Once upon a time I wanted to become a professional philosopher. In the process of seeking this goal, I generated a bunch of manuscripts in varying states of completeness. Some of them cover ideas I have talked about on the blog, others are different. I’m planning to This is one such article
Philosophy Bear is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Punishment at scale and positive retributivism
Abstract: I suggest on the basis of a thought experiment involving large numbers of individuals being punished simultaneously an intuition that considerations of purely retributive justice are never sufficient to justify accidentally punishing the innocent. Since punishment in the real world necessarily involves punishing the innocent by mistake at least occasionally, purely retributive punishment is never permissible.
Introduction
“Finally, Abraham said, “May the Lord not be angry, but let me speak once more. Suppose ten are found there?”
And He answered, “On account of the ten, I will not destroy it.”
–Genesis 32: 18
I will make an epistemological argument against positive retribution. A similar appeal to uncertainty against positive retribution has been made previously by Caruso (2020). However, while Caruso’s argument hinged on uncertainty about metaphysical and moral psychological facts, my argument is built on uncertainty about facts in general, and while Caruso’s argument hinged on substantial uncertainty, mine is concerned with anything less than absolute certainty.
While negative retributivism is the position that punishment can (at least typically) only be justified if someone has done wrong, positive retributivism is the position that a person having committed (at least certain kinds of) wrong itself gives us reason to punish them. In the words of Hoskins & Duff (2021): “Essentially, the state should punish those found guilty of criminal offences to the extent that they deserve because they deserve it.” Colloquially, the positive retributivist endorses, at least sometimes, punishment for the sake of deserts, whereas the negative retributivist asserts only that wrongdoing is a necessary condition for justified punishment. Henceforth, when I refer to retributivism without further elaboration in this piece, I mean positive retributivism.
Like many doctrines, positive retributivism has become more complex over time, and various developments and hybrid theories have developed to which our argument may not apply, or may apply only with modification.
It is worth stressing that our argument only relates to positive retributivism. For example, a moral principle that held that being sure that someone was guilty beyond a reasonable doubt, but not beyond all doubt, was a necessary but not sufficient condition punishment would be, as far as I can tell, unscathed by the intuition we claim to find. It might even be argued that negative retributivism is bolstered by our argument, insomuch as our argument rests on deep-seated intuitions about the wrongs of harming innocents not inherent to pure consequentialism.
I begin with an argument against positive retributivism in general – that retribution for its own sake can never justify harming innocents, but innocents will always be punished by any real justice system. The intuition is based in considering a case of parallel punishmentor the punishment of a large number of people at once, where at least one is innocent, and the punishment is purely retributive.
In response to some counterarguments, I retreat, for the sake of argument, to a more moderate position: viz., that it is immoral to frame a legal code which includes punishment for the sake of punishment. I conclude with a more speculative nod towards a larger principle: it is only permissible to punish the wicked in order to achieve retribution for its own sakeif doing so causes no substantial harm to any innocents – an impossible standard to meet in practice, due to the opportunity costs of punishment. The overall position that this paper develops, then, is a defence of what I term strong practical anti-retributivism, based the way scale affects our view of punishment.
The general point that punishment can catch the innocent, and this might be a problem for retributivism has been made often before (Bagaric & Amarasekara 2000). Various defences from retributivists exist, often focusing on the idea that since the punishment of the innocent is not intended, it is an excusable side effect (e.g. Duff 1986, Walen 2015). My strategy is to try to refute this by constructing a series of thought experiments involving punishing many people at once -punishment in parallel- which, on the retributivist theory, should be permissible, but which I think many people – even some people of retributivist sentiments – will baulk from endorsing.
My argumentative strategy in this work is limited in scope. I claim to derive an intuition that is incompatible with positive retribution in practically possible situations. It is open to critics to claim that such an intuition should be discarded as a result of reflective equilibrium with other intuitions, that the appeal to intuitions in ethics is misguided or at least misguided in cases like this or that they simply do not share the intuition. Empirical study might be called for on the degree to which the public shares the intuitions given here. Nonetheless, for me and anyone else who shares this intuition, and places at least some weight on intuition in ethics, we have a reason to avoid punishment for the sake of just deserts
By appealing directly to an ethical intuition, we leave open complex debates around acts and omissions and harms versus benefits that arise in this area (e.g. Walen 2015), and complex debates about the formulation of retributivism (Alexander 1983). Punishment for punishment’s sake – without any additional benefits – seems, on our intuition, to be wrong if even a single innocent person is caught up. The larger question of how to systematise a broader theory of punishment which includes this intuition and many others is left open.
The judge of the dead and retributive punishment in bulk
Imagine you are the judge of the dead. It is your job to assign departed souls to either heaven or a temporary but agonising period of purgation. By considering this remote case, we render irrelevant non-retributive purposes for punishment, such as incapacitation and deterrence and distill the retributive component. Also, its remoteness removes us from the flawed realm of real justice systems - (see for example Duff 2001, Shelby 2007 for a discussion of how this may affect the rights and wrongs of criminal justice)
One thousand sinners have been bought before you, and you are about to send them off for purgation when an angel bursts in and explains there has been a mix-up. One of these thousand people is innocent, but the angels have lost the paperwork and do not know which. Assuming you have no way of splitting out sinner from saint, should you punish them all or let them all go? Remember that punishment, in this case, will serve no purpose except for achieving ‘just deserts’. The reader may or may not be a retributivist, but I think many retributivists will feel a pull to stay their hand – preferring to let the 999 go rather than punish the innocent.
The argument seems true to me for partially retributively motivated punishment as well. Suppose, in your role of judge of the dead, a certain amount of the punishment you assign serves practical purposes – perhaps deterring those still alive – and a certain amount is purely retributive. Two options seem plausible. You may regretfully assign the 999 sinners and one innocent to that portion of the punishment which is justified with respect to its consequences (and perhaps also cleared through a negative-retributivist principle requiring a certain threshold of proof), but not to the purely retributive portion, or you may make the decision that no punishment is acceptable whatsoever. Nothing about this altered case makes the retributive portion of the punishment acceptable.
The reader may be wondering why this intuition arises. If this situation mirrors the usual situation of punishment – why hasn’t this intuition, then, been noticed previously? The answer is likely to do with quantity [in combination of course with a more prosaic feature of the thought experiment- the clear exclusion of non-retributivist motives]. If you give the final assent to punish a large number of people at once- as in our thought experiment-, the reality that almost certainly one of them is innocent (whether derived statistically or simply on the basis of common sense) becomes clearer. In the real world, almost no one passes direct, final and simultaneous judgment over a large group of people at least one which will, in expectation, almost certainly be innocent. What we do instead is run the process in serial rather than parallel.
It seems to me that this intuition holds regardless of the numbers – it is never acceptable to engage in an act that will result in an innocent person being punished in order to enable the retributive punishment of a number – however large. Retribution, qua retribution, even if attractive in principle, is simply not a good enough reason to punish an innocent if you share the intuition I’ve described, and all real systems of retribution involve punishing the innocent.
Berman (2011) distinguishes between two types of retributivism: consequentialist and non-consequentialist retributivism. Consequentialist retributivism is the view that it is a good for at least certain wrongdoers to suffer, and non-consequentialist retributivism gives a different reason for punishment (e.g., a duty, irrespective of any ‘good’ of punishment, to punish.) With respect to consequentialist retributivism, we might say that our contention is that the good of not punishing the innocent takes lexical priority over the good of punishing the guilty (if such a good exists at all). This means that the consequentialist retributivist should never punish if they are less than certain of guilt (and in practice, they always will be). With respect to non-consequentialist retributivism, the intuition we are proposing would hold that any duty to punish (indeed the very option to punish) is only present where there is a certainty, and there never is.
The argument against retributively motivated punishment in real judicial circumstances, extending on this thought experiment is:
In the circumstances I describe, we have an intuition against dispensing retributive judgment.
Judicial punishment shares with this case the morally pivotal feature that, inevitably, some innocents will be punished unless we know for a certainty that those convicted are guilty. Serial versus parallel punishment is morally irrelevant
As a result, it is never legitimate to punish anyone for retributive reasons through a judicial process unless you are certain that the accused is guilty.
Certainty – at least in matters of complex facts about the external world like those considered in judicial cases – is impossible.
Thus, it is never permissible to punish someone through a judicial process for the purpose of retribution.
The reader may find premise two overly strong. Later we will consider what happens if we relax assumption two and will argue that even if there is a morally relevant distinction between parallel and serial punishment, it would still be wrong to create a justice system like ours.
What we are trying to argue here we might term ‘practical anti-retributivism’or the view that, under prevailing circumstances, we are not justified in engaging in retributive punishment, though other conditions might justify it. The view that actual conditions do not justify retributive punishment, but other, humanly obtainable positions could, could be called ‘weak practical anti-retributivism.’ This view is, for example, held by Duff (2001). The stronger view that no humanly obtainable state of affairs would justify retributive punishment, though there are logically possible conditions under which it mightbe justified is the position of this paper, and could be called ‘strong practical anti-retributivism.’
Objections and the lawmaker for the living
I’ll now review some objections.
The first line of objection comes from the philosopher Alec Walen (personal communication 2023). Namely, there is a longstanding issue in legal theory and epistemology regarding ‘merely statistical’ evidence and whether it is permissible to form beliefs on the basis of them (Enoch et al. 2012, Silva forthcoming). Silva gives the common example of a lottery ticket to motivate the intuition that there may be something suspect about such ‘merely statistical evidence:
“You have one ticket in a very large lottery. The winning lottery number has been selected in a completely random way so that each ticket had an equal chance of being selected. You know all of this, but you have not heard the results. Reflecting on the improbability of your ticket winning, you come to believe (L) that you have a losing ticket” (Silva forthcoming)
The intuition is that forming such a belief is irrational. If this is right, our case is in trouble, because maybe the problem is not that our judge of the dead lacks certainty, it is rather that they only have statistical evidence that each of the accused is guilty.
Another objection is broadly based on the standard line of reply that retributivism has to concerns about punishing the innocent (e.g. Walen 2015, Duff 1986) and runs as follows. There is a moral difference between punishing an innocent for sure and punishing a series of one thousand people, each of which is 99.9% likely to be guilty. One, in a sense, intentionally punishes an innocent person, while the other does not. On some moral frameworks knowingly punishing at least one innocent person may be wholly different from the act of punishing many individuals, such that each is very probably guilty, but in the aggregate, it is likely (perhaps even almost certain) that one is guilty. Hence punishment in parallel where we know for a fact that a set number are innocent is morally different to the serial punishment under uncertainty we see in the real world. In this case, we have more than statistical evidence, and we are not acting with special knowledge a number of the accused are innocent, indeed we have no special knowledge that a normal court wouldn’t.
I will leave aside the truth of both these counterarguments – rather than combating them we will sidestep them with a modified case.
Imagine, once again, you are the judge of the dead. A room full of people all having been found to be worthy of punishment are gathered before your throne, and as you move to punish them, an angel hesitantly whispers in your ear that Sarandiel did the fact-finding for this lot. You catch yourself rolling your eyes. Whereas the other angels are always accurate, Sarandiel always seems to stuff things up. Though you couldn’t put an exact figure on it, and you wouldn’t say you knew it for a fact, it’s implausible that Sarandiel made no errors in judging a crowd this big. Almost certainly, at least one of them is innocent. Do you value punishment for the sake of retribution so much that you proceed? If you have the same intuition here as in the previous case, we have successfully avoided the listed objections.
Perhaps you have another objection- you think there is some real moral difference between the punishment of a large quantity of (almost all guilty) individuals and the punishment of a series of individuals who are all almost certainly guilty, and this will differentiate punishment in real legal contexts from the intuition we are appealing to. Something about punishment en masse, you think, makes the difference. Suppose it does make a difference. This might save the decisions of a judge overseeing sentencing in a single trial at a time considering positive retribution, but it would not justify the actions of the creator of the system of criminal law, who knows any punishment for just desert’s sake in the system of laws she designs will affect both the innocent and guilty alike.
An even broader intuition
Thus far we have supported the view that: An approach to punishment that results in punishment for the sake of putative ‘just deserts’ inflicted on even a single innocent person is unacceptable.
I want to put forward an even broader intuition that I hold. It would not surprise me if most philosophers do not hold this intuition. I know for a fact that many retributivists have rejected it (e.g. Duff 1986), yet I think it is worthy of further discussion than it has thus far received: Even if we believe in positive retribution in principle, it is not permissible to punish the guilty *for the sake of positive retribution* if doing so causes substantial harm to even one innocent. Or more succinctly.
Positive retribution as motivation does not meet the burden needed to impose collateral damages.
From this claim follows strong practical anti-retributivism, because punishing the guilty always causes substantial harm to the innocent. Even were no innocent people caught up, the innocent families of the guilty would be, and even if the guilty had no family, the expenditure of resources necessary for purely retributive punishment constitutes a massive cost in resources that could otherwise save or improve lives. Given the quantity of resources involved even if only a small portion of punishment can be justified only retributively, redistributing such resources would do substantial good.
Or, put in other terms, while punishment may (or may not) be an intrinsic good, the intrinsic good of punishment (if it exists) isn’t the sort of good you’re allowed to foreseeably harm innocents for the sake of – even if that harm does not come in the form of wrongful conviction and wrongful punishment.
How many will find this broader principle persuasive I am unsure, but I put it forward for discussion. Our point, both with this intuition and with the intuition we appealed to earlier in the paper might be summed like so – there is an unattractive brutality in harming the innocent as collateral just so as to punish the guilty, but this is what the retributivist must do.
Philosophy Bear is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Caruso, G. D. (2020). Justice without retribution: An epistemic argument against retributive criminal punishment. Neuroethics, 13, 13–28. https://doi.org/10.1007/s12152-018-9357-8
Duff, R. A. (1986). Trials and Punishments. Cambridge University Press.
Duff, R. A. (2001). Punishment, Communication, and Community (Revised ed. edition). Oxford University Press USA.
Walen, A. D. (2015). Proof Beyond a Reasonable Doubt: A Balanced Retributive Account (SSRN Scholarly Paper No. 2562563). https://doi.org/10.2139/ssrn.2562563
Anti-statistical arguments definitely don't work here, if anything the opposite. You don't need probabilities, merely the possibiility of harming an innocent.
By contrast, if you attach positive benefits to the act of retribution, and the probability of harming innocents is low enough, you can get a consequential case for retribution. But the desire for retribution is against the spirit of most forms of consequentialism
Agreed, and applied practically, every justice system practiced by mankind has been immoral. That said, there are still reasons to keep highly dangerous people separated from society, just as there is a reason to keep a hungry Grizzly bear separated from people. The separation isn't to show the bear it deserves to suffer, but to prevent people from the nature of the bear, and so it is with someone like a serial killer or sex offender.
Anti-statistical arguments definitely don't work here, if anything the opposite. You don't need probabilities, merely the possibiility of harming an innocent.
By contrast, if you attach positive benefits to the act of retribution, and the probability of harming innocents is low enough, you can get a consequential case for retribution. But the desire for retribution is against the spirit of most forms of consequentialism
Agreed, and applied practically, every justice system practiced by mankind has been immoral. That said, there are still reasons to keep highly dangerous people separated from society, just as there is a reason to keep a hungry Grizzly bear separated from people. The separation isn't to show the bear it deserves to suffer, but to prevent people from the nature of the bear, and so it is with someone like a serial killer or sex offender.
Great read!