8

Surprise Torture

Their disbelief was very hard to bear.

—SHEILA CASSIDY, Audacity to Believe

We begin with the equilibrium we found together in Chapter 6, the valuable information, surprise torture equilibrium. This equilibrium occurs only under objective questioning. In this and the next four chapters, we proceed as follows:

1. Describe the equilibrium and some of its features.

2. Identify the substantive outcome(s) (that is, the real-world counterparts) associated with the equilibrium.

3. Narrate an empirical case corresponding to the outcome.

4. Then return to the model features and reflect on the case in light of the model components.

EQUILIBRIUM FEATURES

In this equilibrium, the Cooperative, Resistant, and Innocent Detainee types play (“information,” “no information,” “no information”), respectively; the Pragmatic and Sadistic Interrogators play (“torture,” “torture”), respectively.¹ The Cooperative Detainee’s belief q about the likelihood that the Interrogator is Pragmatic is greater than his threshold ; that is, he is confident enough that he faces a Pragmatic Interrogator who will not torture him if he reveals information. He also believes that the Interrogator believes he has revealed all his information (). The (Pragmatic) Interrogator, however, believes he has not revealed all of his information (hence ) and is also confident that a Detainee who does not reveal information is Resistant and not Innocent () and thus tortures after “no information.”

Figure 8.1 locates the surprise torture equilibrium in the parameter space, with the move “information” of the Cooperative Detainee represented by the dark-gray box. The region covered by dots is the space taken up by torture, either the surprise torture of the Cooperative Detainee or the torture of the Innocent and Resistant Detainees who reveal no information, or both. As Figure 8.1 makes clear, this equilibrium depends on the four thresholds , , , and .

Figure 8.1 Valuable Information, Surprise Torture Equilibrium

Information Revelation Threshold,

As increases (that is, the point at which the Cooperative Detainee is willing to divulge information), the information region shrinks. Conversely, as decreases, the region increases and we see more surprise torture. All things being equal, in other words, the more the Cooperative Detainee is willing to talk (i.e., the lower the threshold ), the more surprise torture we should see:

Implication 8.1 (Torture’s Logic and Surprise Torture). The more willing cooperative detainees are to provide information, the more likely surprise torture is.

This is perverse. The more a torture program works according to its own logic by driving down detainees’ thresholds for talking, the greater the volume of space supporting torture of those detainees. Moreover, since “no information” is met with “torture” in this equilibrium, a change in the threshold alone does not change the total amount of torture in this equilibrium. This is illustrated by the fact that the shading covers the entire space in front of in Figure 8.1. As the dark-gray surprise torture region increases when decreases, all that changes is that more of the region covered by torture includes the torture of a Cooperative Detainee who has provided at least some information.

Information Hiding Thresholds, and

The relative size of this equilibrium also depends upon and . As the Detainee’s version of the Interrogator’s full information threshold increases (moves to the right), it squeezes the equilibrium, making it smaller, whereas increases in the Interrogator’s version make it larger. As the former threshold goes up, as the Cooperative Detainee thinks the Interrogator thinks he has not divulged everything and will torture him, the region in which he decides not to provide information increases. There will be less surprise torture of a Cooperative Detainee because he does not trust the Interrogator enough not to torture him to reveal the supposedly hidden information.²

The equilibrium space gets larger as increases and stays the same because the Interrogator’s threshold, or bar, for not torturing has gotten higher. There is more surprise torture because there is more space occupied by a misunderstanding of the Interrogator’s belief that the Detainee has revealed all the information he has. Obviously as shrinks but grows, this region of misunderstanding grows and so does the relative size of the equilibrium. This makes sense: The greater the misunderstanding, the more surprise torture.

The Interrogator’s own full information threshold is likely to be higher than the Detainee’s. Indeed, we know from Proposition 7.2 that is less than one-half and from Proposition 7.3 that is greater than or equal to one-half. In fact, Observation 7.3 tells us that the two thresholds, the two beliefs, will be the same only in the special case when the Interrogator has understood perfectly the information’s value. All other cases open up the possibility for surprise torture.

As a result, a detainee will believe the interrogator believes he has revealed all he knows “before” or more quickly than the interrogator herself. Indeed, an interrogator’s job is to be skeptical, to try and get every last drop of valuable information out of the detainee. So there is likely to be some daylight between and , making room for f to fall in between and sustain this equilibrium. In other words:

Implication 8.2 (Surprise Torture is Likely). Surprise torture—torture of a detainee after he has provided information, even if it was all of it—is likely.

The implication for real-world torture is that the lower the quality or quantity (or both) of background information, the less interrogators know about a detainee, the more likely is surprise torture. There is something a bit perverse here as well. Presumably, intelligence agencies and their interrogators are most concerned about the information about which they know the least. It can be scarier not knowing what you don’t know than knowing you don’t know some specific threat.

Indeed, this was the implication of Defense Secretary Donald Rumsfeld’s (in)famous dismissal of a question about Iraq and weapons of mass destruction (WMDs) during a February 2002 press conference. Asked about whether there was “any evidence to indicate that Iraq has attempted to or is willing to supply terrorists with weapons of mass destruction,” Rumsfeld tried to brush off the question by saying the following:

Reports that say that something hasn’t happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don’t know we don’t know. And if one looks throughout the history of our country and other free countries, it is the latter category that tend to be the difficult ones. And so people who have the omniscience that they can say with high certainty that something has not happened or is not being tried, have capabilities that are … They can do things I can’t do (laughter).³

Rumsfeld was, understandably, not the first government security official to think this way and will not be the last. If torture is used to find the “unknown unknowns,” however, then surprise torture is more likely. Information will be misinterpreted, and detainees will not be believed even if they have given up important information, even everything they know. In other words:

Implication 8.3 (Information Importance Increases Surprise Torture). The more important the information (in terms of unknown unknowns), the more likely the use of “unnecessary” torture (from the proponents’ point of view) because the value of information is less well understood.

Innocent Detainee Recognition Threshold,

Finally, consider , the Interrogator’s belief that the Detainee is Innocent (and not Resistant) after having received “no information.” As this threshold increases (moves to the back of the cube), the region gets bigger. This too makes sense; for values of p above the threshold, the Interrogator does not want to torture, thinking she has the wrong man. The more hesitant she is to believe that “no information” means an Innocent Detainee rather than a Resistant one, the more likely she is to torture and so the space for the equilibrium increases to the back of the cube.

We know, however, from Proposition 7.2 that if the Pragmatic Interrogator believes that it is more likely that the Detainee she faces is Cooperative than both of the other two types, then is less than one-half and likely closer to zero. Consequently, the back boundary of the dark-gray surprise torture region is under one-half and likely closer to zero (the front of the parameter space). In other words, the equilibrium exists only if the Interrogator believes that the probability the Detainee is Innocent after “no information” is very small. This is indeed likely to be the case.

SUBSTANTIVE INTERPRETATION

Substantively, this formal result covers two outcomes in the real world in terms of the Cooperative Detainee. In the first, the Cooperative Detainee has revealed all his information in order to avoid torture. Nevertheless, he is tortured afterward either because the Interrogator does not understand that the information is valuable or because she believes that he is not telling all he knows (or a combination of the two). Even according to the pragmatic model of torture for intelligence only, this torture is unjustified because a Detainee who has revealed all of his valuable information should not be tortured.

In the second substantive outcome or interpretation of this equilibrium, the Cooperative Detainee has revealed some information, but not everything he knows, and is tortured afterward. There is still surprise here—the Cooperative Detainee thought the Interrogator believed he had revealed all his information and so wouldn’t be tortured—but it is not the same sort of surprise one would feel if one had truly revealed everything and continued to be tortured. According to the pragmatic model, this torture is justified because a Detainee with valuable information is holding some of it back, refusing to reveal it. (Keep in mind, however, that an Innocent Detainee is tortured in this equilibrium as well.)

Note that the two cases are necessarily observationally equivalent to the Interrogator. The fundamental and unsolvable epistemological problem of interrogations generally—with or without torture—is the inability of the Interrogator to know what is “inside” a Detainee’s mind. Interrogators can never know whether a Detainee, even a very cooperative and productive one, has revealed all he knows.

For a case in the real world to match the surprise torture outcome, it must meet the following conditions:

1. The detainee possessed information sought by the interrogator.

2. The detainee was subjected to objective and not leading questioning.

3. The detainee gave up some or all of her valuable information but was tortured afterward from ignorance and/or disbelief on the part of the interrogator.

REAL-WORLD CASE: SHEILA CASSIDY

The story of Sheila Cassidy, an English doctor tortured by Pinochet’s secret police in 1975, meets all three criteria.⁴

On December 4, 1971, tired of the “rat race” of British medical life, Doctor Cassidy set sail for Chile on a cargo ship. She studied Spanish, passed her medical exams, and eventually found work in an urban hospital. It was not long, however, before her growing religious inclinations moved her closer to the Catholic church and she decided to work in a shantytown clinic where she could provide more basic medical services to the poor. Her work and increasing commitment to the Catholic church (she decided to become a nun at one point) led her to befriend many priests and nuns, both Chilean and in the large missionary community.

On October 21, 1975, a Catholic priest asked her whether she was willing to treat a man who had been wounded in the leg by a bullet. The priest did not identify the man, but Cassidy knew it was one of the men from a leftist revolutionary party who were being hunted down by the Pinochet military dictatorship. The man faced certain torture and perhaps death at the hands of the Chilean secret police, the Dirección de Inteligencia Nacional or DINA, if he sought help at a hospital. Cassidy was apolitical, though she was increasingly aware of the misery and hardship imposed by Pinochet’s martial law, and says in her memoirs that she decided to help simply because she was a doctor and doctors treat the injured and ill.

Cassidy did the best she could with limited resources to care for the wounded fugitive, Nelson Gutierrez, in the convent where he had taken clandestine refuge. It was difficult to control the infection and she saw him several times over a period of days. Eventually, he was spirited away (ultimately finding asylum in the residence of the Papal Nuncio), and Cassidy went back to her work at the clinic.

On November 1, 1975, Cassidy was visiting the house of another priest, looking in on a sick nun. She heard a blood-curdling scream and ran downstairs to find Enriquetta, the housekeeper, face down in a pool of blood with a large wound in her back. No sooner had Cassidy reached Enriquetta than machine gun fire from the street began to rip up the house. Plainclothes DINA agents entered the house and demanded her name. When she replied “Sheila,” they said, “she’s the one we want,” and after searching the house they bundled her into a car, slapped her, tied a blindfold on her, and drove off.

It wasn’t long before the car pulled into the courtyard of a colonial-era villa, Villa Grimaldi, now infamous as a major DINA torture center. Cassidy was taken to a room with a desk, a chair, a bare metal bunk frame, and what she would later call “an electric box,” with wire leads snaking from it. In disbelief at her situation, she initially refused to comply with their orders to undress, but relented after they started ripping the clothes from her. They tied her fast to the lower bunk bed, directly to the metal frame, her weight supported by the metal mesh.

And then they started shocking her. This was the infamous parilla, the “grill.” She tried to scream but they had forced a rag into her mouth. With the pain coursing throughout her entire body, they began the interrogation.

“Where did you treat Gutierrez?”

“Who asked you to treat him?” (Cassidy 1977, p. 174).

These (objective) questions let Cassidy know that DINA did not know about the priests’ involvement in particular and the role of the church more generally and she resolved to keep it that way by lying to her torturers.⁵ Two thoughts helped her withstand the pain. The first was the conviction that British officials would soon rescue her. The second was the knowledge that the lives of her friends depended on her.

And so, in response to their questions—“Why did they ask you?” “Where does he work?” “Who owned the house?” “What does he do?” “How did they contact you?”—and in between bouts of painful shocks, she began to make up answers. Finally, though, she could not come up with a street name and thus offered to take them there. At this point, she had no plan other than to stop the pain.

They drove her around the area she had lied about but after a while discovered she’d been lying to them. Angry at having been played and the precious time they’d lost, they promised another round of the parilla and drove her back to Villa Grimaldi. Still, she had bought the fugitives and the priests precious time. Moreover, it so happens that the fictional house she described matched a house in the neighborhood they were searching. As a result, they actually thought they had found the house and mounted a raid, only to discover that they had been misled. This demonstrates dramatically how even objective questioning can, when combined with torture, lead to false information and wasted time.

Back at Villa Grimaldi, they put one of the electrical clips in her vagina and cranked up the current. The resulting pain was “appalling” and the questions came so fast she was unable to make up answers to them. In her words, “they broke me.” Even so, she told the truth only “little by little,” telling “them as little as [she] could” to minimize the number of people in danger (Cassidy 1977, p. 188).

The trouble for Cassidy was that her tormenters refused to believe her. They refused to believe that nuns and priests could be mixed up with Marxist revolutionaries. This “disbelief was very hard to bear” because she “received many gratuitous shocks” for another hour or more and “there seemed no escape from the white hot sea of pain” (Cassidy 1977, p. 189).

Eventually the DINA torturers believed Cassidy and they searched the convent. They found neither the priests nor their main quarries, Gutierrez and revolutionary leader Andres Pascal Allende, and about this they were not happy. She was returned to Villa Grimaldi but this time she was interrogated by two senior officers in an office. They did not torture her and asked about the whereabouts of Gutierrez and Allende. She told them that Gutierrez had received asylum in the Papal Nuncio and that she didn’t know where Allende had fled.

They refused to believe she didn’t know because they assumed she was working with the MIR. They refused to believe her protestation that she only treated Gutierrez because it was her medical duty. And yet, when she told one interrogator that she would even have treated him if he were wounded, he said he believed her.

That did not stop him, however, from sending her to the parilla for a third time.

The pain was worse. They went back to the events of the day to see if they could find any leads. Eventually they hit upon a question that led Cassidy to say that a priest was going to try and find asylum for Allende. In her words, this was “the last information that I had” (Cassidy 1977, p. 192).

While DINA officers went out to find the priest, she was taken before senior officers again. Again suspicious that a church group could be involved, they hit her in the face and threatened to keep torturing her on the parilla, saying she would eventually give in. She was even tied to the bed again, but was not shocked any more.

Two months after her arrest, signed statements, and detention in various centers, she was released and returned to England.

BACK TO THE MODEL

Here, then, is an illustration of the surprise torture equilibrium. In the model, the Interrogator continues to torture after “information” either because she doesn’t understand the information is valuable or because she believes the Detainee is holding more information back. The former belief is the u parameter or variable, while the latter belief is the f variable. Translating Cassidy’s situation back into the language of the model, the value of u for the DINA interrogators was low; in their religious world view, good Catholic nuns and priests would never associate with godless Marxist revolutionaries and so they discounted her information.⁶

Remember that u is between zero and one, so a low value, a value approaching zero, reduces the value of the denominator. As the denominator gets smaller, the value of the whole threshold f gets bigger. This pushes the threshold to the top of the axis, so that the space occupied by torture is greater. More values of f fall under the threshold and thus support torture. The DINA’s estimation of u pushed the threshold up, moving it past the maximum of Cassidy’s belief , and they tortured her despite the fact that she gave them the information they wanted.

Cassidy’s case is like many others under Pinochet. Like French paratroopers in Algiers, DINA wasted no time with other techniques and immediately began torturing. As has been widely documented, torture under Pinochet was widespread, with concentration camps and torture centers set up all over Chile for anyone suspected of anti-regime activity (Muñoz 2008, p. 47; also Ensalaco 2011). The DINA intelligence gathering system relied on torture; as a result, reputation costs r for failing to torture a suspect who did not immediately give up information must have been high.

Even if r were not that high, however, it is clear from the huge apparatus of detention, torture, and summary execution to Pinochet’s own bland dismissals that the costs of using torture c and a were very low. This is perhaps nowhere better illustrated than by DINA’s reaction to the Italian government’s policy of granting asylum to Chileans who fled to the Italian embassy. After DINA tortured an opposition leader by the name of Lumi Videla to death, they threw her abused body over the wall of the Italian embassy in Santiago (Ensalaco 2011, p. 78).

Given the ferocity of the regime’s desire to root out any opposition, as well as the fact that they intended to execute many of those who did not die under torture, it is not surprising that the information hiding threshold was very high. Moreover, given the sheer number of victims swept up by DINA and other Chilean torture units, it would be surprising if they always interpreted the information they received correctly—that is, that u approached 1.

There was also another factor at work here, however. Her torturers’ inability to accept her information as valuable stemmed not from a lack of background information or context in which to place it, but rather from a world view that simply couldn’t square priests and nuns acting in support of leftists. While the Chilean clergy did not directly and openly oppose the regime, many did become active in human rights and became a thorn in the regime’s side. It should not have been such a surprise to DINA, but their own ideology blinded them to the possibility. It is easy to imagine how this conflict between world view and the facts in front of an interrogator might generalize to other contexts and conflicts.

As for Cassidy, she clearly assumed they would believe her when she told the truth. She also appears to have believed that her interrogators were pragmatic, that they would stop torturing once she gave them the information they sought. This may partly be due to the fact that she was not a professional operative or even active in the opposition; she just treated one man. Others may have been less confident that the DINA interrogators were Pragmatic and not Sadistic. It may also be the case that the threshold q dropped over time. She says she resisted initially to give others as much time to get away and only gave up information once she thought enough time had passed that they were more likely to be safe. In other words, the value of her information (to her), v, was very valuable at first, pushing her threshold up, but fell over time, causing the threshold at which point she would tell the truth to drop as well.

The argument thus far has shown that:

1. EITs are torture and the effectiveness of interrogational torture is an open question. (Chapter 2)

2. The Bush program approximates closely the ideal model of interrogational torture and includes limits on torture; the Bush and ideal models provide benchmarks for comparison with the game theory models to come. (Chapter 3)

3. The Bush model generates strange, quixotic outcomes. (Chapter 4)

4. The Bush interrogational torture program is more realistically modeled as objective and leading question variants of an incomplete information game, with three types of detainees, two types of interrogators, and uncertainty about the amount and value of information provided. (Chapter 5)

5. By positing a set of Detainee strategies, calculating the Interrogator’s expected utility using Bayes’ Rule to identify her best response, and checking for incentives to deviate by any of the Detainee types, it is possible to derive a perfect Bayesian equilibrium in which a Detainee is tortured after providing information. (Chapter 6)

6. The RIT model generates nine perfect Bayesian equilibria, the formal and empirical characteristics of which generate important observations, propositions, and implications, including:

(a) The Interrogator’s thresholds for believing that a Detainee is Innocent after “no information” are less than one-half, with close to one-half and closer to zero.

(b) The Interrogator’s information hiding threshold under objective questioning is greater than or equal to one-half, whereas her information hiding threshold under leading questioning as well as the Detainee’s version of are a little less than one-half.

(c) Objective questioning (potentially) provides better information, but is necessarily accompanied by more torture, than leading questioning, which, however, provides less valuable information.

(d) All things being equal, interrogators are more likely to get less valuable information than highly valuable information. (Chapter 7)

7. Surprise torture of a Cooperative Detainee—even if he has provided all his information—is not only likely, but perversely more likely

(a) the more willing the Detainee is to divulge information and

(b) the more important the information is in terms of “unknown unknowns.” (Chapter 8)

The next step in the argument is to examine and interpret the ambiguous information, selective torture and false confirmation, selective torture equilibria.