1
I also have belonged to those groups of men who believe they can produce the truth with white-hot iron. Well, let me tell you, the white heat of truth comes from another flame.
—UMBERTO ECO, The Name of the Rose
Shortly after 3:00 am in the early morning darkness of March 28, 2002, on Canal Road in Faisalabad, Pakistan, Abu Zubaydah heard the front door being kicked in. Grabbing some fake passports and cash, he ran to the roof, sprinted to the edge, and jumped to the roof of the neighboring house one story down. Four Pakistani police officers were waiting. As one of Zubaydah’s compatriots struggled with a policeman’s AK-47, the gun went off. The single bullet hit Zubaydah in the thigh, shattering coins in his pocket, ricocheting off his femur, and sending shrapnel into his groin and stomach.1
As he lay bleeding in the back of a police pickup truck, he was nearly mistaken for a low-level terrorist and summarily shot by a Pakistani officer. FBI and CIA agents, however, recognized him as the target of the raid and he was taken in the back of the pickup to a local hospital. His condition worsening, Zubaydah was flown by military helicopter to a Pakistani military hospital in Lahore. The CIA officer accompanying him tied him to the bed with a sheet and stayed with him as he drifted in and out of consciousness. Within a few days he was loaded into a rickety old ambulance and driven to the Lahore airport.
A white, nondescript Gulfstream jet, probably with the tail number N379P, was waiting on the tarmac for him. The FBI and CIA loaded Zubaydah on the plane and secured him, and the plane took off. Although he was unaware of it, Zubaydah then went on a U.S. taxpayer-funded world tour over the next several days. The CIA changed pilots and stopped off on different continents in order to disguise the final destination: a secret prison near Udon Thani, an hour by car northeast of Bangkok. He was met at the “black site” by a CIA doctor, anesthesiologist, and medic who had prepared a field hospital for him. Two FBI agents, Ali Soufan and Stephen Gaudin, both experts on Al Qaeda, were also waiting.
Soufan and Gaudin got to work right away, questioning Zubaydah when the medical team was not cleaning his multiple wounds. They even helped care for him, holding ice to his lips and cleaning him up after he soiled himself. Their combined efforts were not enough, however, and Zubaydah began to develop sepsis. The CIA station chief hurriedly cooked up a plan to disguise Zubaydah as a soldier and they rushed him to a hospital. Only a tracheotomy and hand-pumping air into his lungs during the ride to the hospital kept him alive. The surgery saved his life and when he awoke, Soufan and Gaudin immediately began questioning him, using an Arabic letter chart to communicate since Zubaydah had a breathing tube in his mouth and could not speak. On April 8 the breathing tube was removed and Zubaydah continued to cooperate and provide information to the FBI agents. On April 10, 2002, he identified a picture of Khalid Sheikh Mohammed (hereafter KSM) as “Mukhtar,” the mastermind behind 9/11.
While Zubaydah lay in the hospital, talking to Soufan and Gaudin, a new CIA team sent from CIA headquarters in Langley, Virginia had renovated the black site to include a special cell, with hidden cameras and microphones to record Zubaydah’s every move. The team from the CIA’s Counterterrorism Center (CTC) included the CIA’s chief operational psychologist, an interrogator, a polygrapher, analysts, as well as support and security staff.
It also included James Elmer Mitchell, a military psychologist who had never even observed an interrogation, let alone conducted one. Mitchell spoke none of the relevant languages, had no background knowledge of, or experience in, the Middle East, and knew nothing about Al Qaeda or terrorism. What he did have was experience with the military’s SERE program and, reportedly, a thousand-plus-dollar-a-day contract with the CIA. SERE stands for Survival, Evasion, Resistance, Escape, a program designed to train service members how to resist torture by actually undergoing limited versions of it under tightly controlled conditions. Mitchell’s contract with the CIA was to reverse-engineer these methods to get Al Qaeda detainees to provide intelligence. Mitchell’s presence, in other words, signaled that Zubaydah would face a very different approach to interrogation upon his return.
On April 15, 2002, a doctor sedated Zubaydah and a CIA team drove him back to the black site. He awoke four hours later, manacled to his bed, in an all-white room, brightly lit with four halogen bulbs. His guards were covered in black from their balaclavas and goggle-covered heads to their boot-encased toes and they communicated with each other using only hand signals. His hands were cuffed and his legs manacled to the bed or chair, with intermittent loud music played in his cell. The new interrogation plan conceived by Mitchell originally called for isolating Zubaydah, but since his medical care precluded this, the team opted the next day for interrogating him around the clock, with short breaks for sleep.
For the remainder of April the CIA and FBI questioned Zubaydah, who provided information on Al Qaeda plans, capabilities, leadership, decision-making, training, and tactics, as well as information on terrorists in Pakistan. On April 20 he provided information which eventually led to the capture of Al Qaeda operative Jose Padilla, the so-called dirty bomber. He also provided information on plots to blow up apartment buildings and the Brooklyn Bridge, as well as other plots.
Despite this cooperation, CIA analysts on the ground at the black site and back at headquarters in Langley remained convinced that Zubaydah was holding back information on pending attacks on the United States and operatives residing in the United States. As a result, in early May the CIA accepted Mitchell’s proposal to start employing coercive measures, including sensory deprivation. Once again Zubaydah was stripped naked, manacled to a chair, and subjected to loud music. This time, white noise was added to the coercion playlist and he was deprived of sleep for 48 hours. At this point a CIA medical officer decided that the methods had crossed the line of what constituted torture and, afraid for his ability to practice medicine in the future, returned to the United States. Soufan himself did the same in May when he discovered a large box in the interrogation area and learned that Zubaydah would be confined in it. Special Agent Gaudin reported through a different chain of command in the FBI and remained for several more weeks, participating in the CIA-directed interrogations until early June 2002, when he too was ordered back home by the FBI. From that point on, the CIA had full control. During May and the first half of June, Zubaydah continued to provide information, but not on pending threats or terrorist cells in the United States. Since both local CIA analysts and their bosses back in Langley were convinced that Zubaydah had this information, they switched tactics.
On June 18, 2002, Zubaydah was put into his cell. He remained there, in isolation for 47 days, until August 4. Not one question was put to him.2
In the meantime, the CIA began seeking written approval from the Department of Justice to employ more coercive techniques. They got it in the August 1, 2002, Department of Justice Office of Legal Counsel memo to the CIA (Bybee 2002). The memo identified ten “Enhanced Interrogation Techniques” (EITs) that CIA interrogators could use on Zubaydah: attention grasp, walling (pushing or slamming a detainee against a flexible wall with the neck wrapped to prevent whiplash), facial hold, facial slap, cramped confinement, wall standing (leaning against a wall supported only by the fingers for lengthy periods), stress positions, sleep deprivation, confinement in a box, possibly with a feared insect, and waterboarding (Bybee 2002, pp. 2–4)). With the approvals in hand, the CIA was ready to go back to work on Zubaydah.
Ten minutes before noon on August 4, 2002, the black-clad “ninjas” entered Zubaydah’s cell. In total silence, without asking a single question, they shackled and hooded him and removed his only article of clothing, a towel wrapped around his waist. Wrapping a towel around his neck, they swung him around and slammed him into the concrete wall. They removed his hood and grabbed his face, making him watch while a large box was brought into the room and placed on the floor to resemble a coffin. Interrogators told him the only way he would ever leave the secret prison was in that box.
Interrogators demanded he provide “detailed and verifiable information on terrorist operations planned against the United States, including the names, phone numbers, email addresses, weapon caches, and safe houses of anyone involved.” Zubaydah’s denials of knowing any of that information were met with slaps and grabs to the face. Over the next six hours he was shut up in the large box, pulled out, walled, forced into a smaller confinement box less than two and one-half feet square, and put in stress positions. Around half past 6:00 pm he was tied to a gurney and waterboarded for the first time. Over the next two and a half hours, he was waterboarded multiple times, repeatedly coughing, vomiting, and involuntarily spasming against his restraints. At one point, he fell unconscious, “bubbles rising through his open, full mouth,” requiring medical intervention to revive him. At 8:52 pm interrogators stopped waterboarding.
“Aggressive interrogations,” however, did not. For the next 20 days straight, 24 hours a day, Zubaydah was walled, grabbed, slapped, and forced into stress positions. Altogether he spent over half of that entire period (266 hours or 11 days, 2 hours) in the large (coffin size) confinement box and over a day (29 hours) in the smaller box. He was waterboarded two to four times a day, with multiple applications each time. When he was left by himself, he was put into a stress position, left on the waterboard with his face covered, or confined in one of the boxes. He was kept nude, dirty, sleep deprived, hungry, and subjected to loud music or white noise. His previous wounds began to reopen under the strain, but CIA headquarters stipulated that interrogations “took precedence over preventive medical procedures.”3
Zubaydah continued to provide general information on Al Qaeda before, during, and after the use of the EITs. He did not, however, ever provide information on terrorist cells or operational plans for attacks against the United States.
Bush administration officials as well as many (but not all) in the CIA have claimed that the techniques saved lives because of the information they generated about future attacks. Did they? Did—does—interrogational torture work?
Quaestio in early Roman law meant only “questioning” or “interrogating” witnesses. As torture worked its way into Roman jurisprudence from the second and third centuries C.E., the term “became synonymous with torture” so that the same word in Medieval Latin had come to mean “torture,” as did the Old French question (Lea and Peters 1973, pp. x–xi).4 Water torture, for example, became known as question d’eau.
Quaestio quaestionum—interrogating (interrogational) torture—is the purpose of this book. The Zubaydah case illustrates some of the questions which must be raised and answered in any appraisal of an interrogation program using coercive techniques. Do the EITs of the CIA and the aggressive counter-resistance techniques of the military rise to the level of torture (Haynes II 2002, p. 1)? Who, exactly, is the detainee? How much does the detainee know (if anything)? What is the value of what he knows? How do you know whether an interrogation technique is working? How does an interrogator know when to stop, when a detainee has told all he knows? Is there a slippery slope problem with interrogational torture or can it be strictly controlled? Will innocents be tortured? Will they falsely confirm information to avoid torture? How reliable is coercive interrogation generally? Does it work most of the time, some of the time, or only rarely? How much bad information is generated?
KANTIANS, KHANIANS, AND PRAGMATISTS
Chapter 2 assesses whether the EITs amounted to torture and what we know about the effectiveness of interrogational torture. For many, of course, the question of torture’s effectiveness is irrelevant: Torture is unjustified whether or not it is effective. Call this rights-focused group Kantians after the famous German philosopher Immanuel Kant. For others, the degree of torture’s effectiveness is just as irrelevant, but for a different reason: The 9/11 attacks provided the United States all the justification it needs to torture suspected terrorists. Call this group Khanians, after Genghis Khan, who is said to have tortured someone who had wronged him by “by pouring molten silver into his ears and eyes.”5 For this group, vengeance, not intelligence, is the goal.
There is also, however, a third group of Americans for whom torture is justified only because, and insofar as, it is effective, and who would oppose it otherwise (call this group Pragmatists). In a 2011 survey, I found that while fewer than one in six Americans consider himself or herself a Kantian and the same is true for Khanians, more than four in ten consider themselves Pragmatists. This group, nearly two-thirds of all those who think that harsh interrogation methods can sometimes be justified, says that the techniques would not be justified if they were not effective.6 A December 2012 Huffington Post/YouGov poll found that a little more than a third of Americans are “not sure” whether information gained from torture is reliable, only one point below the percentage that thinks it is unreliable (a further 29% thought that information from torture was reliable).7 Two years later, after the release of the Senate Intelligence Committee report, Huffington Post/YouGov found that the “not sure” group had jumped to 37%, with the “reliable” vs. “unreliable” camps evenly divided at 31% and 32%, respectively.8 So more than one in three Americans are not sure whether interrogational torture works, and four in ten say it is justified only if it does work.
Thus, determining whether interrogational torture works—and at what human cost—is essential. It may be that interrogational torture cannot be justified under any circumstances, but if it is to be justified at all, it must be effective. First, if it does not work, then the only possible public justification for its use—it is the only method we have left when detainees refuse to talk—disappears and any further use is pure sadism. Second, and just as importantly, if it does not work, then this means that interrogators are not getting the information necessary to save lives. Even its proponents would argue against it under these conditions. As Bagaric and Clarke admit in their pragmatic defense of torture, this is a “knock-down argument” for “if this objection were valid [they] would change [their] minds and not countenance torture in any circumstances” (Bagaric and Clarke 2007, p. 53).
A NEW APPROACH: GAME THEORY
I examine whether or not there is such a knock-down argument in this book. Kantians will find me rhetorically complicit in torture for even considering the question of effectiveness.9 Khanians want to torture out of revenge and will not be persuaded by reason and logic about effectiveness. Pragmatists—perhaps you fall into this group—may very well, however, be open to reasoned and logical arguments evaluating the supposed effectiveness of interrogational torture. Moreover, given the absence of sufficiently reliable and systematic evidence to convince Pragmatists, reason and logic are the only alternatives we have left.
The preeminent vehicles for making reasoned and logical arguments are formal logic and mathematics.10 A branch of mathematics called game theory is particularly applicable to the problem of interrogational torture. If you’ve ever had to merge lanes in traffic, then you already know a little about game theory because you’ve actually played it. Although everyone might be happier taking turns to merge, everyone is afraid that others might jump ahead and that they will be stuck in traffic longer. As a result, everyone crowds up, honks, and yells, and the traffic slows down even more. Everyone is behaving rationally, in their own self-interest, and the result is everyone stressed out and late for dinner.
The film A Beautiful Mind, about the mathematician and game theorist John Nash, provides another, if dated and sexist, example. There is a scene in the movie where Nash and four other (straight male) mathematicians are at a bar and five women walk in, one of whom is a blonde. The men start arguing about who should go for the blonde, but Nash, in a game theory epiphany, tells them that if they all hit on the blonde they will “block each other” and make the other women angry for being made to feel second fiddle. As a result, none of them “succeeds.” Consequently, Nash argues, it makes more sense for them all to court the brunettes and ignore the blonde. The brunettes will respond favorably and everyone is happy (except the surprised blonde).11 Thus, game theory is a way to describe and think about situations in which two or more people have choices to make, and which action they end up taking affects both: (1) what action everyone else wants to take as well as (2) what each person gets in the end, after everyone has chosen.
Cop shows provide a more familiar example. In the third episode of the first season of the popular crime show Law and Order SVU, Detectives Benson and Stabler are questioning a suspect, a woman named Deborah, in the murder of one fashion model and the brutal assault on another. They catch her in a lie and Detective Stabler says to her: “Why don’t you talk to us while we can still help you?” She responds with “I’ve enjoyed about as much of your help as I can take.” Stabler and Benson then walk out of the interrogation room and ask Detectives Munch and Cassidy whether they got anything on the other suspect, Carlo Parisi, whom Munch and Cassidy had interrogated. When Benson learns that neither team had gotten their suspects to admit to anything, the following exchange takes place (with some tweaks to the script)12:
BENSON: Munch, do us a favor.
MUNCH: What?
BENSON: Get Carlo to sit in this chair and tell him his options. We’ve got enough on him with the statutory rape charge to send him upstate for 5 years. If he tells us what happened, we’ll do what we can to get that reduced to a lesser charge so he’s out in a year. On the other hand, if Deborah rolls on him, and he stays mum, we’ll get him on the murder one charge and he’s going away for at least 15. If they both cooperate, we’ll have more evidence against them but still help them out and they’ll each get 11 years.
STABLER: We’ll sit Deborah by Carlo so they see and hear everything and then give her the same choice. She’s already admitted to scoring crank at Till’s party, so we’ve got 5 years on her too, even without anything from Carlo.
MUNCH: Ah, cute. The Prisoner’s Dilemma.
CASSIDY: What’s that?
MUNCH: See, no matter what the other one does, they each spend less time in prison if they talk. They both know this, so they both talk. And we put them both away for 11 years each, even though if they could just both bite their tongues and wait it out, they’d be out in 5. But they never do!
CASSIDY: Damn, I love this job. I’ll get Carlo.
Carlo and Deborah are playing a Prisoners’ Dilemma. If Carlo keeps quiet, Deborah has two options. She gets 5 years if she does not talk and 1 year if she does. So she is better off talking in that case. If Carlo talks, Deborah again has two choices. She gets 15 years if she stays quiet and 11 years if she talks too. So she is better off in this case talking as well. In other words, Deborah is always better off if she talks (assuming, as we are, that she wants to spend as little time in the slammer as possible). So she will choose to rat out Carlo. And since it is exactly the same situation from Carlo’s perspective, he will do the same thing and the result is that each spends 11 years behind bars—6 years longer than if they had both managed to “keep their big mouths shut.”
I have summarized the situation in Table 1.1. Deborah’s choices are the rows and Carlo’s are the columns.
Table 1.1 LAW AND ORDER SVU PRISONERS’ DILEMMA
A row–column combination is an outcome, a combination of Deborah’s choice and Carlo’s choice. Each outcome, that is, each row–column combination, has a pair of payoffs—in this case a jail sentence, one for Deborah (on the left) and one for Carlo (the sentence on the right). Take, for example, the combination of choices where Deborah stays silent (the top row) and Carlo rats her out (the right-hand column). Deborah gets 15 years and Carlo gets 1 year, as just described above.
I have bolded the “Rat Out”–“Rat Out” outcome because that is the equilibrium of this game.13 Equilibria are where game theory’s rubber hits the road. They are the outcomes of the game; they tell you what will happen. The idea is as straightforward as it is important. An equilibrium is just a stable combination of actions; neither player wants to make a different choice given what the other player has chosen. To help fix this idea, go back to the bar scene in A Beautiful Mind for a moment. Is it an equilibrium for all the mathematicians to go after the brunettes (assuming that there is an additional brunette and that the men all prefer the blonde, as in the movie)?
Imagine that you are John Nash. If your friends are playing this strategy—they are hitting on the brunettes—do you have incentive to switch your strategy—from the remaining brunette to the blonde? Yes, you do. You will not be blocked by your friends, so it makes sense to go for it. You would switch your strategy and go for the blonde. But of course, the exact same reasoning is true for your friends, so they would switch too. Consequently, Nash’s proposal in the movie is not an equilibrium at all.14
Return now to the bolded equilibrium of the Law and Order SVU game. Deborah, for example, would not “go back in time” after Carlo has chosen “rat out” and switch her move to “stay silent” even if she could, because that would move her up to the outcome where she gets 15 years instead of 11. The same is true for Carlo, so neither has an incentive to switch her or his strategy. In contrast, the outcome in which Carlo “stays silent” and Deborah “rats [him] out” is not an equilibrium because Carlo does have an incentive to switch his strategy (because he would get 11 years instead of 15 years). Munch was right about it being “cute” (at least for the detectives). Seemingly weirdly, individual rational behavior results in an outcome in which the players are worse off than if they had somehow managed to stay silent.
These are just a few examples, but they should give you the flavor of the game theoretic approach. You have people (players) who have choices to make, with those choices (strategies) leading to various outcomes, each with its own set of rewards or penalties (payoffs). Importantly, the choices that one player makes—that is, the strategy that he or she chooses—affects both (a) the strategy that the other player wants to take and (b) the payoffs to both of the players. Their interaction is strategic. The same is true of interrogational torture, and that makes game theory applicable, useful, and powerful.
Interrogational torture is not a game in the sense of being fun, but it is a game in this strategic sense. The interrogator is trying to figure out what the detainee knows. If the detainee is innocent, he is trying to find a way to convince the interrogator of that fact. If the detainee does have information, he is trying to find a way to avoid giving it up. So game theory is actually pretty well suited to model the kind of interaction that takes place between a detainee and an interrogator. Game theory allows us to model the incentives and the strategies of the detainee and the interrogator and see what would happen.
But it is more than merely applicable. What makes game theory useful are the outcomes that are derived from the game modeled on interrogational torture: the equilibria. These outcomes are basically combinations of detainee and interrogator actions, just like the “rat out”–“rat out” equilibrium in the Prisoners’ Dilemma. So in interrogational torture, one outcome or equilibrium might be “talk,” “don’t torture” while another might be “don’t talk,” “torture.”
We will do the same thing for interrogational torture. We will compare the predictions, the equilibria, emerging from our game theoretic model of interrogational torture to what torture proponents say happens when interrogational torture is used in real life. In the usual course of social science research using game theory, the model generates predictions which are then tested empirically, using data from the real world to assess how useful the model is in explaining the phenomenon of interest. In contrast, we will use the predictions from game theory as substitutes for the data we do not (and never will) have and will compare them to the claims—the predictions—made by torture proponents about its effectiveness (Laver 1997, p. 4).
In other words, we can use the equilibria in place of the data we will never have to identify the conditions (if any) under which a detainee provides clear and valuable information. Modeling interrogational torture also permits us to assess the frequency of torture by identifying how often torture—including torture of an innocent detainee—occurs in equilibrium. Finally, it is possible to use some of the formal mathematical results from the model to derive some implications about the likely severity of torture when it is used. In other words, the outcomes of a game theoretic analysis will help us answer the important questions I raised at the outset.
Game theory, however, offers even more. It is not only applicable and useful, it is also powerful. As a type of deductive reasoning, game theory relies on systematic, logical reasoning to derive outcomes (equilibria) from a set of clearly stated assumptions. You cannot be wishy-washy using game theory. You cannot hide behind vague and unstated assumptions and you cannot waffle in your reasoning from those assumptions to your conclusions. All of this makes a game theoretic argument, like other forms of deductive arguments, valid.
When the assumptions are also true, when they correspond pretty closely to the real world, and when you validly derive outcomes from those assumptions and those outcomes also correspond to what happens in the real world, then you have something powerful: a valid and true argument. That gives us some confidence that we are actually getting at the real-world phenomenon that we are trying to model with game theory. Finally, the formal mathematical properties of game theoretic models often provide unexpected and counterintuitive insights and relationships which might otherwise go unnoticed.
RATIONALITY, TORTURE, AND PAIN
Before saying a little more about exactly how we will do this, it may be helpful first to address two understandable objections to applying game theory to interrogational torture, one ethical and one about the relationship between pain and our ability to make choices. The ethical objection is that it is dehumanizing to reduce the pain of torture to a bunch of numbers. How can you reduce the excruciating pain and fear of waterboarding, for example, to a negative payoff? Is it not monstrous to try and do so?
This understandable instinct is, however, rooted in a common but significant misunderstanding conflating reduction and representation. A game theoretic model does not reduce anything. Reduction entails a complete explanation of the complex whole of one thing in terms of something (or several somethings) simpler or more elemental constituting it. We could, in principle anyway, reduce cooking to chemistry by explaining the transformation of starch, sugar, and yeast into a French pastry. We could go further and explain chemistry in terms of physics.
Fine, but game theoretic models (or other models for that matter) do not reduce anything. The represent; they do not reduce. Models, like other representations of things in the world, simplify, but do not reduce, reality to re-present it in a way that is more intelligible, meaningful, or revelatory for particular purposes (Clarke and Primo 2012, pp. 4–9), passim).
Consider some different models, or representations, of torture. Vann Nath, a former prisoner of the Khmer Rouge in the infamous Tuol Sleng prison in Cambodia, painted scenes of torture in the prison, including waterboarding.15 Or how about a musical representation or reflection on torture? The Palestinian composer Suhail Khoury was inspired by his torture in an Israeli prison to write “Shabeh,” an instrumental piece performed by the group Karloma on the CD Jerusalem after Midnight.16 Finally, Thomas Kennedy offers a fictional literary representation in his moving story of Chilean torture victim Nardo and his struggle to love again in the novel In the Company of Angels. At one point, Nardo speaks in the third person of the tortures he suffered, of having his hands bound behind his back and “his head plunged into a tank of water afloat with excrement” (Kennedy 2010, p. 164).
Each of these is a re-presentation of torture; the first and third are versions of water torture. It would be inaccurate, if not downright insulting, to charge these artists with “reducing” torture to splashes of color, bundles of air vibrations, and blotches of ink. Each re-presents the reality of torture in a different way for different purposes, but does not reduce torture to the medium of that re-presentation. Nor, because they simplify, do they capture everything about torture (Laver 1997, p. 1). The same will be true of a game theoretic model. Our model will not “reduce” pain to numbers, nor will we claim that the model captures every aspect of, or perspective on, interrogational torture.
The second objection also stems from the fact that pain is front and center in torture. The idea here is that the pain and domination in torture are so overwhelming that they rule out any reasonable conception of choice, preference, calculation, or strategy—in short the ability of the individual to make choices, as is required by game theory (Howes 2012). Wisnewski and Emerick (2009), for example, present a critique of pro-torture arguments rooted in the assumption that torture strips its victims of their human agency—that is, their ability to make choices. In doing so, they rely heavily on the philosophers Elaine Scarry and David Sussman.
Scarry, who presents perhaps the most extreme “anti-agency” or anti-choice position, asserts that “torture systematically prevents the prisoner from being the agent of anything” (Scarry 1985, p. 47). The pain caused by torture is “world-destroying” because it results in the “obliteration of the contents of consciousness. Pain annihilates not only the objects of complex thought and emotion but also the objects of the most elemental acts of perception” (Scarry 1985, pp. 29, 54). In other words, the pain from torture is so transformative that people can no longer make choices and can no longer act rationally. Indeed, Scarry goes so far as to claim that the physical pain of torture destroys language itself (Scarry 1985, pp. 4, 6, 19, 49, 54, and passim).
If this were true, and not empirically absurd, there would be no interrogational torture whatsoever: Interrogation presupposes that the person being interrogated can use language. If the actions of actual torture victims are not enough to belie such claims, studies of torture victims demonstrate that many of them not only maintain their agency, but actually enhance it (Rejali 2007, p. 442). Scarry can maintain this position only because she repeatedly rejects the possibility that the motivation for torture could be information, so that “what masquerades as the motive for torture is a fiction” (Scarry 1985, p. 28, also pp. 12, 20, 29, 57, 329–330).
While it is clear that torture is also used for punishment, for intimidation, and to compel confessions, it is just as clear that torture has been employed for information in the past and present. Scarry herself, of course, admits this reality. “But for every instance in which someone with critical information is interrogated, there are hundreds interrogated who could know nothing of remote importance to … the regime” (Scarry 1985, p. 28, also p. 29). This is a claim about the rarity of prisoners with information, not the nonexistence of interrogational torture. This sort of “uncritical indignation” does no favors for the anti-torture argument (Peters 1973, p. xxii).
In short, a desire for information actually does and did motivate much state torture, is perceived by the public and by (some) ethical theorists as motivating torturers, and is the only possible public defense of torture by proponents. Each is sufficient to scrutinize claims of torture’s effectiveness; jointly they constitute a compelling case.
Sussman (2005), pp. 14–15) would also seem to argue against the ability to apply game theory to torture:
[T] orture is not only a violation of the value of rational agency, but a violation that is accomplished through the very annihilation of such agency itself, if only temporarily or incompletely.
The caveat “incompletely” is key here, however. Sussman admits as much when he says that “the victim is presented with a dilemma about submission or resistance.” This suggests that some room for choice, for action, does indeed remain.
The crucial point is that interrogational torture presupposes (rational) agency on the part of the victim. If an interrogator walks into an interrogation room to find the prisoner a babbling madman (and believes the babbling is not an act), it makes no sense to torture him. Sussman himself makes this case quite persuasively. After drawing a nice distinction between coercion, which requires the coercee to possess rational agency in order to respond to the incentives created by the coercion, and brainwashing, the goal of which is to replace the agency of the victim with the agency of the brainwasher, he positions torture somewhere between the two. “Any particular act of torture will tend to shade off in one direction or another: [T]he more effectively the torture undermines the victim’s rational capacities, the less effectively it can also be coercing him by appeal to his incentive structure (and vice versa)” (my emphasis) (Sussman 2005, pp. 10–11)).
In other words, the philosophical position that torturers totally dominate their victims adopts the same caricature of interrogational torture presented by its strongest proponents. According to this view, torturers know how to inflict pain so effectively that all room for agency, calculation, strategizing, and choice is removed. The torturer controls the environment so completely, breaks the victim down so thoroughly, that he does exactly what the torturer desires. The will of the torturer becomes the will of the tortured.
This is exactly what the designers of the Bush interrogational torture program believed: They would “break” detainees who thought they could hold out and make them willing to talk. There are, unfortunately, plenty of first-person accounts demonstrating that such a model—whether in the guise of the television program 24, in the mind of Dick Cheney or SERE psychologists, or in a work of political philosophy such as Scarry or Wisnewski—is a fantasy.
Take the case of the man
who, after repeated tortures and revocations [of his confession under each torture session], when asked by the judge why he retracted his confession so often, replied that he would rather be tortured a thousand times in the arms than once in the neck, for he could easily find a doctor to set his arm but never one to set his neck (Lea and Peters 1973, p. 123).
It is clear from passages such as this, from torture victims’ own accounts, rather than the reflections of philosophers, that they value things—avoiding pain, protecting information, convincing the interrogator that they are innocent—and order them in particular ways. The fact that torturers also want something from their victims—information or even a confession—indicates that they have goals too. Many victims, by their own accounts, think of themselves as having at least two actions, confessing or not confessing, providing information or not. The same is true of interrogators in their selection of whether or not to torture. Finally, both sides understand, even if imperfectly and under uncertainty, how those actions lead to outcomes.
OVERVIEW OF THE ARGUMENT
Heading off the potential ethical and philosophical objections clears the way for applying game theory to the question of interrogational torture’s effectiveness. Since, as the next chapter will show, we cannot answer the question empirically—that is, with data from the real world—then not only can we apply it, but we ought to do so. How will this work?
Our starting point is to build a model of interrogational torture based on the vision of its proponents, of the pragmatists who say it is necessary in some circumstances. This vision is a pragmatic, normative model of how interrogational torture should work, or was supposed to work, according to its proponents. CIA and other government documents are full of safeguards and checks and procedures designed to limit and control torture. The CIA may have exceeded them, but these need to be incorporated into our analytical model just as much as does the reality of what goes on in an interrogation booth. The analytical model can be neither a strawman used to argue against interrogational torture, nor a quixotism used to argue for it.
In addition to identifying the constraints on the game theoretic model, the pragmatic model has another objective: identifying the criteria against which the pragmatic model will be assessed as effective or not effective. As an example of this idea, take the recently created Consumer Financial Protection Bureau (CFPB). Ignoring the partisan bickering for the moment, note that this is a government program designed to achieve certain goals, including “write and enforce rules for banks and other firms, aiming to protect consumers from deceptive and abusive loans and other financial products and services” and “monitor and report on markets for consumer financial goods and services … and how consumers interact with them.”17 The goals suggest the criteria by which it should be measured, the “benchmarks” it should achieve, to use government and corporate management-speak. Data can then be gathered and compared against these benchmarks. For example, according to the CFPB website, “interest rate hikes on existing accounts have been dramatically curtailed,” late fees and overlimit fees have been reduced, and credit card costs are clearer.18 These data suggest that the CFPB may be making some progress toward its goals. If it did not meet these benchmarks, then the CFPB would be considered less successful. Now, you may be a Republican opposed to this agency on principle and thus may disagree with the very goals of the program. But if the program fails to achieve even its own goals, that is a real problem even for those Democrats in favor of it.
Our pragmatic model of interrogational torture will work in the same way. It will identify some of the criteria by which pragmatic proponents of interrogational torture—not opponents of interrogational torture—would judge its success. Once we have done that in Chapter 3—that is, once we have identified both the constraints on the way interrogational torture would be practiced in our model as well as identified some of the criteria of success against which to compare the outcomes of the model (the equilibria)—we can build a game theoretic model of interrogational torture in Chapters 4 and 5.
Having constructed a model that respects the proponents’ view, we then solve it in Chapter 6, just as we solved the Prisoners’ Dilemma when we found the unique Nash equilibrium. Solving the game means finding the equilibria. These equilibria have important general properties we explore in Chapter 7 before examining each equilibrium and its particular properties in Chapters 8 to 11. Each of those chapters also narrates a case study of interrogations and torture from the real world to illustrate the equilibrium under discussion.
The equilibria and these general properties are our substitutes for the missing empirical data to compare against the benchmarks, and this is the task we take up in Chapter 12. Chapter 13 goes beyond describing what happens when torture is used for information to explain just why what happens happens as well as reflects on the use of formal models in political philosophy. Finally, a Postscript compares the predictions of the pragmatic model as well as those of the game theoretic model to the findings of the investigation of the CIA detention and interrogation program by the Senate Select Committee on Intelligence, released after the previous chapters were largely complete.
This sketches the argument that I will make in the rest of this book. To provide a bit more detail, here is a road map through that argument, chapter by chapter. Each chapter ends with this format, providing a short summary of each step reached in the argument to that point as well as identifying the next step.
The argument proceeds by answering the following questions:
1. Are the EITs torture? What do we know about the effectiveness of interrogational torture? (Chapter 2)
2. How closely does the Bush program approximate the ideal model of interrogational torture? What are the limits on torture in the pragmatic model? What benchmarks define the success of the pragmatic model? (Chapter 3)
3. What are the outcomes of the Bush Interrogational Torture (BIT) model? (Chapter 4)
4. What does a more realistic model of interrogational torture (RIT)—but one still faithful to the pragmatic model’s core principles—look like? (Chapter 5)
5. How do we solve the RIT model? (Chapter 6)
6. What are the equilibria and outcomes of the RIT model and what can we say generally about them? (Chapter 7)
7. What does each of the outcomes look like? What are its properties? What is an example of the outcome in the real world of interrogational torture? (Chapters 8 to 11)
8. How do the RIT model outcomes compare to the pragmatic model’s benchmarks in Chapter 3? (Chapter 12)
The first step in the argument, then, is to examine whether the EITs used by the CIA and similar techniques used by the U.S. military amount to torture and what we know about the effectiveness of interrogational torture in eliciting valuable information.