Solzhenitsyn’s puzzle

And the simple step of a simple courageous man is not to take part in the lie, not to support deceit. Let the lie come into the world, even dominate the world, but not through me (Solzhenitsyn, 1972).

It is natural to read Solzhenitsyn’s remark as splitting morality into two parts. The first part concerns making a difference, changing the world, having a causal influence. It would be best if the world contained no wrongs, so it would be best if nobody committed them. If one can do anything to prevent the wrongs, or some of them, then so much the better. But what if the wrongs have been or will be committed no matter what one does? What if it is out of one’s hands? That is where the second part of morality comes into play. The second part is concerned with one’s own participation in wrongdoing. The wrongs are going to be committed anyway; there is nothing to be done about that. But one must not be implicated in them. One’s own hands must remain clean even as the world falls.

There is something puzzling about this way of dividing up morality. How does one come to be implicated in something without having made any difference to it? How can wrongs conceivably come ‘through’ me, to use Solzhenitsyn’s vivid expression, except by my making a causal contribution to their commission? How do my hands get dirty if the wrongdoing is out of my hands? Is there another mode of participation in wrongdoing other than by causal contribution? We should begin by setting aside a possible distraction. Arguably it is possible to get one’s hands dirty on the wrongs of another by profiting from them, or by helping to conceal them, or in various other ex post facto ways. If a country employs slave labour to build its infrastructure, multinational companies that take advantage of the cheapness of the infrastructure to increase their returns are arguably committing a moral wrong that involves compounding another person’s moral wrong. It is possible, although it can also be misleading, to think of this as a kind of complicity. We can think of complicity broadly as including any kind of wrongdoing that consists in any kind of association with the wrongs of another. But Solzhenitsyn already narrows the topic down to exclude so-called ‘complicity after the fact’. He is interested only in how the wrong comes into the world, not what is done with it once it is there. His suggestion is that there is a way that a wrong can come into the world through me, and make my hands dirty in the process, without my having made any difference to it.

In this paper I will suggest that there is no such way of participating in (=being complicit in, being an accomplice to, being implicated in) the wrongdoing of another. A common worry is that causal contribution is what characterises a principal wrongdoer, i.e. a wrongdoer who is not an accomplice. So if complicity is not going to collapse into principalship, some other mode of contribution—a non-causal mode—must characterise a wrongdoer who is an accomplice. I will suggest, on the contrary, that the difference between principals and accomplices is a causal difference, i.e., a difference between two types of causal contribution, not a difference between a causal and a non-causal contribution. And I will suggest that there is a moral difference that supervenes upon this causal difference. So I will agree with Solzhenitsyn’s view that morality can be divided into two parts: principalship and complicity. I will merely draw the line in a different place. Both principals and accomplices make a difference, change the world, have an influence. The essential difference between them is that accomplices make their difference through principals, in other words by making a difference to the difference that principals make.

Principals and accomplices

Let me come to the topic of complicity through another topic, at first sight only very remotely connected. Are there any wrongs that are unjustifiable and inexcusable in principle? For instance, can we imagine any adequate justifications or excuses for raping a child or massacring prisoners of war? Isn’t it the case that if one commits these wrongs, even in terrible fear of the alternatives, even in the thrall of terrible ignorance, one is automatically thereby exposed as a coward or a knave? Shouldn’t one submit to death, even to one’s own death and the death of all around one, even to the extinction of all life on earth, rather than do it? Perhaps. It is hard to know how to go about comparing the extinction of life on earth with the rape of a child or the massacre of prisoners of war. Fortunately one doesn’t need to be able to make such comparisons to come up with a general argument to challenge the idea that the commission of these latter wrongs cannot imaginably be justified or excused.

After all, one can always imagine a scenario in which one needs to commit a certain wrong in order to avoid more and/or worse commissions of the very same wrong. My sadistic kidnappers warn me (and they are not bluffing) that if I don’t torture one innocent person agonizingly to death then they will torture a hundred even more innocent people even more agonizingly to death. And if I don’t fly a large plane into a large skyscraper full of people, they will have fifty even larger planes flown into fifty even larger skyscrapers even more full of people. When all this even greater wrongdoing is perpetrated, my kidnappers tell me, it will be my fault.Footnote 1 Although they are not speaking the whole truth in saying this—obviously it will also be their fault—they are surely speaking the truth. For surely there is only one justified thing for me to do in such a case, and no apparent excuse for not doing it. I must commit the single wrong myself and thereby avoid the multiple and/or worse commissions of the same wrong by them. If that is true then there are no wrongs that cannot imaginably be justified or excused. Of every wrong it is true that one can imagine extreme circumstances in which one would be justified in committing it, and a fortiori extreme circumstances in which one would be excused. To say that a certain wrong is unjustifiable (or inexcusable) is only an emphatic way of saying that the relevant justifying (or excusing) circumstances do not in fact obtain. It is not to say that they never imaginably could.

There is much to be said about this argument before its success can be adjudicated. Here I will raise only one issue. It is the issue of whether it can be morally relevant, in determining what if anything I am justified or excused in doing, whether I would be flying the plane into the skyscraper, whether I would be doing the torturing of innocents, and so on. All else being equal, shouldn’t I care more about what I do, about my own wrongs, than about other people’s? So shouldn’t I give the avoidance of my own wrongdoing, all else being equal, more weight in my reasoning than the avoidance of wrongdoing by others? Agent-relativists are those who say that I should. Agent-neutralists are those who say that I should not. All else being equal, say agent-neutralists, it is self-indulgent or squeamish to care more about one’s own wrongs than about those of other people, just as it is self-righteous or hypocritical to care more about the wrongs of other people than about one’s own. Some agent-neutralists are strict consequentialists who think that actions can only ever be made wrong by their consequences, e.g. pain, misery, or death. But one may also believe, more plausibly, that actions such as torture and massacre are intrinsically wrong (wrong irrespective of their consequences), and yet still be an agent-neutralist about their avoidance. One may still think that what each of us should care about, and aim to achieve, is that there be less torture and less massacre in the world, never mind whether it is perpetrated by ourselves or by someone else. Under such a doctrine we are still interested in the consequences of wrongdoing, to be sure, but the consequences we are interested in are further wrongs, which are not made wrong by their consequences (and the wrongness of which is therefore missed or misrepresented by strict consequentialists).

Now an agent-neutralist can of course admit that one’s own wrongs are the ones for which one is responsible, i.e., for which one owes a justification or an excuse and in respect of which one can be at fault.Footnote 2 But even though I may not owe a justification or excuse for other people’s wrongs as such, the agent-neutralist story goes on, I may still owe a justification or excuse for my own failure to prevent other people’s wrongs. I may still be responsible for other people’s wrongs as an accomplice, a secondary wrongdoer who contributed to the commission of their wrongs. The real debate is then about whether, given the choice between my being an accomplice in wrongdoing (by failing to prevent its commission by others) and my being a principal in wrongdoing (by committing it myself), I should care more about, and hence give extra rational weight to, not committing it myself. The agent-neutralist offers a negative answer. It is no good replying on behalf of the agent-relativist that merely failing to prevent wrongdoing is not enough to make one an accomplice in that wrongdoing. The agent-neutralist will only point out that whether failing to prevent wrongdoing is enough to make one an accomplice in that wrongdoing must depend, inter alia, on whether one should care as much about the wrongs of others as one should care about one’s own wrongs—and that, of course, is exactly what is at issue.

My own sympathies have always been with the agent-neutralist. It seems to me that the avoidance of wrongdoing by anyone is fundamentally everyone’s concern. So there is no general boundary of responsibility such that merely failing to prevent another’s wrong cannot make one an accomplice. And yet a paradox seems to emerge as soon as we begin to think of the issue in terms of responsibility and fault. If I should not care about my own wrongs any more than I care about the wrongs of others, why should I care about my own responsibility or fault any more than I care about the responsibility or fault of others? ‘Torture that innocent person,’ say my kidnappers in an agent-neutralist spirit, ‘or we will torture a hundred. If the hundred get tortured, it will be your fault.’ I don’t have to be an agent-relativist to resist this line of persuasion. Instead I can simply out-agent-neutralise my captors. I can reply: ‘My fault or your fault, it’s all the same to me. After all I don’t care about what I do any more or less than I care about what you do, and that extends not only to wrongs but also to justifications and excuses. I don’t care whether I am justified or excused any more or less than I care whether you are justified or excused. I don’t care whose fault it all is. I’m not squeamish about such things.’

If this reply really contains a consistent extension of the agent-neutralist thought, then agent-neutralism has a big problem. Because now it turns out that an agent-neutralist can’t tell the story that I suggested she could ‘of course’ tell about responsibility. She can’t after all say that my morally salient connection with the wrongs of my kidnappers is as an accomplice, a secondary wrongdoer who contributes to the commission of their wrongs by failing to prevent those wrongs. It turns out that this way of looking at the problem was an agent-relativist trojan horse. For thinking about the problem in terms of complicity by failing to prevent already assumes an agent-relative standpoint. It assumes that I should care about my wrongs in a way that I shouldn’t care about the wrongs of others. Otherwise why is it morally salient that, by failing to prevent other people’s wrongs, I too do wrong? As an agent-neutralist I should simply care to prevent wrongs; whether they are mine is irrelevant. Only an agent-relativist can be animated to avoid wrongdoing by the thought of her own involvement in it, by the thought that at least some of the wrongdoing would be her own.

So it begins to look like we forced to agent-relativism by agent-neutralism itself. It looks like my kidnappers can’t get any hold over me with their agent-neutralist arguments unless I am agent-relativist enough to care especially about my own involvement; but if I am agent-relativist enough to care especially about my own involvement, their agent-neutralist arguments anyway lose their hold over me. To put it another way, it looks like one cannot be a consistent agent-neutralist. Fortunately, it only looks that way. The foregoing remarks build a paradox out of an ambiguity. There are two different senses in which particular wrongs might be ‘especially mine to care about’. Consider my promise to submit this paper for publication. My promise creates a duty on me that only I can fulfil. Of course, this doesn’t necessarily mean that only I can do the actual submitting. Maybe that can be done for me by an assistant. Whether that is true depends on how exactly we interpret the promise. But what it does mean is that, however we interpret the promise, there is some action of promise-keeping that is mine to perform (either submitting the essay myself or having it submitted for me). If I don’t do this thing, the promise wasn’t kept, and the duty it created wasn’t fulfilled. So the reason that my promise gives me to keep my promise is, as I will put it, personal in respect of conformity. Thanks to my promise there is a reason for me to conform to. What does not follow is that the reason is also personal in respect of attention, that only I should feel the force of the reason in my practical reasoning. On the contrary, it may be that everyone in the world owes my reason the same rational attention that I do. It may be that my keeping my promise is everyone’s business, so that everyone should, all else being equal, help me to do so with the same vigour as if the promise were theirs. They could in principle do this by encouraging me in my promise-keeping, or by facilitating me, reminding me, coercing me, etc. There are many possible ways for other people to give my reason the full attention it deserves, assuming that the reason is impersonal in respect of attention. The only principled limit to what they can do to help me is that they can’t actually keep the promise for me. For, to reiterate, my reason to keep my promise is personal in respect of conformity.

The debate between agent-relativists and agent-neutralists is a debate about whether reasons (or some of them) are personal in respect of attention. Agent-relativists say yes; agent-neutralists say no. Agent-relativists say that, all else being equal, others have less reason (or less compelling reason) than I have to contribute to my keeping my promise; agent-neutralists say that, all else being equal, others have as much reason (and as compelling reason) as I have to contribute to my keeping my promise.Footnote 3 But both sides can readily admit that keeping my promise is something that only I can do. Even those who insist that all reasons are impersonal in respect of attention can still consistently accept that at least some reasons are personal in respect of conformity. So the agent-neutralist victim of the sadistic kidnappers can still say that her actions are her responsibility and the actions of her kidnappers are their responsibility. All that this means is that there are some reasons for her to conform to, and other reasons for them to conform to. This is not an agent-relativist trojan horse. She is not saying, or suggesting, that in deciding what she should do she should give more emphasis or weight to what she does than to what the kidnappers do. Indeed, she could consistently assert that her main responsibility for the time being is to stop her kidnappers carrying out their nefarious plan. If she fails, she might say, she is responsible for failing to stop them committing their wrongs. She is then asserting her responsibility as an accomplice. But she is not asserting her responsibility as a principal. For the wrong she admits to is a wrong of contributing to the kidnappers’ wrongs (by failing to prevent them). It is a secondary wrong. It is her wrong but it is her wrong of failing to minimise the wrongdoing of all.

Another way of saying this is to say that, for agent-neutralists and agent-relativists alike, the distinction between principals and accomplices is embedded in the structure of rational agency. As rational beings we cannot live without it. I am responsible for what I do, and you are responsible for what you do. But on any credible view I need to give attention, in what I do, to what you will do in consequence. And you need to give attention, in what you do, to what I will do in consequence. In that sense, there are two parts of morality. There is what I should do simpliciter, and then there is what I should do by way of contribution to what you do. If I fail in the first I am a principal. If I fail in the second I am an accomplice. The truism ‘I am responsible for my actions’ cannot mean that I am responsible for my actions, never mind your actions. For my own actions inevitably include my actions of contributing to your actions. This much is (or should be) uncontentious. What is contentious is the further issue of how, in thinking about which actions I should perform, I should count those actions of yours to which I would thereby be contributing. Agent-relativists say that, all else being equal, I should count your actions for less—give them less rational attention—than I would count my own actions. Agent-neutralists say that, all else being equal, I should count them just the same.

In principle, an agent-neutralist view creates greater scope for wrongful complicity than does an agent-relativist view. On an agent-neutralist view, those who contribute to the wrongs of others can claim no allowance in respect of rational attentiveness for the mere fact that those wrongs are not their own. All else being equal, on an agent-neutralist view, I should not prefer my failing to prevent you committing a grisly murder than over my committing the same grisly murder myself. All else being equal, I shouldn’t care whether it will be you or me holding the knife. All else being equal, I am complicit in murder—in effect a murderer myself—if I don’t prevent you from murdering.

But all else is rarely equal. For by and large I am better-placed to prevent my own wrongs than I am to prevent yours. Of course there are many contingencies involved in working out how well-placed I am to prevent your wrongs. Something depends on who you are to me, what kind of wrong you are planning, where you are physically located, whether I can trust you to keep your side of the deal, how much I know about what is going on, etc. The point is only that it would often be an inefficient use of rational energy for me to pay the same rational attention to your wrongs as I pay to my own. I would be more productively employed, as the saying goes, keeping my own house in order.

This is the main consideration that militates against a radically expanded scope for complicity, even on an agent-neutralist view. It explains, in agent-neutral terms, the conspicuous element of superficial agent-relativity in everyday moral experience. Often—but not always—a mere failure to prevent a wrong committed by another is not enough to make me an accomplice to that wrong. Some more active contribution is required. This reflects the fact that, to put it crudely, stepping in to prevent another’s wrong typically has more costs and fewer benefits, even in terms of wrongdoing-avoidance itself, than does refusing to procure or enable the commission of another’s wrong. The reasons at stake here—the reasons to avoid wrongdoing—are ultimately agent-neutral. But they include agent-neutral reasons for people to reason and react agent-relatively over a range of cases (see Parfit, 1984, pp. 111–114). So they yield derivative agent-relative reasons. There are agent-neutral reasons why, sometimes, we should regard other people’s wrongdoing (agent-relatively) as being less our business, and less worthy of our rational attention, than our own wrongdoing would be. And these reasons are reflected, quite properly, in how we judge people and their actions morally, and in particular whether and when we judge them to have been wrongdoers by their complicity in the wrongdoing of others.

Doesn’t the case against a radically expanded domain of complicity have more to do with freedom than it has to do with the efficient use of rational energy? Doesn’t respect for our freedom require that we be counted as accomplices only in a narrow range of cases? Certainly freedom is important in deciding how much of the morality of complicity we should institutionalise in the law (in the same way that freedom is important in deciding how much of the morality of anything we should institutionalise in the law). But freedom has little bearing on the morality of complicity apart from the law. True, in order to avoid being complicit, one might sometimes have to stop someone committing a wrong by coercing him or manipulating him, both of which would invade his freedom. But that consideration has already been implicitly folded into our discussion under the heading of rational efficiency. Imagine that the wrong about to be committed is one of coercion or manipulation, and it is a wrong I could possibly prevent by coercion or manipulation. Then the simple question we have been addressing is whether, given that there is going to be coercion or manipulation either way, I should prefer not to be the coercer or the manipulator myself. For an agent-relativist the basic answer is yes. For an agent-neutralist the basic answer is no. The main qualification for the agent-neutralist is the one I mentioned already. Am I best-placed to intervene? Sometimes yes, sometimes no. Sometimes it would be better, agent-neutrally, for me to leave the business of wrong-prevention to others, such as the wrongdoer himself, or his mother, or his teacher, or the law. That being so, it might be better for me to think agent-relatively, to downgrade the wrong in respect of the rational attention I give it. So there need be no differences in eventual moral judgment—in determining who counts as an accomplice and who does not—as between those who start from an agent-neutral position and those who start from an agent-relative position. The only essential difference lies in whether the element of agent-relativity in ordinary moral experience (which is what admittedly limits the domain of complicity) calls for, or does not call for, a deeper agent-neutral defence.

Causality and the elimination of complicity

So there are two parts of morality. There is what I should do simpliciter (the morality of principalship), and then there is what I should do by way of contribution to what you do (the morality of complicity). ‘Contribution’ here means ‘causal contribution’. What is at stake in the debate between agent-neutralists and agent-relativists is how to count, in determining what each of us should do, what others do in consequence or as a result.Footnote 4 If what I do has no consequences or results for what anyone else will do, there is nothing for agent-neutralists and agent-relativists to disagree about. My action is not linked to anyone else’s actions in such a way as to raise the question of how anyone else’s actions are to be counted in determining what I am to do.

Spelling this feature out may lead some to think that the discussion has gone down the wrong path. For shouldn’t this feature lead us to doubt whether the examples we have been discussing so far need be analysed in terms of complicity at all? Consider the story in which my captors will have fifty planes flown into fifty skyscrapers if I don’t fly one plane into one skyscraper. For simplicity’s sake, let’s assume agent-neutrality. And for simplicity’s sake, let’s assume that there is a wrong of killing per se, a wrong with no other ingredient but killing. In the example, there is admittedly killing on both sides. It is a case in which my responsibility for the higher number of killings (where the planes are flown by other people) would be as accomplice, whereas my responsibility for the lower number of killings (where I fly the plane myself) would be as principal. But why can’t it be the ordinary responsibility of a principal on both sides? After all, my refusal to kill just a few hundred people will admittedly have the deaths of all those thousands of people among its consequences. In other words, I will have caused all those thousands of deaths. In other words again, I will have killed all those thousands of people, as surely as if I had flown the planes myself. That being so, the choice that my kidnappers give me is simply a choice between my killing a smaller number of people and my killing a larger number. Either way, I am a principal, a killer. It’s true that my killing the larger number is by omission (my not flying and crashing a plane), whereas my wrongfully killing the smaller number is by positive act (my flying and crashing a plane). But this distinction can clearly be accommodated within the scope of principalship. On any credible analysis of killing, there can be killing by omission as well as by positive act. That being so, complicity can be eliminated from the story altogether.

This strategy of elimination works for some imaginable wrongs, but not for others. If there is a wrong per se of acting with fatal consequences then clearly anybody who procures the commission of that wrong also commits the wrong as a principal. With some provisos, the same can be said of the wrong of causing death, if causing death is a wrong per se. But the same can’t be said of the wrong of killing. Killing is not merely causing death, and causing death, in turn, is not merely acting with fatal consequences. In saying this I am not relying on the idea that to be a killer one must have a mens rea of some kind, or be at fault. I am assuming that it is possible to be an accidental and faultless killer. What I mean is that killing is causally different from merely causing death and that causing death, in turn, is causally different from occasioning death. Roughly: killing is causing death other than by making a causal contribution to a killing by someone else; causing death, in turn, is making a causal contribution to a death other than by making a causal contribution to an abnormal action or event (often known as a novus actus interveniens) that itself makes a causal contribution to the death.Footnote 5

You can see at once, even in these rough renditions, what I mean by saying that the differences are causal differences. Causing death is a causally refined way of causally contributing to death. It requires a distinctive causal route from causer to death, free of any novus actus interveniens. Killing in turn is a causally refined way of causing death. It requires a distinctive causal route from killer to death, not mediated through a killing by someone else. This is consistent with the observation that it is possible to kill by omission. If I go away for a month and leave my toddler locked in the house without food, I kill her by failing to feed her. I am clearly a principal in her killing, a killer. The causal route from me to the death is not interrupted by a novus actus interveniens nor is it mediated through killing by another. But things are quite different if I fail to stop someone else from poisoning my toddler. Then my causal contribution is mediated by someone else’s act of killing. It follows that I do not commit the wrong of killing myself, assuming always that there is such a wrong, but am at most complicit in its commission by failing to prevent it. Failing to prevent a child’s being killed is not the same as killing a child by failing to prevent her death. In fact the two are mutually exclusive. Either I killed the child myself or I contributed to someone else’s doing the killing; it can’t be both. (We may say, when I contributed to someone else’s doing the killing, that I was a killer ‘in effect’. I used this idiom myself above. But a killer in effect is not a killer. It is a person whose actions had the same result or consequence as those of a killer.)

This intriguing causal refinement that distinguishes killing from causing death also distinguishes various other nominate action-types from their less causally refined neighbours. To take another example from my own earlier discussion: the causal element of torture is not merely causing pain, let alone acting with painful consequences. It is inflicting pain. Inflicting pain is causing pain other than by making a causal contribution to an infliction of pain by someone else; causing pain, in turn, is making a causal contribution to pain other than by making a causal contribution to an abnormal action or event that itself makes a causal contribution to the pain. So when I fail to prevent or procure the torture of fifty people, I am not myself their torturer. Following Sandy Kadish, I will call actions like these ‘nonproxyable’ (Kadish, 1985, p. 373), Kadish uses the label to refer to actions (such as having sex and eating) that do not consist in making a causal contribution to anything. Since they do not consist in making a causal contribution to anything, one obviously cannot engage in them oneself by contributing causally to another person’s engaging in them. But what Kadish does not notice is that there are many actions that do consist in making a causal contribution to something, but which also have the feature of being nonproxyable, because the requisite causal contribution is a refined one. This is true of not only of killing and torturing, but also of coercing, enslaving, inducing, destroying, igniting, and countless other action-types. Just leaf through the dictionary for plentiful examples. Where nonproxyable actions are wrongs per se or necessary ingredients of more complex wrongs, one cannot be a principal in these wrongs by making one’s causal contribution through another principal. In such cases, whoever acts through a principal must be an accomplice.Footnote 6 So in such cases the attempt to eliminate complicity from the moral landscape, in favour of a more capacious domain of principalship, fails.

Why, we may wonder, are so many common-or-garden nominate action-types nonproxyable? Part of the answer surely lies in the fact that such action-types are often wrongs per se or at least necessary ingredients of more complex wrongs. Where the wrongs in question are common, it saves a lot of time if such action-types have names of their own. But that answer just invites a restatement of the question. Why are there so many wrongs that have actions of nonproxyable types as necessary ingredients? Why, in other words, are our moral classifications so often organised around keeping principals segregated from accomplices? We already know the answer. The distinction between principals and accomplices, as we discovered, is embedded in the structure of rational agency. As rational beings we cannot live without it. It enters our thought as soon we begin to think about responsibility. I am responsible for my actions, and you are responsible for yours. My actions are mine to justify or excuse, and your actions are yours to justify or excuse. And yet my actions include my actions of contributing to your actions. So there is a sense in which my responsibility for my actions can extend out to your actions. I can be accused of failing to attend to reasons that are yours to conform to even though I cannot be accused of failing to conform to them myself. I fail to attend to them, in the relevant sense, by contributing to your non-conformity with them. The question which divides agent-relativists and agent-neutralists is merely how much of this kind of rational attention I should be giving to your reasons. Which contributions to your non-conformity make me complicit in it? The two sides differ in their response. But they can do so without disagreeing about what counts as principalship.

This much is repetition. It repeats the explanation of why the contrast between principals and accomplices is embedded in the structure of rational agency. The pay-off that we have just discovered is that the distinction between principals and accomplices is therefore often marked in morality. Many moral wrongs, such as torture and rape and betrayal and deceit, are committed only by performing nonproxyable actions, such that anyone who contributes to their commission through another person’s commission of them is an accomplice, not a principal, in their commission. If moral theory has lost sight of this distinction in recent times, we have the consequentialist revolutionaries to thank. Strict consequentialists insist that it matters only what consequences my action has, not how those consequences come about (where this has no further consequences). We have just found one reason why they cannot be right about this. They cannot be right because if they were right, an elementary truth about responsibility would be falsified. The elementary truth is that I am responsible for my actions, while you are responsible for yours. Each of us has a different relationship to our own actions from the relationship that we have to the actions of others. The relationship we have to our own actions is direct: we answer for them as such. The relationship we have to the actions of others is indirect: we answer for them only inasmuch as, by our own actions, we contribute to them. If strict consequentialists want to abandon the distinction between these two modes of responsibility for actions they need to abandon, not merely agent-relativism, but also the idea that we each have a special relationship to our own actions. In the process strict consequentialists take their own moral position to the edge of intelligibility. Who is strict consequentialism addressed to, if not to each of us in respect of our own actions? How does it survive as a moral view without presupposing the elementary truth that we are each responsible for our own actions?

Many people, bending over backwards to avoid this strict consequentialist error, are tempted to say that people, as responsible agents, have free will and defy causality. Causal chains therefore cannot run to or through their actions. One wipes people out of the story, as responsible agents, if one insists on subjecting their actions to causal explanations or on running causal explanations through their actions. So it is often said. But this strikes me as an equal and opposite mistake, a way of fending off strict consequentialism that throws the baby out with the bathwater. For the strict consequentialist clearly has one thing right. When I pay a hitman to kill an enemy it is a straightforward consequence of what I do that the hitman kills my enemy, and hence that my enemy dies. Barring the special case of novus actus interveniens, I procure the killing and I also cause, through the hitman, the death. There is a straightforward causal explanation of how it all unfolds, action and reaction, antecedent and consequent. None of this wipes out the hitman’s part in the story as a responsible agent. On the contrary, it presupposes it. The hitman’s part in the story as a responsible agent is wiped out only if we go on to claim that I was myself a killer (not just a killer in effect, but a killer). The reason why this wipes out the hitman’s part in the story as a responsible agent is not that the hitman has free will and can change his mind after I give him the go-ahead. His change of mind after I give him the go-ahead is no different, causally speaking, from a change in the wind after I throw a ball. No: the reason why claiming that I was the killer wipes out his part in the story as a responsible agent is that he was the killer—and there can be only one. He committed the primary and nonproxyable wrong. My part in the causal story was not as principal but as accomplice. In fact I straightforwardly caused him to commit the wrong. To say that is to confirm, not to deny, his responsible agency.

Complicity beyond causality?

Procuring or causing or inducing someone to commit a wrong; failing to prevent or permitting or enabling the commission of a wrong: these are straightforwardly causal modes of complicity. Where these modes of complicity are concerned, the principal would not have committed her wrong but for the accomplice’s intervention. The accomplice clearly made a difference to the overall incidence of wrongdoing; the world clearly contains extra wrongs thanks to her. But there are other modes of complicity (for example, assisting and encouraging the commission of a wrong) which are prone to fail this test. I lend someone the tools that she uses to break into a house; but she didn’t exactly need them; she would have got in by other means even without my tools. I egg the kidnapper on with reminders of the great rewards his crime may bring; but he didn’t exactly need my encouragement; he was going ahead with the kidnap plan anyway. English law says, and I think rightly, that I am complicit in these cases only because my assistance actually assists and my encouragement actually encourages.Footnote 7 A failed attempt at assistance or encouragement is just that. It may still be wrongful but it is not complicity, any more than a failed attempt at murder is murder. It is a failed attempt because it has no effect, no impact, on the principal. This helps us to see that there must be something causal going on in these cases. I am making some kind of causal contribution to the principal’s wrong that serves to turn my attempt into a success. But the problem is that I don’t seem to have made a difference to the overall incidence of wrongdoing; the world contains no extra wrongs thanks to me. This may make us think again of Solzhenitsyn’s way of dividing up morality. The wrongdoing is coming into the world anyway; perhaps one can’t change that; and yet there remains a sense (doesn’t there?) in which the wrongdoing comes into the world through the accomplice—with his assistance or encouragement—and that is what implicates him in it as an accomplice.

Granted that I made some kind of causal contribution to the principal’s wrong, why is it a problem for the causal view of complicity (as we may call the view defended in this paper) if I didn’t thereby make a difference to the overall incidence of wrongdoing? It is a problem because it seems odd that someone should be expected to pay attention (in her practical reasoning) to features of the world that will come out no better whatever she does. Whether one is an agent-neutralist or an agent-relativist about the avoidance of wrongdoing, shouldn’t one advocate (under the heading of the efficient use of rational energy) that people aim to make differences to wrongdoing they can make rather than differences they can’t make? Of course one may reply that assistance and encouragement often do make all the difference between commission and non-commission of the principal wrong. Often they tip the balance. The risk that one will tip the balance in any particular case arguably justifies a general rule against providing assistance or encouragement to potential wrongdoers, a general rule that applies even to cases where the help or encouragement proves unnecessary, because the wrong will be committed anyway. No doubt this consideration is valid and important. It figures not only in justifying the law of complicity but also in explaining the morality of complicity apart from the law. But there is also a deeper consideration in play that has nothing to do with the risk of tipping the balance. For there is a sense in which, even when the assistance or encouragement furnished to a wrongdoer is unnecessary, it does (in spite of appearances) make a difference to the overall incidence of wrongdoing, a difference that warrants attention in the accomplice’s practical reasoning.

As a first step towards seeing this deeper consideration, think about the infamous ‘arms dealer’ defence: ‘If I don’t do it, somebody else will’Footnote 8 This defence strikes most of us as lame. But why? Surely the arms dealer has a point. Surely it makes no difference to the overall incidence of arms dealing whether this particular dealer deals in arms or not. There are plenty of others waiting in the wings. Maybe they will even be able to deliver the same arms on the same day at the same price. If this arms dealer were shot today there would scarcely be a blip in the statistics for arms dealing this year. So the arms dealer doesn’t really change anything. Or does he? We are invited to think of the problem impersonally, perhaps calculating the net number of arms dealt, or the net number of arms deals done. Possibly this arms dealer makes no difference to those impersonal figures. But it is important to remind ourselves that the arms dealer himself has no business thinking of the problem so impersonally. After all, some of the arms deals in question are his arms deals and this makes him responsible for them as a principal. They are his to justify or excuse. Other deals that might take place if he did not deal would not be his. At most he would be responsible for them as an accomplice. This certainly opens up the question that arose, mutatis mutandis, in the case of the sadistic kidnappers: whether and when the avoidance of (complicity in) arms dealing by others is capable of justifying one in dealing arms oneself. In answering this question, it clearly matters how much arms dealing is involved on each side, and how bad the arms dealing is. But notice that it also matters why one engages in the arms dealing. One justifies one’s own arms dealing by pointing to the arms dealing of others thereby avoided only if one deals in arms in order to avoid arms dealing by others. So one needs to pay attention (in one’s practical reasoning) to the facts upon which one relies for justification.Footnote 9 That is why the ‘arms dealer’ defence seems so lame. The arms dealer who makes it clearly does not deal in arms to avoid arms dealing by others.

I said before: It is not obvious that someone should pay attention (in her practical reasoning) to features of the world that will come out no better whatever she does. But it is obvious that she must pay such attention if she relies on those features by way of justification for what she does. Justifying one’s wrongs is not cancelling them; it is showing that one committed them for a sufficient reason, that one was not unreasonable in committing them. That being so, we should not allow the arms dealer to get away with the idea that, by dealing arms, he made no difference to the overall incidence of wrongdoing. In the relevant sense of ‘overall difference’, he did make an overall difference. He added his own arms dealings. True, he also subtracted like arms dealing by others, by competitors of his who would have filled the space in the arms market if he had moved into another line of work. But he cannot be allowed simply to treat the subtraction as cancelling out the addition, as yielding a zero sum. That is an abdication of responsibility. It is an abdication of responsibility because it is a refusal to accept that the relationship he has to his own wrongs is different from the relationship that he has to his competitors’ wrongs: in particular, that he can rely upon theirs to justify his only if he commits his in order to avoid theirs.

How does this help us understand cases in which the accomplice’s assistance or encouragement is, so to speak, surplus to the principal’s requirements? There are some dissimilarities. But all the cases teach the same lesson. In the assistance and encouragement cases, as in the case of the arms dealer, we should resist the idea that the accomplice made no difference to the overall incidence of wrongdoing, in the relevant sense of ‘overall incidence’. She added the wrongs (of the principal) to which she (as an accomplice) made a contribution. True, she also subtracted wrongs to which she did not make a contribution, wrongs which, ex hypothesi, would have been committed without her assistance or encouragement. But the wrongs that she subtracted cannot be regarded as literally cancelling out the wrongs that she added. By virtue of her contribution to them, she stands in a different relationship to the wrongs that were added. It is an abdication of responsibility for her to rely on the wrongs thereby subtracted unless she is pointing to them by way of justification or excuse for her own contribution. In which case she has to explain how the avoidance of those subtracted wrongs figured in her practical reasoning. So she cannot help herself to the proposal that she need not attend (in her practical reasoning) to features of the world that will come out no better whatever she does. She needs to have attended, in her practical reasoning, to whatever features of the world she relies upon to justify her own acts. If they did not figure in her practical reasoning, they also do not help her to defend her contribution to the principal’s wrongs.

You may say that this line of thought is question-begging. It assumes that our assister or encourager is an accomplice in order to explain why, on the causal view, she is an accomplice. But that is not quite true. The argument certainly assumes that the assister or encourager contributes to the wrongs of the principal. But that assumption was built in at the start. We followed English law in inisisting that it is a necessary condition of being an assister or encourager that (as the case may be) one’s assistance actually assists or one’s encouragement actually encourages. The fact that one made such a contribution already gives one something to justify. Does it already make one complicit? The issue is tricky. Of course, there are tricky issues of intention and knowledge to settle. But that is not the area of trickiness I have in mind. The trickiness I have in mind is this: Should we consider someone’s justification for performing an act before or after deciding that the someone in question is an accomplice? Is complicity the thing to be justified or the unjustified thing? In surveying the debate between agent-neutralists and agent-relativists in Sect. ‘Principals and accomplices’ above, I sided with the view that at least some questions of justification need to be settled first, before classifying an agent as an accomplice. That is why (I claimed) agent-neutralism throws, in principle, a wider net of complicity than agent-relativism: it allows each of us fewer justified abstentions from preventing the wrongs of others. If this view is right, then the argument of this section was not question-begging. It certainly assumed that an assister or encourager contributes to the principal’s wrong. But the argument did not assume that an assister or encourager was thereby made complicit in the principal’s wrong. Rather, before classifying her as complicit, it allowed her the opportunity to justify her assistance or encouragement of the principal’s wrongs by reference to the alternative wrongs that were thereby avoided: in other words, by reference to the difference that she made to the overall incidence of wrongdoing. So we are still in part one of Solzhenitsyn’s two-part morality. We are still talking about those who make a difference to the state of the world.

After Solzhenitsyn

Solzhenitsyn’s remark makes us think of the defensive arms dealer in the following way. The arms dealer doesn’t really make a difference to the incidence of wrongdoing. The arms dealing that he is about to embark upon will ‘come into the world’ whatever he does. The only question for him is whether to be implicated in this particular arms trading, whether it is to ‘come through him’. How is he implicated in this particular arms deal given that he makes no difference to it? The puzzle vanishes if we think of the principal wrongdoer as something much bigger than him, say the arms industry as a whole. It is the arms industry that is necessary for any particular arms to be dealt. One arms trader is just a tiny cog in the huge machine of wrongdoing. It is being a cog that makes him complicit. The question then becomes: does he need to make any difference at all, anywhere in the industry, to be implicated? Does he have to make at least a tiny marginal difference? Or is it enough that he is just part of it, dealing arms, joining in the industry’s ventures, albeit making not even a marginal difference to the extent of arms trading? Can one perhaps even be a wrongdoer simply by association?Footnote 10

I have already suggested how to approach these questions. No doubt there are plenty of cases where the principal is a collectivity and the individuals who make up the collectivity are accomplices. Perhaps they throw up special difficulties connected with the marginality of individual contributions.Footnote 11 But such cases proliferate out of control if one begins as Solzhenitsyn begins. In Solzhenitsyn’s vein one thinks of the problem in two parts, beginning with the moral shape of the world—the wrongs that will be there anyway, with or without my personal influence. One shifts afterwards to thinking about my moral position, and that of other particular people. Given that the wrongs will come into the world anyway, how can I avoid getting my hands dirty on them? In place of this way of dividing up morality I have been trying to lay the groundwork for a different picture in which the moral shape of the world is simply the moral position of the people in it. Wrongdoing never just ‘comes into the world’ and accomplices never just get their hands dirty on it on the way past. Accomplices themselves bring wrongdoing into the world. So too do principals. The essential difference between accomplices and principals is that accomplices bring wrongdoing into the world through principals. And that is where morality cleaves in two. An accomplice is one who acts with the consequence or result that the principal commits the wrong. I see no reason to believe that there is any other way of being complicit in another’s wrongdoing than by making such a difference.