prescribing in certain ways. The fentanyl disaster began with such laws, so those who wanted the drug switched to the illegal market, where quality, dose, and hygiene couldn't be controlled.
McDonalds, no, but Medicaid, yes.
People should be free to make bad decisions, why not have them feely available in supermarkets?
I can buy enough liquor to kill myself in most stores.
> People should be free to make bad decisions
The problem is that unless we also eliminate the welfare state people being free to make bad decisions means everyone else ends up paying to clean up the resulting mess.
I used to have the mantra "legalize and tax". A staunchly libertarian friend asked "why tax". To reduce demand and suicides. The Abrahamic religions forbid suicide. We can call that a cultural feature that has survived.
The tax part has become questionable since illegal drugs are cooked in labs and strength is enormous compared to what was once the norm. They are cheap to smuggle. Legal drugs, conforming to FDA rules, will never be cheap enough to outcompete illegal drugs, where we don't know strength, dose, and so on.
So, subsidize consumption of legal drugs through Medicaid! But it has to be done in a controlled manner. It's a way of preventing accidental suicides.
Prison is a bad metric, people overly focus on it. You need to include probation, parole, convictions, and registries. Most innocent people required to register for example would happily take a reasonable amount of time in prison over $2 fine and lifetime registry and yet the plea offered is normally a $2 fine lifetime registry OR max sentence at trial and then when you get out, still lifetime registry. And then have people on lifetime parole, lifetime probation, etc. You should be tracking scarlet letters, not incarceration as it vastly understates the problem.
(Edit: I can't find the data I seen years ago, might have been counties, not states. Removed this section)
Puffy is in prison for giving a professional sex worker a free ride. Martha Steward did time for lying about a crime she didn't do. Our President is a felon and yet somehow mandatory legal prohibitions on felons don't apply to him. etc.
There are no such states. The highest incidence is in Georgia, at 5.5% of the population. Nationwide it's 2.7%. Source is google.
Deleted as can't find and don't want to spend more than 15 minutes on it, might have misremembered, I edited and deleted. Even 5.5% is insane though.
I used your terms to search: "court supervision of one sort or another".
Sure, given.
But how many of them are in for literally, things I don't think should be a crime. I don't expect an answer. I am just sorta wondering with no way to figure it out.
Maybe we should make a list, things that lead to prison/jail/court supervision, that we don't think should be crimes.
The problem is that's nearly all of them and no two people are going to agree on what should remain outside the basics of common law; which we violently rejected in the early 20th century with the Progressive Era.
The problem is not the bargaining over the trial per se, it is over the fact that poor innocents can spend years in prison awaiting trial and richer ones can be found innocent and yet be ruined financially (often after years awaiting trial).
There are lots of things wrong with our legal system. This post was on a much narrower subject — how plea bargaining could make a legal system work better, why in our system it doesn't, and how that could be changed.
My assertion is that these factors are so large that they make your model too simple to support your proposals for changing plea bargaining.
The problems you and Eugene describe may be real but why do they undermine my proposal?
I would argue that, apart from the cost issue in my other reply, the problem with your proposal is that people who are acquitted are not normally compensated and certainly they almost never come out better off than before no matter how weak the case against them. You have tried to address the incentives from the POV of the prosecutor but that doesn't help the innocents who go to trial and are acquitted and so doesn't affect their decision when plea bargaining - They still lose even when they win and you are advocating for more trials and hence more losers even if the proportions improve
Taking imprisonment whilst awaiting trial: It is not uncommon for people to be convicted at trial and yet serve little or no extra time because the time spent waiting is greater than the sentence on conviction. In this case, the guilt or innocence is irrelevant to the decision to accept a plea bargain.
Or, consider cost: The cost of a defence (in criminal cases) is, or should be, independent of your guilt or innocence. It may be better to have money and a criminal record than no money and no criminal record. This doesn't eliminate the value of the trial verdict but it makes it a much smaller part of the overall value proposition thus, again, making guilt or innocence less relevant to the plea bargaining decision.
(And I'm not even going to go into the inconvenient truth that spending on lawyers IS correlated with desirable [for the defendent] outcome).
Time spent in custody, and legal costs, are economic losses and you are explicitly advocating for more trials and hence greater economic loss.
That slowness is one of the worst things about our judicial system. Eugene Volokh likes to say lawyers' true super power is the ability to turn every question into a question of procedure, and it shows, with it seemingly common on his blog for an appeals court deciding whether a district judge got some piffling motion wrong, several years after the charges were filed, and they probably won't get to trial for several more years.
That's not the current biggest problem with the legal system.
The biggest problem is that too many violent criminals are given slaps on the wrist, if even that, and put back on the street.
“It probably does not increase the accuracy of the system, the extent to which the guilty are punished and the innocent not punished, may well decrease it.”
I followed all of this piece save this bit about “may well decreases it”.
Of course anything “may”, but why is it that it is likely (“may well”) to decrease the accuracy of the system?
If your point is that it may well increase the latter, ok, but I fail to see how it likely reduces the extent to which the guilty are punished.
Not to mention that changing to a system with a lot more trials, will - given the cost in both elapsed time and in terms of the talent of the lawyers on each side - likely have a material effect in decreasing the accuracy of the system.
I think the likelihood of innocent people being punished increases due to prosecutors offering plea deals with 1) bonus charges and 2) slightly less costly expected value.
For 1, prosecutors often will throw in a list of stuff they threaten to charge you with, knowing that while they could never make those charges stick, it raises the possible costs and makes the innocent worried about how badly things could go. Especially if they are at all afraid of being convicted though innocent, which of course happens. If the prosecutor had to go to trial, none of those pointless charges would ever come up, but the defendant offered the deal doesn't know that. They might plead guilty to something they never did just to avoid the process as punishment aspect, as well as the outside chance of conviction.
For 2, it is related to 1, how confident is the innocent that they will not be wrongly convicted, that the prosecutor won't cheat, that they can afford the defense without bankrupting their family, etc. How often does a prosecutor pressure a defendant and just throw a plea deal out to see if he gets lucky, knowing he probably can't convict, and gets lucky?
Your 1 was what I was thinking of.
This I understood from the beginning.
What I don’t understand is the claim that the current system likely (“may well”) decreases the extent to which the guilty are punished.
Or was your initial wording simply vague, and you didn’t mean to suggest that part?
That part was my question/claim.
I did not mean to suggest that it made it less likely that the guilty would be punished. I was implying some measure of the summed error rate without being very precise about it.
Compared to what? Some hypothetical world where the judicial system has the infinite resources necessary to bring every case to trial?
Why would it be infinite resources? There isn’t that much crime. You are rather over stating that I think.
My qualm was not with the claim that in the current system innocent people may well be more likely to be punished. It was with the claim that the extent to which the guilty are punished is likely (“may well” be) decreased.
Oh, that is easier, though. Say a DA has 100 cases a month, staff and court room time for maybe 20. He decides to plead out as many as he can to get his case clearance rate up. Some he might plead out roughly the same punishment, say 80%, but that rate goes down until he is pleading people for possession of stolen property instead of grand theft auto and car jacking.
Now, the real fix would be for the government to pay attention to what its actual job is and spend more money on courts and criminal justice such that he has the bandwidth to try all 100 cases. The DA’s incentives given his situation however is to just get as much through as possible and only try the big cases that help his career.
> Now, the real fix would be for the government to pay attention to what its actual job is and spend more money on courts and criminal justice such that he has the bandwidth to try all 100 cases.
That's how things used to work. When the seventh amendment was passed, the average trial lasted about an hour. Then the bleeding hearts insisted on giving everyone the right of infinite appeals and other measures designed to decrease the rate of false convictions that had the effect making the process longer and more expensive.
In part, but I was more referring to how long things take to actually get to trial in the first place, and how long they drag on. A big part is just how long it takes to find space in a court room. There just don’t seem to be enough judges and prosecutors to cover requirements.
> A big part is just how long it takes to find space in a court room.
Which brings us back to trial lasting longer, hence creating a backlog of cases awaiting trial.
It's a crossover error rate issue. A 2x2 of Accept vs Reject and Innocent vs Guilty. The current plea bargain system is pretty good at reducing false negatives and very bad at reducing false positives, so it is extremely inaccurate despite having a much lower rate of one of those types of error.
Sorry, maybe you just misread that specific paragraph? That one was about the current plea bargain system, not the proposed one.
“The current plea bargain system is pretty good at reducing false negatives and very bad at reducing false positives”
Possibly I don’t understand which you mean by false positives and false negatives.
Either way I don’t see how one can be at all confident of DF’s claim that the current plea bargain system increases the number of innocent falsely punished. (I of course concede that it *might* be true.)
Unless one is counting solely the numbers of innocent people punished rather than the total cost (in years served or fines paid) of innocents punished. In that case obviously allowing an innocent person to choose a sure lesser punishment over the possibility of false conviction for a far worse sentence will result in greater pure number of individuals falsely punished.
A false positive is concluding that someone is guilty when he isn't. A false negative is concluding that someone is not guilty when he is.
As it happens I have an unfinished post on false positives in an entirely different contest — the tendency of humans to see patterns that are not really there. I need more examples.
Clouds!
Perhaps a view from ML.
This looks a lot like a low-base-rate binary classification problem. Suppose you want to find internet articles about chinchillas. Almost all articles are negatives, so you do not start with the most accurate and expensive review on everything. You use a funnel: keyword rules and crude statistics first, then small models, then expensive models or humans at the end. In other words, high-recall low-cost filters first, high-precision high-cost checks last.
The important part is what happens over time. Each layer only sees data already filtered by the previous layer. So if you change one upstream component, downstream components can suddenly get out-of-distribution inputs and behave badly.
A standard robustness trick is to use multiple tools at each stage with similar average performance but different, ideally weakly correlated, error modes. In practice that means models trained differently, rules written by different people, different heuristics, and so on.
What is interesting to me is that the legal system seems to naturally have some of these properties: different police departments produce different cases, different prosecutors bargain differently, judges make different evidentiary decisions, and juries are inconsistent in their own ways. That sounds messy, but it also creates the kind of variation that can make local changes and experiments more robust than in a fully homogenized pipeline.
Obviously the legal system is much less controlled than the ML settings I work with, but that is exactly why I am surprised lawyers and judges do not seem more interested in this kind of systems view.
Separate question: this kind of funneling and selective filtering also happens in science. Should we have an equivalent of plea bargaining for papers? Maybe a fine for a rejected paper proportional on the scale of the claim becoming a default if paper rejected by 3 reviewers. First reviewer can offer to reject you for only a fraction if you accept. If not next 2 check?
Another problem with the current system is that the trial is a punishment in itself, whether the accused is convicted or not. The medieval system was better in this respect, assuming it was indeed well rigged.
Game Theoretically Sound plea bargaining is both funny and an improvement. It is slightly worrisome because this type of weighting makes the most sense when adversarial epistemics have frustrated any better means of discerning something. It has the additional advantage, however, of being easily measurable by any outsider, with a few extra assumptions about available information to different parties. Some of which, interestingly, exist already to different degrees in different times and places. Similar structures also allow for a formalizable and hence accountable system of conditional sovereignty: can you beat the GTO strategy? So yes, this could be an incentive and structures issue, and it could be solved by additional and particular formal processes. Thank you. I know this post doesn't get into questions of private legal systems but I feel like it offers a productive place to go with developing those. Which I assume already has literature about it, I don't know, I haven't read anything in half a decade.
I have one book in part about how a modern private legal system might work and another with one chapter on the logic of stateless legal system and several describing real world examples.
“when adversarial epistemics have frustrated any better means of discerning something” why are you so confident that this is in fact the case?
While I’m not exactly sure what you mean here, this assertion seems to imply that you are ignoring the fundamental reality behind public choice theory. I.e. that government officials themselves act based on their personal incentives.
Which given your reference to incentives in the following paragraph seems particularly odd.
How does literally anyone ascribe a level of confidence to my statements?
Also I must indeed be missing some colossal tower of assumptions because I have no idea what the relevance of this response even is.