Causeless complicity

Abstract

I argue, contrary to standard claims, that accomplice liability need not be a causal relation. One can be an accomplice to another’s crime without causally contributing to the criminal act of the principal. This is because the acts of aid and encouragement that constitute the basis for accomplice liability typically occur in contexts of under- and over-determination, where causal analysis is confounded. While causation is relevant to justifying accomplice liability in general, only potential causation is necessary in particular cases. I develop this argument through the example of the role of U.S. legal officials in abetting the acts of unlawful interrogation that have taken place since 2001. I also suggest that there may be a limited justification for ex post ratificatory accomplice liability.

1. The category of “complicity,” in both criminal law and in ordinary ethical thought, picks out ways one person can be liable to sanction for bad things (criminal acts among them) done through the agency of another. We become complicit in the wrongs of others, thus becoming—in a variant formulation—the accomplices of those principal agents. In the central cases of complicity, whereby we become accomplices by assisting others in their wrongdoing or encourage them to engage in the wrong, the justification for liability is uncontroversial: through our acts, we participate in their wrongs, and so become liable for them as well. But there still is something puzzling about this liability. If, as John Gardner argues, individual wrongdoing involves individuals making a (wrongful) difference to the world, then accomplices present a problem. For, frequently, although accomplices do actively associate themselves with others’ wrongs, the wrongs would have occurred regardless—with or without the encouragement or assistance. Thus, a full justification of accomplice liability would seem obliged to make one of two choices: insist that, appearances notwithstanding, accomplices do make differences to the wrongs their principals do; or deny that making a difference to a wrong is a necessary condition of responsibility for that wrong. Gardner opts for the first choice. In this paper, I defend the second. We can be complicit in others’ wrongs without making a difference to the occurrence of those wrongs.

Now, making a difference to an event’s occurrence is a central form of causation. Thus, the question in play here is whether complicity is a causal relation, or something else. Doctrinally, the question is whether proof of causation (at least sine qua non, perhaps proximate) is a requisite for liability. Since I do not think accomplices always make differences to their principals’ wrongs, I argue that the relationship is non-causal. And although some of the textbook writers, especially in the United Kingdom, will argue that courts require, or presuppose, some form of causal connection, I also argue that this is a misinterpretation of doctrine. Causation may be present, I argue, and causal relations feature in the justification of accomplice liability overall, but causation is not necessary to complicity. Put yet more generally, causation does play a role in determining patterns of the overall distribution and justification of accomplice liability, but not in particular instantiations. Once this point is seen, some new territory opens up as well. In particular, my argument reveals a normative space that can justify liability in cases of attempted complicity (for instance, when the principal fails to complete the crime), as well as limited scope for after-the-fact liability¹.

2. The claims I have just made are broad and apply, as I will show, to a great range of cases. But there is an underlying agenda to my argument as well, and I will disclose it by relating some now-familiar history.

In the spring of 2002, the first reports of extensive abuse of detainees emerged from the Abu Ghraib prison complex in Iraq, with the publication of photographs showing low-level military personnel engaged in depraved and sadistic acts towards these prisoners. Initial government claims that these photos represented anomalous behavior by a few “bad apples” were immediately belied by the leak of documents less graphic but no less shocking. These were legal memoranda, principally drafted by lawyers working for the Department of Justice and the Department of Defense. These memoranda, in turn, set out a legal theory under which large categories of detainees captured in the “Global War on Terror,” could be effectively stripped of their protections against torture and other forms of abuse by American military and civilian personnel.

The leaked memoranda gave a clue to what has been since confirmed and reconfirmed, that U.S. officials had assembled a comprehensive program of counter-terrorist intelligence that relied on techniques forbidden under bedrock international law, including the Geneva Conventions and the Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment, and U.S. criminal laws implementing these anti-torture and abuse agreements. After the attacks of September 11th, the order went out from the White House to “take the gloves off” in counter-terrorist policy, to “work through, sort of, the dark side,” in the words of Vice President Dick Cheney (Mayer 2005).

These techniques, according to news reports, include stripping prisoners naked and subjecting them to extreme heat, cold, and noise; “Palestinian” hanging, whereby prisoners are shackled at the wrists behind their backs and suspended by their shackles, thus dislocating their arms; and “waterboarding,” or using wet towels or running water to give prisoners the sensation of drowning (Danner 2004; Jehl and Johnston 2004). Various means of inducing fear and humiliation have also reportedly been used, such as the use of dogs, sexual degradation, threats to family members, and desecration of holy objects. The subjects of these techniques were suspected high level al Qaeda operatives, held and interrogated by the CIA; suspected al Qaeda and Taliban fighters from Afghanistan, held by the military; and insurgents from Iraq, also held by the military (though sometimes also interrogated by the CIA).

The gloves that needed to be taken off were not just the moral inhibitions of military and CIA interrogators. There are also U.S. criminal laws implementing the international agreements banning torture and prisoner abuse. It was the lawyers’ job to perform this task.² The lawyers did so with enthusiasm, drafting literally reams of memoranda whose range of arguments concerning the inapplicability of quite obviously applicable laws was notable for creativity, if not legal quality.³ Indeed as others have noted, the memoranda display an ideological conformity and certainty that is highly peculiar given the contentious nature of the topic—a topic so contentious that the White House, Department of Justice, and Defense views were soundly rejected by lawyers within the military and at the State Department—and were later repudiated by the Justice Department itself.⁴

The poor quality of the legal reasoning in the memoranda makes their purpose somewhat puzzling, especially since they were drafted by a cadre of the top administration lawyers, those at the Office of Legal Counsel (OLC) in the Justice Department. Conventionally, OLC lawyers respond to requests from within the executive branch for interpretations of the scope and implications of law, and when different executive agencies disagree, the OLC issues its own official statement of the law, which then binds all executive agencies. Interrogation practices presented such a case of inter-agency conflict, since both military and State Department lawyers vehemently disagreed with the analysis initially broached by the OLC lawyers, and worried about the effects of U.S. abrogation of anti-torture standards. But, at the end of the day, the OLC’s opinion would lay the groundwork for policy, silencing the inter-agency dispute. This much is clear. But the memoranda appear also to have served a less conventional purpose: although the OLC memoranda could not bind courts in their independent interpretation of U.S. law, their very existence could provide political cover for the administration—“our lawyers said our policy was o.k.”—as well as the basis of a criminal defense of reasonable reliance on official orders for officials who in fact engaged in abusive practices. And thus the lawyers played a very significant role in the torture policy.

Now, I have been painting with a broad brush. U.S. civilian and military officials appear to have abused large numbers of suspected terrorists and insurgents in a broad variety of contexts, with much but not all of the abuse a matter of official policy. Because my interest lies in the complicity in these abuses of one central cohort of actors, namely the government lawyers, it is worth distinguishing the categories. Captives from the battlefields of Afghanistan, as well as terrorism suspects handed over to cooperating intelligence agencies overseas, were the intial primary target of these memoranda, and their interrogations took place at official, as well as highly secret, military bases around the world, most famously including the U.S. Naval Base at Guantanamo Bay, Cuba, but also including “black sites” in (perhaps) Eastern Europe and Asia. The abuses at Abu Ghraib appear to have been counter to the mainstream of policy, insofar as the official U.S. position was that prisoners from the Iraq theatre were entitled to the protections of the Geneva conventions, which should have barred all these forms of abuse. (On the other hand, the abuses took place after Major General Geoffrey Miller, commander at Guantanamo Bay, was sent to Abu Ghraib to “Gitmoize” the intelligence collection—that is, to increase the aggressiveness of interrogation practices (Hersh 2004).)

So there is already a distinction between those cases of abuse occurring as the intended result of policies cleared by the lawyers, and the unintended, if foreseeable, effects of those policies on military personnel reasonably confused by the blurring of the formerly bright lines of proper and improper detainee treatment. There is also a temporal distinction that complicates the issue. While some abusive interrogations occurred after the writing of the memoranda, for instance the interrogation of Abu Aubaydah, said to be an al Qaeda leader, others, including perhaps the interrogation of Khalid Sheikh Mohammed, may have taken place before their writing. That is, the memoranda may have been requested because CIA officials were worried about their liability for acts already performed.

Across all these categories, common questions of moral and criminal responsibility arise. For those who actually made the policy decisions to have these acts done, the questions of complicity are easily answered in causal terms. In my view, those officials are morally culpable for the atrocities, and—were a forum and willing prosecutor available—ought to be criminally liable for them as well, under both international and U.S. domestic law. But the liability of the lawyers is complicated in two dimensions. For most cases, they did not formulate policy, but merely eased its implementation. Given the aggressive posture of U.S. policy after 9/11, it is likely that many of these abuses would have taken place even if the lawyers had refused to sign off. There is, at least, little basis for thinking the lawyers causally responsible for the abuses. Second is the temporal issue: for some of these cases, the lawyers’ actions may have served as ratification rather than stimulus. As a moral matter, the lawyers who tried to provide cover for these policies are tainted by the degradation and torture that occurred. For many legal professionals, the disclosure of the memoranda was particularly disheartening, for while law has not always stood on the side of justice, the legal prohibition of torture stands as one of the signal achievements in modern humanitarian law, as well as a proud legacy of the Enlightenment. That legacy, at least in the U.S., has now been squandered. But as a matter of legal doctrine, matters are less clear. In fact the scandal of detainee abuse squarely presents the question of the scope of accomplice liability I began with: whether accomplices must cause (or causally contribute to) the acts of their principals, or whether instead complicity can outrun causation.

3. Arguing this will take some work, in particular refining some points about the proper treatment of morally significant causation in complicity contexts. The general lessons I hope to extract from pondering the torture case bear principally on issues of responsibility in institutional settings, but also, I hope, illuminate more generally the justification of complicity law.

I first describe briefly the general shape of accomplice liability at common law, then probe the underlying concept of causation. Anglo-American criminal law distinguishes two central ways of becoming liable for a crime. First, one may commit the crime oneself, by performing the prohibited act (the actus reus requirement) with the requisite culpable state of mind (the mens rea requirement)—for example, breaking into another’s house, with the specific mental state of intending thereby to remove the property of another. Such a person is a principal. Alternatively, one may play a role in another’s commission of a crime, by intentionally aiding or encouraging another, himself with criminal intent, to perform the prohibited act. Such a person is an accomplice (also called a secondary party). The basic requirements are deeply rooted in common law, but have been formalized in the U.K. in the “Accessories and Abettors Act of 1861,” (amended by the Criminal Law Act of 1977):

Whosoever shall aid, abet, counsel, or procure the commission of any indictable offense ... shall be liable to be tried, indicted and punished as a principal offender.

And in the U.S., in the Model Penal Code §2.06:

(3) A person is an accomplice of another person in the commission of an offense if:

• (a) with the purpose of promoting or facilitating the commission of the offense, he

  • (i) solicits such other person to commit it, or

  • (ii) aids or agrees or attempts to aid such other person in planning or committing it, or

  • (iii) having a legal duty to prevent the commission of the offense, fails to make proper effort so to do: or

The fundamental point is that accomplice liability is a basis of liability for another’s substantive crime, not a crime in its own right. Once a defendant’s liability as an accomplice is established, he is thereby treated as a principal—that is, as if he had committed the actus reus of the crime itself. There has been, and continues to be, substantial debate in the literature over the exact specifications of the doctrine, principally the mental state requirement. While it is generally acknowledged that accomplices need not have the “same” intent as their principals (hardly possible in any case, since accomplices are intentionally performing acts of assistance or encouragement, not the prohibited acts themselves), the alternative mental state required of accomplices is in dispute. In the United Kingdom and Commonwealth, a secondary party’s knowledge that his acts will tend to aid or encourage another’s crime is sufficient for liability, while in the United States the dominant rules requires that the accomplice take a purposive stance towards the principal’s acts, thereby intending to aid in those acts.⁵ But our focus here is on the actus reus, on what the accomplice must do in order to render himself liable for another’s act. And here common law and code, U.K. and U.S., agree that there are two basic categories for complicity-generating behavior: rendering assistance to another planning to or already engaged in committing a crime; or encouraging that other to commit the crime, without necessarily assisting its commission.

Complicity doctrine is therefore marked by a dramatic asymmetry in the positions of accomplice and principal. The principal’s actions are only a basis for liability if they satisfy the relatively constraining templates of substantive criminal law: they are burnings or batterings or killings or rapings or robbings, or attempts thereof. By contrast, virtually any kind of act, speech or otherwise, can satisfy the act requirement of accomplice liability, for virtually anything one person does can be a form of assistance or encouragement to the other. Thus, the potential scope of liability for accomplices is both exceptionally wide and vague, and one of the primary roles of the mens rea requirement is to limit this scope and render it more predictable.⁶

This also suggests a role for a causal requirement in complicity cases: namely, as another limit on the scope of accomplice liability. If a requirement of causal contribution could be made out, it might also contribute to a deeper retributive basis for accomplice liability—linking accomplices by effect, as well as intent, to actual harms. In paradigm cases, causation can readily be found. At its core, and before any fancy philosophical analysis, the notion of causation involves making a difference to what happens in the world. By my acts, I can often make a difference to what others do. If I contract a hit man, who otherwise would have no interest in the target, I can be fairly said to have caused the killing (through my solicitation) committed by the hit man. If a criminally-minded acquaintance is debating whether to burgle a house, I tell him I know of a fence who can help dispose of the loot, and this consideration is dispositive in his deliberations, then I causally contribute to his act (even if, as a pragmatic matter, we might withhold the claim that I “caused” him to commit a crime he was already disposed towards). And if I provide him with the burglary tools, or give him a ride away from the crime scene, then I causally contribute to the robbery. Perhaps my assistance was not a necessary (sine qua non) condition of the principal’s committing a similar crime. But it was a necessary condition of his committing that crime: the crime using my burglary tools, or arrived at in my car.

So the paradigm cases are clear, and perhaps because the paradigm cases have been fixed so firmly in the English judicial mind it is frequently asserted that English law requires accomplices to contribute causally to the resulting crime.⁷ John Gardner happily endorses this requirement as well.⁸ But these paradigm cases, stipulatively specified as they are, mark the limit of causality’s role. For in many more cases, causal relationships between accomplices and principals are too indeterminate (or undetermined) epistemologically, physically, and metaphysically to satisfy any real doctrinal requirement. Second, a too-insistent focus on a causal requirement, even if it could be proved, would distort complicity doctrine by rendering it oblique to the expressive and preventative dimensions of principal liability. Principals can be convicted for regulatory violations that cause no harm, and for all manners of attempt, including attempts that cause no harm. U.S. law, both common and reconstructed (M.P.C) has largely recognized this point. At common law, an accomplice need only act in a way that might have made a difference to the outcome. The Model Penal Code explicitly fails to require causation, as do the jurisdictions that follow the M.P.C. on this point.⁹ The M.P.C. further recognizes fully inchoate accomplice liability, when an accomplices aids, encourages, or attempts to aid or encourage a principal whose crime was never committed.¹⁰ But causeless complicity plays an implicit role, I suggest, in non-M.P.C. British and U.S. criminal law as well. Put another way, the M.P.C. recognizes (though perhaps exaggerates) the already present logic of causeless complicity.

Consider first assistance. Complicity doctrine imposes liability not just for the paradigm cases I mentioned above, but also for cases where the accomplice’s assistance is in reserve, or otherwise superfluous to the principal’s crime. Denoting the accomplice as “S” and the principal as “P”, S might give P a second set of burglary tools, which he will use incase his own break, or stand ready to assist in case P cannot handle the robbery on his own, or provide money that P may need for his criminal scheme if expenses are greater than his own funds. The point is well-illustrated by a New Zealand case, Larkin v. Police.¹¹ Larkin was convicted after trial as an accomplice in the breaking and entering of a liquor store, where he acted as a look-out on the street. What makes the case interesting is that he was an unannounced volunteer for the role. Larkin had overheard some men at a party talking about robbing a liquor store when, by his account, he decided to stand watch outside the store, ready to shout a warning if the police arrived. Larkin ran when he heard someone else shout “Cops!” and was arrested shortly thereafter. He argued two central points in his defense: first, that because he did not communicate to the principals his intent to keep watch, nor did he actually warn them, he did not render either assistance or encouragement; and second, that the robbery had been completed by the time he arrived on scene, and thus his help was too late to count as assistance in any event.

Let us put aside Larkin’s second claim that he arrived too late to be an accomplice. As a semantic point, and as the court ruled, the defense is valid, for one cannot assist in the production of an act that is already complete—one can only assist in some further act, such as transporting the loot or hiding the principals. As the appellate court found, Larkin could be convicted as an accessory after the fact, but not as an accomplice to the burglary. And, indeed, the appellate court instated a conviction for him as an accessory after the fact.¹² (As a normative matter, as I will argue, there might be room for such an expanded conception of complicity.) More interesting is Larkin’s other defense, that an uncommunicated act of act of assistance is insufficient. Here Larkin was resting on the usual case, in which acts of assistance are communicated and so may often be deemed to have a causally significant role in encouraging the principals’ crime. But accomplice liability can be grounded in assistance or encouragement, and not necessarily both. Uncommunicated assistance is still assistance, and so, if Larkin’s act of standing watch had taken place before the crime was completed, it would be grounds for conviction as an accomplice. In this complicity differs from conspiracy, where the gist of the offense is an agreement between the parties, rather than an act by one aimed at producing some effect in another.

There are actually two key points here: assistance can be a basis of complicity without communication; and assistance can suffice for responsibility even if the assistance does not play a causal role in the crime. Assume Larkin stood watch during the burglary. His contribution to the burglary might have played a causal role in its commission, to be sure: had he shouted a timely warning, the burglary might have been completed successfully, or at a slightly different time, and so forth. Relying on an unanalyzed notion of causation, as making a difference to some event’s occurrence, Larkin’s act played no causal role—or at least not in relation to the burglary. To be sure, Larkin caused many things by standing in front of the storefront: he moved some air molecules, perhaps made a difference in the perceptual fields of passersby, wore down his shoe soles a bit. But the principals’ act was unaffected, no matter how finely it is described.

Nor will more philosophically-refined theories of causation help here. The two dominant theories of causation today both descend from David Hume’s famous pronouncement, “We may define a cause to be an object followed by another, and where all the objects, similar to the first, are followed by objects similar to the second. Or, in other words, where, if the first object had not been, the second never had existed.”¹³ As many have noted, the two definitions are not at all equivalents, and instead stand for very different concepts of causation. The first provides what is now known as a “regularity” account, and as refined by J.L. Mackie, H.L.A. Hart and Tony Honoré, and Richard Wright, such an account holds that one event (or factor) is a cause of another event when it is an insufficient but necessary element of a set of conditions actually sufficient but not necessary for the occurrence of the second event. (See Hart and Honoré (1985); Mackie (1974); Wright (2001)). On such accounts, particular (singular) causal claims are seen as instantiations of generalizable, law-governed processes: the thrown ball can be said to have caused the window to break because a thrown mass, coupled with such other conditions as the brittleness of the glass, instantiates a regularity involving many glass fragments. Similarly, a poisoner who slips poison into someone’s drink performs an act that, given the vulnerabilities of the victim, is also an element of a set of conditions instantiating regularities involving toxins and organ failure.

The latter half of Hume’s definition, which makes causation turn on the question what might have happened had the potential cause not been present, is the basis for the second path of causal theorizing, in terms of counterfactuals. On counterfactual accounts, causation is seen as, at root, a matter of dependence (or in David Lewis’ refinement, a matter of influence.) An event c is said to be (a/the) cause of another event e if the two can be related to each other, either directly or by a chain of mediating events, such that if c had not occurred, e would not have occurred (or the events on which e depended would not have occurred, etc.).¹⁴ Counterfactual views furthermore tend to be highly permissive, allowing absences and omissions to count as events, and permitting a reasonably high degree of refinements among the dependent effects. If Zidane had not headed the ball at 28:30, France would not have scored at 28:32; thus Zidane caused the goal at 28:32 (even if someone else might have scored at 28:34). If the FBI and the CIA had communicated effectively, the 9/11 hijackers would never had boarded their planes; their failure to share information causally contributed to the disaster (even though it is also true that the failure of the hijackers’ friends and families to dissuade them from their path also contributed to the outcome). Clearly, it can be difficult to assess the truth of such counterfactuals, though we do so all the time. The counterfactual theory does not insist that we know their truth before we assess causal claims, but instead offers it as a basic interpretation of what it is to make a difference.

Because of commonly noticed difficulties in regularity accounts—such accounts have trouble distinguishing effects from correlates, and in making clear the direction of dependence, from cause to effect—counterfactual accounts, which solve these problems for the most part, are probably more prominent today. But both regularity and counterfactual accounts have trouble with issues that arise frequently in complicity contexts: pre-emption, overdetermination, and underdetermination. Larkin can be seen as a pre-emption case. Assume, again, that Larkin was present at the time of the burglary, but that he did not alert the burglars to the arrival of the police because one of their confederates shouted a split second beforehand. Both counterfactual and regularity accounts will treat Larkin’s aid as a preempted, potential cause, rather than an actual cause: the burglary neither depended on Larkin’s act, nor was it in the circumstances a necessary element of a set actually sufficient for the burglary’s success, since the burglary was already a success.¹⁵ As a matter of causal analysis, this seems correct: Larkin’s act might have been a cause (and the accounts can explain why), but it was not. The analysis in terms of causation thus fails to indicate the basis of responsibility.

Someone insisting upon causation’s relevance might still make a stand here, in two different ways. The first is by re-describing the crime as it actually occurred, to include all the surrounding circumstances. A burglary that occurs with Larkin standing in the street differs from a burglary occurring with Larkin sleeping in a car. In Gardner’s nice phrasing, S has successfully subtracted from the world a crime without his aid, and added one with it.¹⁶ But a thicker description will not support causation. It is analytically true that whatever S does makes a difference to the world, by making actual the possible world in which S acts. But it is not an analytical truth that this difference is a causal difference. Making a causal difference means changing the properties of P’s criminal act, and that means showing more. Compare: S might give P a red shirt to wear, instead of the green shirt he has on, and so subtract the world in which P commits the murder wearing a green shirt, substituting the world in which he wears the red. Such a difference has no causal bearing on the killing itself, just on the description of the actor. Or even, suppose S simply says “hello” to P, as P goes off to the crime. S now replaces a world in which the crime is committed without his greeting, for one committed with it. This is again a difference, resting on an act (albeit a speech act) by S, but it bears no causal relation to P’s act. Voicing a greeting is, categorically, not an adequate actus reus for accomplice liability, even though it does make a difference to the world. While making differences matters to complicity, not all differences matter.

The second defense of a causal role is to insist on the significance of the counterfactuals that S makes true in such cases: by making it likelier that P would succeed in the crime, S causally contributed to the crime. This is the route taken in the famous U.S. case of Attorney General v. Tally, in which defendant Tally, seeking to aid his brothers-in-law, the Skeltons, who were intent on killing the lover of his wife’s sister, sent a telegram to a telegraph operator telling him not to deliver another message, warning the intended victim.¹⁷ The court treated this as a case of pure aid, independent of encouragement; moreover, the facts of the case left unclear whether Tally’s telegram had in fact played a significant role in the success of the murder. No matter:

The assistance given, however, need not contribute to the criminal result in the sense that but for it the result would not have ensued. . . . If the aid in homicide can be shown to have put the deceased at a disadvantage, to have deprived him of a single chance of life which but for it he would have had, he who furnishes such aid is guilty, though it cannot be known or shown that the dead man, in the absence thereof, would have availed himself of that chance; as, where one counsels murder, he is guilty as an accessory before the fact, though it appears to be probable that murder would have been done without his counsel ....¹⁸

The court here treads a familiar path, treating the elimination of a possibility of escape as a contribution to murder. This echoes common sense: Tally surely did contribute to Ross’ death, through his efforts. But if we are to speak strictly of causation, it stretches the concept into incoherence to equate the elimination of one possibility of escape, if indeed that possibility might not have been “availed of” with causation of the death. In counterfactual terms, if there is every reason to think the murder would have happened anyway, then the nearest possible worlds in which Tally does not attempt his assistance are worlds in which there is no causal dependence of the killing on Tally’s blocking the warning. In INUS terms, Tally’s help would appear to be a redundant contribution to a set of circumstances sufficient for the death, hence not a cause.

What does go on, and what constitutes the basis for our normative assessment of “contribution,” is a change in the distribution of possible worlds in which Ross escapes (ones in which he listens and adheres to the warning), rather than ones in which he doesn’t receive it, or fails to grasp its urgency, suspects a trick, etc. Tally might have causally contributed to Ross’ death, but it is not necessarily the case that he did. Similarly, had Tally handed the Skeltons an extra rifle, to be used in case theirs jammed, he might have causally contributed to Ross’ death as well—but only if, in fact, the Skeltons’ guns had all jammed. The relevant test for assessing S’s causal contribution is to subtract S’s assistance from P’s resources, to see if P’s crime still occurs: would the Skeltons have killed Ross without Tally’s aid, given that their guns worked? To ask instead, would they have killed Ross with Tally’s aid, if their guns hadn’t worked, is to change the subject, to ask about causation in another possible world. The test is not to remove P’s resources independent of S’ aid, or to alter the world so that S’ aid is rendered more effective—that is question-begging. In the world in which their guns fire, his act of assistance (clearly adequate for accomplice liability) is potentially but not actually causal.¹⁹

There is, finally, a familiar third route, which Larkin himself sought out, in order to show the relevance of causation: to transform assistance cases into encouragement cases, and thus make the causal relation one of difference-making to the principal’s deliberations. Encouragement of a principal who has already formed an intention to commit a crime may make a difference, even a but-for difference, if P’s flagging confidence meant that without S’s encouragement P would not have committed the crime at all, or would have dithered and committed it later. In Larkin’s case, the argument was to no avail, since a basis in assistance was possible. But posit now a pure encouragement case, where Larkin shouts to the burglars, “Make sure you get the good stuff!,” with the intent that this emboldens them in their actions. Doctrinally, common law complicity requires that encouragement have an “effect on the mind” of the principal.²⁰ Shouting “shoot him!” at a deaf person does not make one an accomplice to a homicide. But the requisite “effect on the mind” just has to be the registration of S’s encouragement on P’s mind. And this, while a difference S makes, is frequently not a causal difference in relation to the principal’s act, but only a causal difference affecting the salience of various items in the principal’s deliberative field.

As to the principal’s act (and decision), encouragement places us again firmly in the territory of underdetermination. While it might be possible, after the fact, to assess whether S’s input contributed to the act, the deliberative process is famously inscrutable and unpredictable.²¹ The most we can generally say is that the accomplice’s encouragement raised the likelihood of the principal’s act—a claim far short of causation. But even this counterfactual support is not required, for liability would lie even if it could be shown that S’s attempt at encouragement were actually dissuasive—P might think anything S thinks is a good idea is actually a poor one—or P might think that S’s encouragement is simply irrelevant, as his heart is already hardened to the crime. So long as P heard S, and S intended his words to be encouraging, S is an accomplice. In all these cases, S’s encouragement makes a difference to the world, but it does not make a difference to the realization of the criminal event. The role of complicity is to locate liability despite these processes, not to hitch liability to their indeterminacy. As Sanford Kadish has famously argued, complicity and causality reflect complementary modes of inculpation—though, as I have suggested, one might instead rightly interpret complicity in terms of possible causation.²²

4. In short, liability overflows the causal relationship. What we require for responsibility by Larkin or Tally are acts of a certain type, and thus we require them to make differences to their worlds, creating through their actions worlds in which they exert agency, as opposed to one in which they reside passively. That difference is the basis of responsibility, and the object of responsibility is the target of the accomplice’s act—here, the crime or the cover-up. There are various ways in which one can attach oneself responsibly to another’s acts, before, during, or after the fact, and complicity doctrine recognizes a limited variety of these ways. While some are causal, not all are; and while most involve physical action, speech or otherwise, not all do, for a guard can render himself complicit in a burglar’s theft by doing nothing, deliberate failing to sound the alarm. What binds together all the complicity cases is the mental state of the accomplice—a mental state directed both towards the accomplice’s own agency (including the agency involved in refraining) and towards the agency of the principal.

This double-relation is what makes accomplices specially responsible both for themselves and for their principals. To insist on a causal requirement is to do violence both to criminal law doctrine as well as to the understanding of moral responsibility, which can attach in the more complex inter-relations I have sketched above: through the indeterministic processes of conversational deliberation and inter-personal reaction, or through the counterfactual possibilities of readiness to aid, as here. Indeed complicity doctrine might—as conspiracy doctrine very clearly does—dispense with an act requirement altogether, provided the would-be accomplice’s intent could be sufficiently proved. (This is the lesson of the omission cases, in fact.) But the act requirement is supported by many humane and liberal considerations, none of which is put into question here. Finally, one might ask, as I do here, whether even the boundaries of complicity might stretch further, to encompass acts of ratification and endorsement, as well as to attempts to assist uncompleted crimes. And it is to those topics I now turn.

While my discussion has been couched in terms of criminal liability, it applies to moral responsibility as well, a fortiori. Because of the way assessments of agents’ moral responsibility for wrongs bleeds into assessments of agents’ characters, causation (as opposed to expressions of agency) is even less at play in the moral domain. While some might want to deny S’s moral responsibility for the harms caused by P in some of these cases of overdetermination, they will generally be willing to grant S’s moral responsibility for something, for instance for S’s wrongful acts of aiding in themselves, independent of their consequences, and so they will underwrite the legitimacy of S’s punishment at law. We do, after all, frequently hold people responsible for who they are and the attitudes they express, not just the harms they cause. And we hold ourselves responsible for much that it would be out of place for others to fault us for. We answer for ourselves and our acts in many ways, from hanging our heads low, to committing ourselves to do better in the future, and we are approached in sadness as much as anger for much of what we do.

These are all part of the spectrum of responsibility, and I have discussed a lot of these responses in other work. But I want to return here to the focal case of liability to actual punishment. Given the centrality of harm to criminal law, it would be absurd to deny causality’s centrality to the paradigm of criminal responsibility. The difficulty is picking out the right role for causation to play.

The key, as I suggested above, is to see causation’s relevance in terms of the theory of criminalization, not the theory of responsibility. Causation does distinguish some modes of manifestation from other non-culpable modes of endorsement, such as declarations. But it does so at the level of ex ante consideration as to what types of conduct by S are likely to enhance risks of harm or wrongdoing, taking heed of the special dangers of group polarization and divided criminal labor. That causal generalizations underlie the liability rules does not mean they must underlie particular instances of liability. Against these risk management concerns is balanced our general interest in maximal freedom of action, and clear notice of acts bringing us within the scope of criminal liability. S’s acts of assistance must be ones that, in Kadish’s words, “could have contributed to the action of the principal,” even if in fact there is no reason to think that they did so in the particular case.²³

Similarly, mere undirected endorsement of another’s criminal intent is shielded from liability for reasons derived mainly from concerns with protecting free expression, even expressions that reveal social dangerousness and contribute to general social unease. The presence of causation and absence of complicity in such cases bespeaks the complex relationship between the two. In U.S. law, the role of immediacy in the separate crime of solicitation, or in incitement as a basis of complicity, derive directly from political morality, specifically from libertarian concerns for recognizing the inherent risks of robust political activity.²⁴ To the extent other jurisdictions differ in their treatment of incitement, such differences are more likely to reflect different weights accorded to liberty of speech versus security, not different assessments of causation or moral responsibility.²⁵

Recognizing causation’s limited role in standard doctrine should not just leave matters where they are. To the contrary, once causation is properly located at the type- rather than token-level of risk, there is enormous rational pressure to move in the revisionary direction of the M.P.C., and so to recognize liability for attempted aid. Indeed, I have implicitly suggested that inchoate accomplice liability already exists covertly, in overdetermined encouragement cases. P’s resolute decision to commit a crime will undercut S’s attempt at encouragement, for to attempt encouragement is to attempt successful encouragement. (Since S will also likely intend the success of P’s criminal venture, S may not mind or notice the failure.) From the point of view of normative justification, once we see S as complicit in an attempt to encourage an overdetermined act, there is no reason to withhold liability for cases of attempted aid as well. This is simply an application of a general principle of limiting moral luck in criminal liability, of not distinguishing in terms broader than necessary between actors identical in their morally salient characteristics.

To illustrate, take four variants of the Tally case.²⁶

• (1) Tally blocks the warning telegram with his intercepting order, thus contributing to the Skeltons’ murder of Ross.

• (2) Tally blocks the warning, a Skelton brother shoots at Ross but misses.

Clearly, in both these cases, Tally is guilty as an accomplice, for murder and attempted murder respectively. Now look at two more cases.

• (3) Tally tries to send the intercepting telegram, but a break in the wire blocks its transmission. The Skeltons are nonetheless able to kill Ross.

• (4) Tally tries to send the telegram, but it is not transmitted. Ross leaves town before the Skeltons can find him.

In these two latter cases, the common law would provide for no liability for Tally, in (3) because there is merely an unsuccessful attempt at aid, which could not have been effective under the circumstances; and in (4) because there is an unsuccessful attempt to aid a non-existent crime. However, the M.P.C., and the minority of state jurisdictions that follow it on this point, would convict Tally of murder and attempted murder, respectively.²⁷

Given that Tally’s conduct is identical across all instances and that even the common law grounds liability when chance intervenes between the Skeltons’ shot and Ross’ life, the M.P.C. would seem to have the better position. Tally’s criminopathic agency is pledged to the murder in equal measure, across all four cases. Put aside the chestnut whether punishment ought to be reduced in cases of unsuccessful attempt: To refuse to acknowledge accomplice liability short of failed attempt (4) is to deny application of all the considerations that make attempt liability persuasive in the case of principals. Inchoate liability operates across the dimension of possibility, treating defendants who might have caused harm as if they actually did cause harm. It thus constitutes a central form of causeless criminality. Inchoate liability is fully personal, grounded in the agency of the defendant, and reflects both a retributive response to the unrealized risk of that agency in the bats, and a deterring response to similar projections of that agency in the future. Even if there are disputes at the margins of inchoate liability, most notoriously with regard to the distinction between preparation and attempt, that there should be such liability is nowhere in dispute.

Since actual causation is not required for principal inchoate liability, nor, as I have argued, for accomplice liability, there seems no reason to preclude liability in any of the instances. Evidentiary concerns alone will not tip the balance either. The evidence of S’ attempt (as to both act and intent) may be equivalent in all cases, and is not affected by the commission (or not) of the principal’s crime. If there is a reason to limit liability short of the full reach of the M.P.C., to cover failed attempts, it comes not because S is causally inefficacious, but because in the final instance, the absence of any criminal act from which S’s liability can be derived effectively converts the retrospective structure of criminal liability into a forward-looking model of risk control. While the retrospective and prospective dimensions of criminal law always exist in tense balance with one another, this might be viewed as a case in which the balance has shifted too far. But that conclusion is hardly obvious, and the risk-reduction aspect of criminal law might just as well be firmly embraced, while maintaining vigilance against excessive threats to individual liberty.

The principled defense of inchoate complicity is clear enough, even if it reflects a doctrinal minority, and I will not belabor it further. What I would like to elaborate concerns the dimension of time, rather than possibility. Consider now a last variant of Tally:

• (5) Tally sends the telegram, but it arrives moments after the Skeltons shoot Ross.

No jurisdiction would convict, even under the M.P.C., for aid that arrives after the crime. As the Larkin court said, accomplice liability only attaches to subsequent, not previous, criminal acts. But this is actually mysterious. Once we recognize accomplice liability for no crime at all, seeing it as grounded in actual exercises of criminal agency, there would seem to be a basis for recognizing liability for actual harms, grounded again in actual exercises of agency, even if the agency comes too late. Indeed, even if one is not prepared to recognize unilateral, inchoate accomplice liability (attempted aid for non-existent crimes), there should be room for post hoc aid. Return, one final time, to Larkin. Larkin’s attempted assistance came too late, so the court found him liable as an accessory after the fact, not an accomplice. At the bottom line, this seems normatively unobjectionable: it recognizes Larkin’s moral fault, general social dangerousness, and attachment to a particular criminal act. But, arguably, it distorts the category of accessories after the fact, who are typically involved in certain forms of parasitic, post-crime aid (hiding or disposing of crime evidence) and general thwarting of the machinery of justice, to lump Larkin in there, rather than to recognize him as someone more ambitiously committed to a crime he did not realize had yet occurred. To put the matter more crudely, there seems no reason to give Larkin the benefit of his luck. I do not deny the linguistic peculiarity on the other side of treating someone as “aiding or encouraging” a crime that has already occurred, but normative rather than linguistic considerations would seem the more persuasive.

5. Now return to the torture memoranda, and to the culpability of the lawyers who produced them. Start with the easier cases, the abuses committed after the drafting of the memoranda. The basic defense on behalf of the lawyers was that they did not try to dictate policy, but only provided policy-neutral information concerning the legal parameters.²⁸ However, as I have said, the aggressiveness, one-sidedness, and argumentative overreaching of the memoranda clearly reveal an argument for a specific, and highly controversial position: that the detainees are to be denied Geneva protections, and that interrogations should proceed in light of military necessity, without regard to contrary domestic and international law. As such it amounts to encouragement of the war crimes and domestic crimes that occurred, moreover successfully communicated encouragement. By providing cover, the memoranda also amounted to aid given to the executive branch in the acts of abuse it ordered committed. Mens rea is present; the question is thus whether the actus reus suffices as well.

So the second line of the defense must be that whether or not the O.L.C. lawyers intended the abuses, they did not cause them. And we now have the resources to dispose of that argument as well. It may be true that the lawyers did not cause, or causally contribute to, the wrongful acts of any CIA or military personnel. Perhaps the policy would still have been enacted, even without the political and legal cover the memoranda provided; secrecy might have been even tighter, or greater inducements offered to legally exposed personnel. But even if lawyers’ acts did not (in fact) make a causal difference to the occurrence of abuses, there is still a robust basis for their implication. They stand in the same position as any provider of redundant encouragement or unneeded assistance to a resolute or well-supplied principal. That their advice could have made a difference will be clearly sufficient for liability.

Next take the unintended abuses at Abu Ghraib that were a causal consequence of the executive policy decisions. The issue here is not causation, then, but mens rea. In the U.S., although perhaps not in the U.K., complicity would require lowering the mens rea standard from intent (or knowledge, in the U.K.) to recklessness.²⁹ This standard could probably be met, as the risks of loss of military discipline and unregulated aggressive and abusive detentions were clearly flagged by the military lawyers who objected to the policy. There are strong arguments against lowering the mens rea standard of accomplice liability to knowledge, let alone to recklessness; but if there is a case for lowering it, it ought to apply here, where both magnitude of the relevant risks and awareness thereof were extremely high.³⁰

The conceptually tricky case involves acts of torture that might have occurred before the drafting of the memoranda. The lawyers’ liability for these abuses would have to be established in terms of post hoc ratification, if it can be established at all. The moral case for their taint by previous acts of torture is clear. But there is a legal case to be made as well. Their endorsement of those acts was far more than an expression of approval: it was an act with real institutional consequences, among which were the reductions in the risks of domestic prosecution for those acts by personnel engaging in them in the future. Since the memoranda were yoked to both (possibly) prior acts as well as future ones, there seems no good reason to acquit in the former instance and convict in the latter.

The Larkin alternative, of resting with accessory after the fact liability, is always a possibility. But this seems to me a limited context in which actual post hoc accomplice liability is present; and indeed, despite the disfavor of ratificatory criminal liability, it gets some recognition in civil law. Ratificatory liability, incurred for ex post approvals of unauthorized exercises of agency, serves both a retrospective function of giving third-parties recourse on the basis of their reasonable expectations concerning an agent’s authority, as well as a prospective function of putting principals on guard against incautious agents. While ratification has not been a recognized basis of criminal liability, there are good reasons to recognize it in well-defined contexts, where the act ratified is one likely to be repeated—indeed more likely, given the encouragement of the ratification.

Indeed, the case for ratification is well supported on moral, practical, and epistemic grounds. Morally, ratification expresses an agent’s identification with the acts of another, reflecting here an endorsement of a wrongful act. Practically speaking, ratification reveals the dangerousness of the ratifying accomplice and exacerbates the dangerousness of future principals (in the criminal sense, now). Epistemically, evidence of ratification may be all that surfaces, and may give rise to a warranted inference that there was prior subtle encouragement, and hence full concert. Here we must bear in mind that this is fundamentally an institutional context, and that in such contexts acts of ratification have forward-looking effects as well. In this case, the forward looking effects are obvious: they just consist in the further acts of abuse that occurred. Conceptually, ratification of prior acts in contexts where that ratification may have future effects looks a lot like a failed attempt to aid or encourage the prior acts, where the aid or encouragement is of a type that is generally appropriately targeted by criminal law because of its possible consequences. The reason for the failure of the attempt is its temporal location, not some spatial or causal impedance.

Whatever we conclude about the limiting case of ex post liability, the principled argument for some form of liability for the torture lawyers is clear, even for the crimes that came before their own efforts to torture the law. While causal generalizations about the effects of their acts make punishment fair, we must not lose sight of one important basis of responsibility: they were asked what they thought about what their government was already doing, and instead of saying at least, “wait a minute,” they instead answered, “good idea, boss.” That way responsibility lies.