Unbalanced Policy Priorities and the Interrogation of Terror Suspects

Unbalanced Policy Priorities and the Interrogation of Terror Suspects Abstract Why do states shift from opposition to torture to its employment in the face of widespread abhorrence and claims that it generates unreliable information? The expected value of acting upon unreliable information depends on the state’s appraisal of the type of error it might be committing. States may value avoiding Type I errors, which result from acting upon incorrect intelligence, differently than avoiding Type II errors, which result from failing to act on an undetected threat. Interrogations yield a mix of truth and lies; operations employing this information will avoid making Type II errors but will also make Type I errors. A state’s preference over error type changes with circumstances (as we show, occurs in Turkey and the United States), leading states to adopt coercive interrogation techniques despite secular abhorrence of torture. Our results suggest anti-torture advocacy might be more effective targeting preferences over error types than reinforcing anti-torture moral imperatives. A majority of states, many democratic, torture prisoners (Conrad et al. 2011). Why do nations—especially those motivated by liberal norms—practice torture in the face of broad and intense condemnation (Sussman 2005)? Factors such as retribution (Liberman 2013), intimidation, and false confessions might encourage states to employ torture (Rejali 2009). Torture, however, is most often justified as an information tool (O’Mara 2009), with many stating that intelligence efficacy alone provides a liberal justification for its practice. The “only one rationale for torture that might conceivably be acceptable to a liberal: torture as a technique of intelligence gathering from captives who will not talk” (Luban 2007, 1436). In an increasingly networked world, the value of this information has grown: “The global nature of transnational terrorism necessitates that in addition to reacting to directed threats from terrorism, states harness information about the ties of origins’ nationals in attacks worldwide” (Avdan 2013, 6). Thus, a critical answer to the why-do-liberal-states-torture question seems to be the claim that sometimes valuable, timely information necessary for their security can only be obtained using coercive methods. The question, however, is even more complex as states’ policies toward torture vary over time. What conditions shift those opposed to torture to its support? Focusing on policy change requires that we examine political factors that vary (Avdan 2012; Blanken 2012). These political factors can be external or internal to the state (Bayer and Onis 2010). Most literature in this area is structural and thus cannot explain change in torture policy (Conrad and Moore 2010). For example, what altered to cause the notable increase in the global use of coercive practices such as waterboarding following the attacks of September 11, 2001? Some might see this question as trivial—obviously, the deaths of almost 3,000 people on 9/11 generated tremendous feelings of anger, hate, and fear that led many to desire revenge. However, changing counterterrorism policy is difficult (Blanken and Overbaugh 2012), and after 9/11, a majority of the American public did not support the use of torture (Gronke et al. 2010). Thus, it seems unlikely that revenge alone can explain coercive interrogation policies post 9/11 in the United States, let alone in nations that were not directly affected by the attacks (Conrad and Moore 2010). We develop a simple game theoretic model that links policy preferences to the decision to torture and, in so doing, illustrates informational motivations for changing torture policy. As such, our work embraces the viewpoint that it is important to understand the possible rational underpinnings of a state’s decision to torture (Dragu 2011; Schiemann 2012). Our intent is not to justify as rationally desirable an obviously abhorrent practice, but instead to gain a deeper and more nuanced understanding of the informational conditions under which states—even those that abhor torture—will resort to such extreme measures (Wantchekon and Healy 1999, 597). What differentiates us from previous game theoretic treatments of torture is our focus on error types. Errors can be either Type I (mistakes of commission) or Type II (mistakes of omission). For example, a Type I error would include executing someone for a crime they were later found not to have committed. A Type II error, in contrast, would be failing to arrest someone for a crime they had committed. In the first case, the state commits an error of action, and in the second case, the error is that the state fails to act. Although institutions and individuals always try to avoid mistakes, states have preferences over the type of error they want to avoid most, and these preferences fundamentally affect their calculations about the usefulness of information gained through torture. In particular, rational leaders who are strongly biased against Type II errors will still consider “dirtying their hands” with torture to get what they perceive may be vital information necessary to avoid an error of omission (Walzer 1973). Our goal is to use states’ desire to avoid Type II errors as a conceptual tool for developing a better understanding of how unreliable information must be before states decide to forgo torture in the face of potential security threats, avoiding its reputational and ethical costs. We begin by examining the role of torture in intelligence gathering, error types, and how the two interact. We then present a model that analyzes decisions to use torture in view of the quality of the information it likely reveals. We illustrate the game with brief vignettes of torture policy changes in the United States and Turkey and conclude with thoughts on future research and policy. Torture A common definition of torture draws on the United Nations “Convention against Torture and Other Cruel, Inhuman Degrading Treatment or Punishment” (Danelius 2008, 1)1: any act by which severe pain or suffering, whether physical or mental, is intentionally inflicted on a person for such purposes as obtaining from him or a third person information or a confession…when such pain or suffering is inflicted by or at the instigation of or with the consent or acquiescence of a public official or other person acting in an official capacity. (United Nations General Assembly 1984) There are two main reasons provided supporting the prohibition of torture: ethics and efficacy. Ethics States have long viewed torture as inhumane and cruel and banned its practice. The Geneva Conventions (signed in 1949 and ratified by virtually all countries) ban the torture of prisoners of war (Yingling and Ginnane 1952) and promote the “humane treatment and interrogation of prisoners” (Kelly et al. 2013). International prohibitions of torture were expanded by the Torture Convention cited above.2 The most basic normative argument against torture is that it vitiates “the inherent dignity of the human person” by violating peoples’ “inalienable rights” (United Nations General Assembly 1984). Torture does not just reflect an extreme version of other types of punishments but represents a categorical, “special” bad (Sussman 2005). Shue (1978, 124) argues that “no other practice except slavery is so universally and unanimously condemned in law and human convention.” International law places the condemnation of torture among jus cogens or “peremptory norms,” a prohibition that cannot be overturned by treaty or law and is binding regardless of whether or not torture-specific prohibitions are adopted (Foley 2003, 4–5). Torture is seen as irreconcilable with modern democratic values because “[t]orture aims to strip away from its victims all the qualities of human dignity that liberalism prizes” (Luban 2007, 1430). But, democracies do nasty things (covert actions, colonialism, etc.) when they feel like they are unobserved by their constituents (Bueno de Mesquita et al. 2003), suggesting anti-torture normative constraints are part of the exogenous “strategic context” within which rational choice actors maximize utility. Effectiveness Effectively obtaining information represents one of the most commonly identified objectives of torture. “As a method for extracting information, torture currently is quite widely used” (Wantchekon and Healy 1999, 597). While some believe that torture’s effectiveness at procuring information is irrelevant to its prohibition (“a democracy that legalizes the use of torture in its desperation to gain information loses something more important—the trust of its people, the foundation of a democracy” Rejali 2004), others clearly disagree and see torture, even if intrinsically bad, as a vital intelligence tool (Shue 1978; Dershowitz 2003). As Schiemann (2012, 3) states, “determining whether interrogational torture works—defined as providing clear and valuable information to the state—is essential. First, if it does not work, then the main rationale for its use—it is the only method we have left when detainees refuse to talk—disappears, and any further use is pure sadism.” Armshaw (2011, 139) agrees, writing that, “Ethically, torture could only be justified as a means of generating information that could not be found (usefully) in any other fashion.” Even Rejali (2009, 23) states that “The heart of the matter is whether an organization can apply these techniques…scientifically and professionally to generate true and reliable intelligence.” So does torture work? The widespread answer is no. Intelligence agency members and interrogators frequently point to the highly inaccurate nature of the information obtained through torture (Morgan 2000; Finn 2008; Rose 2008; Alexander 2009; O’Mara 2009; Soufan 2009a; Schiemann 2012). For example, interrogation practitioners such as those of the US Army state that “the use of torture and other illegal methods is a poor technique that yields unreliable results, may damage subsequent collection efforts, and can induce the source to say whatever he thinks the interrogator wants to hear” (US Army 1992, 1–8). This anti-torture view is not just policy but reflects actual Army and other organizations’ evaluations (Irvine 2005). For example, recently released Central Intelligence Agency (CIA) documents “fail to show that [enhanced interrogation] techniques stopped even a single imminent threat of terrorism” (Soufan 2009b). physical and/or psychological duress are viewed as an alternative to the more time-consuming conventional interrogation process. The error inherent in this line of thinking is the assumption that, through torture, the interrogator can extract reliable and accurate intelligence. History and a consideration of human behavior would appear to refute this assumption. (US Joint Personnel Recovery Agency 2002, 1) Reflecting these views, US military personnel oppose torture at higher rates than American civilians—especially military personnel who served in Iraq (Gronke et al. 2010, 438, 441). But not everyone agrees. Some believe torture should play a role in intelligence (Hanson 2010). Others maintain that torture is vital for obtaining critical national security information. For example, former CIA Director Hayden and Attorney General Mukasey (2009) argue that “coercive interrogations” generate “intelligence” and add that it is vital to corroborate all intelligence information. Bagaric and Clarke (2005, 588) state, “The main benefit of torture is that it is an excellent means of gathering information.” Republican Senate leader McConnell advocated dropping legal restrictions on interrogating a suspect in the Benghazi attack saying, “The most valuable thing we can get from this terrorist is information” (Baldor and Benac 2014). Some make exceptions to torture prohibitions in the face of potentially deadly and time-dependent threats in what is widely known as the “ticking bomb” scenario (Silver 2004). In its classic form, Walzer states that a leader is asked to authorize the torture of a captured rebel leader who knows or probably knows the location of a number of bombs hidden in apartment buildings around the city, set to go off within the next twenty four hours. He orders the man tortured, convinced that he must do so for the sake of the people who otherwise die in the explosions—even though he believes that torture is wrong, indeed abominable, not just sometimes, but always. (1973, 167) In this scenario, the ability to obtain information that prevents the imminent loss of life creates allowable circumstance for using torture—even if it is recognized as morally wrong (Dershowitz 2003). Conditions can justify “dirty hands” without legitimizing torture (Walzer 1973): “it cannot be denied that there are imaginable cases in which the harm that could be prevented by a rare instance of pure interrogational torture would be so enormous as to outweigh the cruelty of the torture itself” (Shue 1978, 141). Thus, the critical factor is the utility of the information. Types of Error An operation to rescue Americans held captive by ISIS failed when US Special Operations Forces arrived at a base shortly after the prisoners had already been moved (Mulrine 2014). Among the prisoners relocated was journalist James Foley, who was later beheaded in an act that lead to a significant increase in the US military presence in Iraq (Harnden 2014). Wartime intelligence mistakes like these are unavoidable. “Intelligence failures are not only inevitable, they are natural” (Betts 1978, 88). Although errors are inevitable, the type of mistake committed can vary (Kam 1988; Bar-Joseph and Kruglanski 2004; Bar-Joseph and Rose McDermott 2008). A key question is what type of error is more acceptable—Type I (commission) or Type II (omission) (Pollard and Richardson 1987; Sobel and Leeson 2006; Zhao and Olivera 2006)?3 Decision makers have to choose between the “cry wolf” problem and “miscalculated escalation” (Wirtz 1989, 834). By understanding the error type that actors try most to avoid, we can make predictions about their likely behavior (Kam 1987; Babad and O’driscoll 1992; Mercer 2005). This approach has been successful in predicting decisions on foreign policy (Lampel and Shapira 2001), conflict resolution (Avruch 2003), legal systems (Persson and Siven 2007), sniper rules (Roberts 2013), and drug approval (Intriligator 1996; Carpenter 2004; Sobel and Leeson 2006). Gartner (2013) points out that an effort to avoid one type of error may inadvertently promote the other type. In the case of information obtained under torture, it is the reliability of the information that can cause mistakes. Trying to avoid a Type II error by torturing to obtain new information does not promote Type I errors. But the mix of truth and lies that a tortured prisoner may disclose can lead a state that acts upon this information to risk Type I errors as it tries to avoid making Type II mistakes. The Game We focus on rational incentives for the State based on its evaluation of error types and for the Prisoner based on their perception of disclosure as an “extension of their jihad” (Weiser 2014). We develop a simple game model designed to capture the State’s tolerance for faulty intelligence, comparing situations in which avoiding errors of omission take precedence over making errors of commission to situations in which such policy priorities do not hold true. The model allows us to evaluate the link between unbalanced priorities over error types (disproportionately avoiding Type II errors) and the decision to use torture to obtain possibly poor quality intelligence. The State in our model must decide to use torture to obtain noisy information that it must subsequently act upon or not. Interrogated terror suspects, if they talk, are assumed to seek to incite the State to act upon the information, as overreaction facilitates recruitment (Weiser 2014). However, the State cannot ignore the possibility that it is facing a prisoner who will never talk. Cases of resistance under torture have been documented (Cardina 2013). Such prisoners will not inform regardless of whether they have valuable information to disclose or not. If the Prisoner remains silent, it is the result of character and inner resolve in the face of extreme duress. The State, therefore, makes decisions regarding the resort to torture with uncertainty about the type of Prisoner it is facing and has the following initial beliefs: with probability β, the Prisoner will break down under torture and talk; with probability (1 – β), the Prisoner is resilient and will never talk. If under torture the Prisoner talks, he discloses a mix of true and false statements, but the State cannot reliably sort out the truth from the lies. There could be many reasons for the State’s inability to validate perfectly all information. For example, in the classic “ticking bomb” scenario, interrogators have insufficient time and resources to validate all the information they extract: “the less time the authorities have to make use of the information gleaned from coercion, the less incentive they have to investigate its veracity, and the less incentive the captive has to tell the truth” (Armshaw 2011, 138; see also Koppl 2006). But it could also be the case that the information received is by its very nature hard to verify ex ante. As Mulrine (2014) points out “good US military intelligence on…militant groups is in short supply.” For example, a prisoner could disclose a plan to attack a particular site or to change a known headquarters by a particular date. In both cases, the information may be verifiable ex post, after the fact, when it is too late to protect the site or storm the group’s headquarters. The attempt to rescue Foley exemplifies this dynamic as US Special Operations Forces found a “dry hole” where American hostages were expected to be held (Harnden 2014). As a result, the State can anticipate that proportion x∈[0.1] of the Prisoner’s disclosures are false and (1-x) are true. If the tortured Prisoner talks, the State must decide whether to act upon the information received with uncertainty about its reliability. Under what conditions would a State still decide to use torture despite the cost in terms of reputation and international standing? Reputation costs incurred as a result of a state’s use of torture are varied and multi-tiered. Costs are incurred with the public acknowledgment of the state’s practices, such as domestic audience costs that compromise a democratic leader’s chances of re-election. For example, the disclosure of abuse of prisoners at Guantanamo Bay was met with immediate public outrage with attendant damage to the Bush Administration’s credibility and trustworthiness (Malinowski 2008). A state that practices torture also risks losing moral authority on the world stage. Forsythe (2006) likens the risks associated to the Bush Administration’s use of enhanced interrogation techniques to those taken by France during the Algerian war of independence. France, by engaging in such practices, lost “self-respect and reputation in the world” and “generated increased opposition at home and abroad” (Forsythe 2006, 465). Although increased opposition at home in the United States was slow to develop in the aftermath of the 9/11 attacks, “Bush policy towards enemy detainees contributed to a highly negative reaction in the Arab-Islamic world and amongst virtually all European publics” (Forsythe 2006, 475). Thus, states face a combination of domestic and international costs (Gartner and Regan 1996). In the words of Forsythe (2006, 491), “the United States gravely damaged its reputation as a champion of human rights, humanitarian norms and the rule of law.” States that do not have the stature of the United States on the world stage may incur lower reputation costs of this sort, but audience costs at home and the potential for internal dissent and destabilization remain core ingredients of the reputational costs associated to a state’s use of torture. The State’s decision to torture depends on the expected value of the information it hopes to receive. The game model is pictured in Figure 1. The State’s payoffs are in black, whereas the Prisoner’s payoffs are recorded in dark gray. Figure 1. View largeDownload slide The game model Figure 1. View largeDownload slide The game model The State makes an initial decision at nodes S1 and S2 on whether to employ torture with uncertainty about the Prisoner’s type. If the State decides against the use of torture, its payoff is E, which we interpret as the value of the information the State has currently obtained given conventional interrogation techniques. The Prisoner’s initial payoff is normalized to 0. If the State decides to torture and the Prisoner remains silent, nothing is gained and the State incurs a cost c for having resorted to these means. Its payoff is then E-c, while the Prisoner’s payoff of  -y is assumed greater than any payoff he could get by talking and having the State subsequently consider the information. If the Prisoner is not resilient and talks, he makes a decision from node P to disclose information that is part truth part lies with a proportion x of lies. The State must then make a decision at S3 to use the information obtained under torture or not. If the State does use the information received under torture, it must expect that proportion x is misleading and will lead to a Type I error. For example, the State might decide to preemptively guard against an attack that will not happen wasting resources that could have been used elsewhere. Or it could destroy an alleged terror cell safe haven killing innocent parties instead. In that case, the State’s payoff is E1-d-c as the value of the information it does have is diminished by the inaccuracies introduced by the false disclosures of the Prisoner. But proportion (1-x) of disclosures is reliable and use of these new facts increases the State’s effectiveness. In that case, a Type II error is avoided and the State’s payoff is E1+t-c. For example, the information can lead to the arrest of a high-level operative or the destruction of a cache of arms held by a terror organization. In sum, use of the disclosures made by the Prisoner under torture yields an expected payoff to the State of   x[E(1−d)−c]+(1−x)[E(1+t)−c].The Prisoner receives  -z1 if the State uses the disclosures. If the State chooses not to use the information obtained under torture, its payoff is E-c as nothing has been gained by torturing the Prisoner but a reputation cost has been incurred. The Prisoner’s payoff in that case is  -z2. What drives Prisoner and State choices? First, the Prisoner makes a strategic decision only if he does not remain silent. In this case, he must decide on the mix between truth and lies in his disclosures. We assume that the overarching motivation for the Prisoner is to see the State conduct an excessive, imbalanced response (Gartner and Regan 1996). A violent, disproportionate State reaction gives visibility to the terror group and enhances its appeal with potential recruits. We, therefore, assume that  -z1>-z2. This preference for action implies that enough truth be embedded among the lies or that x be small enough. More precisely, the Prisoner will want to choose x so that   x[E(1−d)−c]+(1−x)[E(1+t)−c]>E−c.This means that   x<tt+d or equivalently (1−x)>dt+d. (1) Note that we do not assume that the Prisoner’s payoffs depend on the magnitude of x; instead the Prisoner’s incentive, if he cannot keep silent, is to mobilize the State to act against the terror group as this furthers the cause (Adler 2010). Although there may be a preference for lies over truth, truth telling in and of itself may actually be perceived as more likely to lead to visible action against the terror organization than lies. Moreover, some terror suspects may take pride in boasting about their organization’s plans and achievements (Weiser 2014). By choosing x to satisfy (1) the Prisoner’s priorities are satisfied, however high or low x might be. This modeling choice has an added advantage in that we avoid requiring that a tortured Prisoner be assumed to pick a value for x to precisely maximize an objective function. Rather the choice is to target an ordinal preference. The lower bound on the reliability of the disclosures obtained under torture dt+d depends on the valuation of error types. It is immediately apparent from (1) that a State that values avoiding an error of omission (high t) far more than it does making an error of commission (low d) will also tolerate using information whose reliability is low. From the Prisoner’s point of view, inciting a Type II error-phobic State to action can be achieved while adding a high proportion of lies or misleading information to the mix of disclosures. How unreliable can the Prisoner’s disclosures be before the State decides to forgo the use of torture altogether? The State will only resort to torture if it anticipates that it will make use of the information received by these means. This is because torturing without using the information obtained yields a payoff of E-c<E as reputation cost c>0. The State that tortures must expect a low enough proportion of prisoners to be resilient and a high enough proportion of disclosures by prisoners who talk to be reliable. Torture becomes a rational choice if   (1−β)[E−c]+β[x(E(1−d)−c)+(1−x)[E(1+t)−c)]>E.This leads to the following inequality:   c<Eβ[t−x(t+d)]. (2)As the Prisoner’s priorities lead him to choose x<tt+d relationship, (2) can hold for small enough values of the strictly positive cost c. Stated differently, if reputation costs are small enough, the following behaviors form a subgame-perfect equilibrium (SPE) of the game pictured in Figure 1. The Prisoner who talks, discloses a mix of truth and lies where lies comprise proportion x<tt+d of the information communicated to the State. Anticipating the Prisoner’s disclosure under torture, the State knows that it will use the information obtained by these means for sure. It will, therefore, decide to resort to enhanced interrogation techniques with probability 1.Having established the existence of an SPE we ask: How reliable must information received via torture be to justify its use? And how is reliability linked to States’ valuations of Type I and II errors? In order to examine the consequences of relationship (2), we must first make assumptions about the proportion of prisoners who will not talk under torture. A State considering the use of torture must assume that some prisoners will break down under duress and that a sufficient proportion of valuable information will be revealed by these individuals. One would expect the State to expect that few prisoners will remain silent. However, whether a prisoner will talk or remain silent when tortured is a matter of debate. Rejali (2007) claims that the idea that “everyone talks sooner or later under torture” is a myth. We will, therefore, examine two cases: in one the State assumes that at most 20 percent of prisoners will be able to resist and remain silent, and in the other, the State believes that as many as 80 percent of prisoners will resist and remain silent under torture. Consider the first case. If 20 percent of prisoners remain silent, then β=0.8 and relationship (2) can be rewritten as   c<E * 0.8[t−x(t+d)], (3)where x is the proportion of prisoner disclosures that are unreliable. If the State believes that 80 percent of prisoners will remain silent, then inequality (3) reads   c<E * 0.2[t−x(t+d)]. We explore the relationship between reliability, valuation of error types, and reputation cost both numerically and graphically. In order to visualize the impact of parameter shifts on the decision to use torture, we identify the threshold x for which relationship (3) is binding, assuming first that 80 percent and then 20 percent of prisoners are expected to talk. Table 1 provides threshold values for x for various values of c and t, holding d constant at d=0.1. These are the values obtained when relationship (3) is binding. Parameter E is also held constant at E=2.5. Table 1. Threshold values for proportion x of unreliable information Case 180% of prisoners talk  t=0.2  t=0.3  t=0.7  t=1.5  c=0.3  16.7%  37.5%  68.8%  84.3%  c=1.5  The State will not torture  46.9%    Case 220% of prisoners talk  t=0.2  t=0.3  t=0.7  t=1.5    c=0.3  The State will not torture  12.5  56.3  c=1.5  The State will not torture  Case 180% of prisoners talk  t=0.2  t=0.3  t=0.7  t=1.5  c=0.3  16.7%  37.5%  68.8%  84.3%  c=1.5  The State will not torture  46.9%    Case 220% of prisoners talk  t=0.2  t=0.3  t=0.7  t=1.5    c=0.3  The State will not torture  12.5  56.3  c=1.5  The State will not torture  View Large Consider the two cells shaded in dark gray on the right side of Table 1. When the value of avoiding a Type II error t is large ( t=1.5), torture can be a rational choice even if the prisoner is unlikely to talk. If 80 percent of prisoners talk and cost c=0.3, the use of torture is rational even if as much as 84.3 percent of prisoner disclosures are misleading. If a much lower 20 percent of prisoners are assumed to talk, then torture remains a rational choice if the few prisoners that talk reveal useful information about half of the time. Of course, information reliability must be higher if fewer are expected to talk, but the rationality of the resort to torture when few prisoners are expected to talk at all comes as a surprise. As expected, if it becomes more costly to violate international torture norms, the reliability of information obtained by these means must increase considerably even if the value attributed to avoiding a Type II error is high. When only 20 percent of prisoners are expected to talk, then an increase of costs to c=1.5 rules out the use of torture under any of the parameter values examined for  t. At the other extreme, consider the cell highlighted in dark gray in the top left side of Table 1. For this cell, 80 percent of prisoners are assumed to talk under torture; the cost of violating the torture norm is set at c=0.3, but the benefit associated to avoiding a Type II error has dropped to 0.2. Now, given d=0.1, torture can only be justified if there is at least an 83.3 percent chance that information obtained via torture is valuable (it is unreliable at most 16.7 percent of the time). Clearly, increases in the reputational cost of violating the torture norm and reduced benefits to avoiding Type II errors relative to the cost of making a Type I error favor ruling out torture as a means to gather noisy intelligence information. In Figure 2, we provide a visual representation of the circumstances in which a state might resort to torture when seeking new information. In Figure 2, we fix (1-x) or the proportion of reliable information that can be expected from the disclosures of a tortured prisoner and identify the areas for c and t that satisfy cost constraint (3) assuming that 80 percent of prisoners are expected to talk under torture and given values E=2.5, and d=0.1. The figure limits t to 1.5. Figure 2. View largeDownload slide Information reliability and torture Figure 2. View largeDownload slide Information reliability and torture Figure 2 is constructed as follows: we consider various values for x, the proportion of prisoner disclosures that are false or misleading. For each value of x, we obtain a relationship between c and t for which inequality (3) is binding. These are the upward sloping lines labeled according to the value of (1-x), the proportion of disclosures that are truthful. For example, if 1-x=25 percent, meaning that the State anticipates that 25 percent of the information obtained under torture is reliable, (3) is binding when    c=0.8 * 2.5 * [t−x(t+0.1)] or c=−0.15+0.5t.This relationship between c and t is represented in Figure 2 by the bolded upward sloping line. In the shaded area below this line are all combinations of c and t for which the State finds it in its interest to use torture. For combinations of c and t above the bolded line the state will not resort to torture given the 25 percent anticipated reliability of information. If the anticipated validity of the information obtained through torture decreases to, say, 10 percent, then c must decrease and t must increase for torture to be a rational decision. The combinations of c and t for which the State tortures in this case is shaded in dark gray for values of t that remain below 1.5. Our game theoretic analysis suggests that rational states strongly biased against Type II errors will consider torture to extract information they know to be highly unreliable even if they believe that a high proportion of prisoners may not talk. As Senator Feinstein, Chair of the Senate Select Committee on Intelligence, wrote in the “Committee Study of the Central Intellgence Agency’s Detention and Interrogation Program,” “…prior to the attacks of September 2001, the CIA itself determined from its own experience with coercive interrogatons, that such techniques ‘do not produce intellegence,’ ‘will probabably result in false answers,’ and had historically proven to be ineffective. Yet these conclusions were ignored” (Feinstein 2014, 3). We unpack the calculation whereby torture is simultaneously recognized as generating unreliable information and yet still employed. Our results are not US specific. For example, we think this calculus may help to explain the hardening of interrogation practices by other democratic states following the trauma of the 9/11 attacks. We elaborate on these dynamics in the vignettes that follow looking at both the United States and Turkey. Vignettes We provide two vignettes to show how the model helps to explain actual situations. These brief descriptions are not full case studies but rather suggestive illustrations of the link between policy decisions and priorities over error types. Vignettes are especially useful with game theoretic models as they provide illustrations of different types of choice calculations. Lacking data on individual torture cases, the vignettes focus on the policy implications of the model and not its specific dynamics. The goal is not to match each aspect of the model to a historical event, but rather to capture the general flavor of the policy trade-offs we examine. United States Since 9/11 The 9/11 attacks altered US leaders’ preferences over error types and torture. 9/11—Bush 9/11 forced the United States to confront “a new kind of threat” (former Homeland Security Secretary Tom Ridge 2009) that fundamentally altered leaders’ and organizations’ views. As a result of 9/11, priorities changed. As Vice President Cheney stated, “I’ll freely admit that watching a coordinated, devastating attack on our country from an underground bunker at the White House can affect how you view your responsibilities” (Clarke 2009). The attacks altered attitudes toward types of error across American leadership and intelligence organizations. For example, in a “dramatic shift in focus of the Department of Justice (DOJ),” the FBI shifted from a emphasis on criminal activities such as catching bank robbers to a new Type II avoidant anti-terrorism perspective: “Since September 11, the Attorney General and the FBI Director have elevated counterterrorism and the prevention of future terrorist attacks against United States interests as the top priority of the DOJ and FBI” (US Department of Justice, Office of the Inspector General 2009, 11—emphasis added). This change represented “a huge paradigm shift within the DOJ from prosecution to prevention,” which we would argue maps directly to a shift from avoiding Type I to avoiding Type II errors (US Department of Justice, Office of the Inspector General 2009, 11). As former CIA Director Hayden stated, “I was in government for ten years after 9/11, and let me tell ya, a phrase I never heard from anybody in a position of authority: ‘Whatever you guys do about this terrorism threat, please, please don’t overreact’” (Kelly 2014). In other words, have no fear about committing Type I errors. Other organizations and leaders similarly shifted their perspective on intelligence errors. US National Security Advisor (and later Secretary of State) Rice stated, “Unless you were there, in a position of responsibility after September 11th, you cannot possibly imagine the dilemmas that you faced in trying to protect Americans…you were determined to do anything that you could that was legal to prevent that from happening again” (Kessler 2009). The National Intelligence Director on 9/11, Dennis Blair, echoed this perspective: “Those methods, read on a bright, sunny, safe day in April 2009, appear graphic and disturbing. But we will absolutely defend those who relied on these memos [supporting torture]” (Buckley 2009). Although the Bush Administration’s support for employing torture in interrogations following 9/11 has been widely described (CIA 2004; Department of Justice 2008; US Senate 2008), we think that an important aspect of the story is missing. Specifically, following 9/11, the Bush Administration became extremely wary of committing a Type II error—failing to take actions that might prevent additional attacks on the United States and American citizens—and thus became more willing to commit Type I errors (Suskind 2006; Gartner 2013). As President Bush’s initial appointee to lead the 9/11 inspired Homeland Security Department stated, “The criminal justice system is about prosecution and counter-terrorism is about prevention” (Ridge 2009). The goal of prevention is to avoid Type II errors—avoid failing to take actions that would have prevented an attack. Prevention requires acting before a crime is committed, thus opening up the door for the commission of Type I errors and torture: “On February 7, 2002, President Bush signed a memorandum stating that the Third Geneva Convention did not apply to the conflict with al Qaeda and concluding that Taliban detainees were not entitled to prisoner of war status or the legal protections afforded by the Third Geneva Convention” (US Senate xiii). The use of torture was national policy, implemented not by “a few bad apples” as was often repeated after Abu Ghraib, but by leaders fearing making Type II mistakes. “Interrogation policies endorsed by senior military and civilian officials authorizing the use of harsh interrogation techniques were a major cause of the abuse of detainees in US custody” (US Senate xxv). Despite recognizing the low quality of the information gained through its use, American officials, profoundly fearful of committing another Type II error, systematically endorsed the practice of using torture to obtain counterinsurgency information. Finally, it is important to recognize that, despite the national hurt inflicted by 9/11, the American public did not support torture. “Not once during the eight years of the Bush administration was there an American majority in favor of the use of torture” (Gronke et al. 2010, 441). Rather than public opinion, the motivation was an informational logic, driven by fear of committing a Type II error, which made the data error rate from torture acceptable to US leaders and the information it provided appear useful. 9/11—Obama Focusing on error type leads to a different way of considering the change in torture policy implemented by the Obama Administration. In 2009, the President issued an Executive Order requiring personnel to follow the US Army Manual on interrogations, which explicitly prohibits torture (Semel 2013). Although Obama Administration levels of coercive interrogation and actions such as the extraordinary rendition program still exceed that of pre-9/11 (especially at “black site” prisons), they clearly represent a change in US actions from the Bush Administration. The standard explanation for the change in policy is that the political costs of violating the norm became too large, so the administration changed policy. “Torture was only ended (if indeed it has ended) by the public outcry after the revelation of torture by U.S. soldiers, as well as the leaking of the Red Cross reports to the media, made the costs to elites too high to bear” (Armshaw 2011, 137). Our approach confirms that a shift in the perceived cost of violating torture norms will inhibit its use, but this inhibition will only be reinforced if the administration’s assessment of errors becomes less biased toward avoidance of Type II errors. As a result, the threshold for information reliability beyond which a state’s calculus favors torture dramatically increases. Figure 3 illustrates a possible Bush–Obama shift in t and c and the implications for thresholds on information reliability. We assume for both administrations that d=0.1, E=2.5, and they anticipate that 80 percent of prisoners will talk. Figure 3. View largeDownload slide Possible shifts in costs and benefits from Bush to Obama Figure 3. View largeDownload slide Possible shifts in costs and benefits from Bush to Obama Using (3), we can then identify the minimum level of reliability that each administration requires given its evaluation of the avoidance of a Type II error as well as the costs of torture. For example, given the assumed priorities of the Bush Administration, minimum information reliability must satisfy cBush=2.5*0.8*[tBush-xtBush+0.1] or 1-x=cBush+0.22tBush+0.2. Clearly, the constraint on information reliability increases under the Obama Administration, and this favors foregoing the use of torture in the interrogation of terror suspects. With reference to Figure 3, given anticipated reliability of information, the combination of costs and valuation of Type II error avoidance for the Bush Administration falls in the shaded area in which the state will torture, whereas those same costs and valuations for the Obama Administration fall outside of the torture zone. Assessments of the ineffectiveness of torture to produce timely and important results made it easier for the administration to oppose its use as far more reliable information would have been needed to justify the practice. In terms of the model, minimum required reliability CObama+0.22tObama+0.2 exceeds anticipated reliability of the information received under torture. As a result, one of the four “central themes” in the US Senate’s recent report on the CIA’s enhanced interrogation techniques following 9/11 is titled, “The CIA’s ‘enhanced entorrogation techniques’ were not effective and did not prodcue any ‘ticking time bomb’ information crucial to saving lives” (Kelly 2014). The War on Terror builds on the fear of committing an error of omission. Our analysis contributes to explaining the widespread alleged use of torture since 9/11. Turkey The 1980 military coup led to a rise in the use of torture by the Turkish government. The magnitude and severity of torture at the hands of the military government in the early 1980s has been referred to as “the period of barbarity” (Zeydanlıoğlu 2009, 7). Zeydanlıoğlu in his characterization is referring to the experiences of prisoners in Diyarbakır Military Prison No. 5 in Turkey’s Kurdish region in the early to mid-1980s. The return to a civilian government at the end of 1983 and the restoration of a multi-party political system did not end the dispute with the Turkish Kurdish population, a dispute that has long cultural, historical, and even linguistic roots (Fernandes 2012). Martial law remained in place in provinces in the South-East, where the Turkish government continued its use of torture in its efforts to suppress Kurdish secession attempts. The high incidence of reported torture cases continued into the first half of the 1990s. A 1998 European Union (EU) progress report noted that “persistent cases of torture, disappearances and extra-judicial executions are regularly recorded despite repeated official statements of the government’s commitment to ending such practices” (European Commission 1998). The use of torture began to decrease in 2002. As shown in the ITT database, Turkey’s use of torture decreased from the highest level “Systematic/level 5” from 1995 to 2002 to “Routinely/level 4” in 2003 and “Widespread/level 3” in 2004 (Conrad and Moore 2011). In 2001, the government began to make progress in protecting the population’s human rights and decreasing the practice of torture as observed by both the European Commission (2002) and Amnesty International (2001). A 2004 Human Rights Watch notes that “Turkish legal protections for detainees are better than in many EU member states.” In 2005, the Turkish government amended provisions in the constitution to improve the legal framework regulating torture, which criminalized torture and reduced the maximum length of detention from 90 days in 1980 to 24 hours (Human Rights Watch 2006). In 2012, the EU reported that “overall, there has been a downward trend in torture and ill-treatment in places of detention” (European Commission 2012; see also UN Manual on the Effective Investigation and Documentation of Torture—“The Istanbul Protocol”). What motivated this change? One factor that did not change is membership in anti-torture conventions and legal prohibitions against torture. Turkey is a signatory to the United Nations and European Conventions against Torture and Other Cruel, Inhuman, or Degrading Treatment, signed and ratified in 1988. The Turkish constitution explicitly forbids the use of torture (Article 17, 1982). In addition, torture occurred under both civilian and military rule (although not at equal levels). The changes that did occur were a reduction in the threat posed by the Kurdish terrorist group, the PKK (and the cost of Type II errors), and an increase in the government’s assessment of the costs of Type I errors (and their potential impact on the likelihood of EU admission). In 1998, the PKK threat was “escalating” (Bruinessen 1998, cited in Kocher 2007, 1), supporting fear of Type II errors. The security situation changed quickly and dramatically. “As of 2002, the Kurdish Southeast of Turkey is largely pacified. PKK guerrillas have withdrawn across the Iraqi border” (Kocher 2007, 2). The result is that, “for all intents and purposes, PKK is no longer a military presence in Turkey” (Kocher 2007, 2). The dramatic reduction in threat posed by the PKK made Type II errors more acceptable. At the same time, the EU put specific conditions on human rights protections in Turkey as a condition for admission (Schimmelfennig et al. 2003). “The EU expected Turkey to…abolish the death penalty and torture” (Bayer and Onis 2010, 187). This high level of scrutiny put a premium on avoiding Type I errors. Thus, the decreasing nature of the security threat posed by the PKK and the desire for Turkey to enter the EU changed the incentives for error types and altered the decision to torture prisoners. As a result of changes such as adopting a revised penal code that criminalized torture, in 2004 “the European Council agreed to open accession negotiations with Turkey on the grounds it had sufficiently fulfilled the Copenhagen criteria” (Bayer and Onis 2010, 189). In terms of our model, parameter t decreased, whereas parameter d increased as the threat represented by the PKK was diminished. At the same time, the costs of violating torture norms increased. Graphically, the situation is similar to that of Figure 3: Information obtained under torture must now be far more reliable to justify the practice. More recently, however, the threat of Kurdish revolt has increased the heavy handedness of Turkish security forces, whereas the economic crisis in the EU has made membership less attractive. The result is that Turkey is increasingly accepting of Type I and fearful of Type II security errors as evidenced by a powerful new domestic security law that entered force in April of 2015 (MENAFN 2015). The two vignettes demonstrate that variation in the incentives for error type can alter preferences for unreliable information. Both vignettes also show that torture policy varies—and thus requires explanations that also vary. Conclusion Why might those who agree on the ethical imperative of anti-torture policies disagree on the utility of torture? We think the two sides talk past each other; each focuses on minimizing a different type of error. Those motivated by the ticking bomb scenario want to do everything possible to avoid a horrific Type II error—failing to stop an event that will have deadly consequences. Those who see torture as ineffective point out the Type I errors committed by acting on the bad data resulting from torturing captives. The debate then is really about the utility of Type I versus Type II errors and not about the morality of torture. By not addressing the error type issue, the two sides fail to engage the key issue driving their disagreement. Prisoners must also make a choice. Interrogated prisoners, who want the state to respond with an imbalanced “over the top” reaction, need to provide enough true information in the mix with lies as incentive for the state to act. This creates a strategic question: How little truth can the states anticipate from prisoners and still employ torture? This question links information reliability, error types, and the choice to torture despite the cost. Because our approach is not regime-type specific, we can help explain variation in the use of torture between and within regime types, which is critical given the increasing practice among democracies. Although the number of democratic states increased in the last quarter of the twentieth century, the practice of torture also increased. “During the final quarter of the twentieth century, freedom from torture was the most widely violated human right to the physical integrity of the person” (Conrad and Moore 2010, 459; although there may have also been changes in monitoring standards, Fariss 2014). Our approach represents a general process that is not unique to the calculus of torture; the treatment of information in many situations depends on the balance between fears of making errors of commission and omission. For example, our approach helps to explain why the media, politicians, firms, and individuals react to non-random surveys, such as Internet polls that are recognized as unrepresentative and highly erroneous. If these surveys address an unbalanced situation where the cost of a Type II error dramatically outweighs a Type I error, our model can identify the conditions in which this information—even if recognized as inaccurate—leads to action. More generally, the model can be adapted to fit a wide variety of topics in which the key dynamic is an excessive fear of making a Type II error. Does this study provide a rational justification for torture? Absolutely not! We provide an examination of the decision-making structure underpinning choices over torture. Our objective is to make opaque arguments more transparent so they can be more clearly viewed and more effectively evaluated. Just like studying war does not make one a hawk (Gartner 1998), systematically studying torture does not make us torture champions (Wantchekon and Healy 1999; Schiemann 2012). Only by understanding the rational dynamics of torture can we appreciate why states value it. In fact, understanding these underpinnings might help torture opponents to have greater influence. Opponents of torture diligently advocate anti-torture norms and treaties. But these actions have seemingly failed to stymie the practice: the “spreading of an international norm against torture has, according to our findings, not limited the practice of torture to any extent” (Gilligan and Nesbitt 2009, 467–95). Our argument suggests that these appearances might be deceptive; a state can have a high cost for Type I errors, a relatively low benefit for avoiding Type II errors, and still employ torture if the cost of violating the norm is low enough. Steps taken to enhance the norm might thus help to reduce torture, even if this cannot be observed directly. At the same time, if efforts to promote anti-torture policies occur simultaneously with dramatic increases in the preference for avoiding Type II intelligence errors, states might practice torture despite embracing anti-torture norms. For example, one reason anti-torture norms appear ineffective may be that after 9/11 states across the globe raised their estimates on the costs of Type II errors, thus leading to an increase in the practice of torture and a greater willingness to accept more unreliable information. Given the vivid fears generated by terrorism, those attempting to decrease the practice of torture might be more effective trying to lower the costs of Type II and raise the costs of Type I intelligence errors, rather than continuing to champion already widely accepted anti-torture norms. Footnotes 1 Adopted by the General Assembly December 1984 and entered force in June 1987. Signed (1988) and ratified (1944) by the United States. US reservations included: “in order to constitute torture, an act must be specifically intended to inflict severe physical or mental pain or suffering” (United Nations 1984). 2 Although the US Supreme Court found in Hamdan v. Rumsfeld that Common Article III applied to the US War on Terror, the Bush Administration developed ways to work around this prohibition (Vázquez 2007). 3 Error type is different from prospect theory, which argues that how an error is framed influences risk propensity (Kahneman and Tversky 1979). For a critique of applying prospect theory to torture, see Johnson and Ryan (2015). References Adler Emanuel. 2010. “ Damned if You Do, Damned if You Don’t: Performative Power and the Strategy of Conventional and Nuclear Defusing.” Security Studies  19: 199– 229. Google Scholar CrossRef Search ADS   Alexander Matthew. 2009. “ Torture Doesn’t Work. The Daily Beast.” Accessed May 2, 2014. http://www.thedailybeast.com/articles/2009/04/20/torture-doesnt-work.html#. Amnesty International. 2001. “ Turkey: An End to Torture and Impunity is Overdue.” Accessed April 06, 2016. https://www.amnesty.org/en/documents/eur44/026/2002/en/. Armshaw Patrick. 2011. The Torturer's Dilemma: Analyzing the Logic of Torture for Information. Diss., Tallahassee, Florida: Florida State University. Avdan Nazli. 2012. “ Human Trafficking and Migration Control Policy: Vicious or Virtuous Cycle?” Journal of Public Policy  32: 171– 205. Google Scholar CrossRef Search ADS   Avdan Nazli. 2013. “ Controlling Access to Territory Economic Interdependence, Transnational Terrorism, and Visa Policies.” Journal of Conflict Resolution  58: 592– 624. Google Scholar CrossRef Search ADS   Avruch Kevin. 2003. “ Type I and Type II Errors in Culturally Sensitive Conflict Resolution Practice.” Conflict Resolution Quarterly  20: 351– 71. Google Scholar CrossRef Search ADS   Babad Elisha Michael Hills, O’driscoll Michael. 1992. “ Factors Influencing Wishful Thinking and Predictions of Election Outcomes.” Basic and Applied Social Psychology  13: 471. Google Scholar CrossRef Search ADS   Bagaric Mirko, Clarke Julie. 2005. “ Not Enough (Official) Torture in the World? The Circumstances in Which Torture Is Morally Justifiable.” University of San Francisco Law Review  581– 616. Baldor Lolita C., Benac Nancy. 2014. “ U.S. Prepares for Trial of Captured Benghazi Attack Suspect.” PBS . Accessed August 13, 2014. http://www.pbs.org/newshour/rundown/u-s-prepares-trial-captured-benghazi-attack-suspect/. Bar-Joseph Uri, Kruglanski Arie W.. 2004. “ Intelligence Failure and Need for Cognitive Closure: On the Psychology of the Yom Kippur Surprise.” Political Psychology  24: 75– 99. Google Scholar CrossRef Search ADS   Bar-Joseph Uri, McDermott Rose. 2008. “ Change the Analyst and Not the System: A Different Approach to Intelligence Reform.” Foreign Policy Analysis  4: 127– 45. Google Scholar CrossRef Search ADS   Bayer Resat, Onis Ziya. 2010. “ Turkish Big Business in the Age of Democratic Consolidation: The Nature of Limits of Its Influence.” South European Society and Politics  15: 181– 201. Google Scholar CrossRef Search ADS   Betts Richard K. 1978 . “Why Intelligence Failures Are Inevitable.” World Politics  31: 61– 89. Google Scholar CrossRef Search ADS   Blanken Leo J. 2012. Rational Empires: Institutional Incentives and Imperial Expansion . Chicago, IL: University of Chicago Press. Google Scholar CrossRef Search ADS   Blanken Leo J., Overbaugh Justin. 2012. “ Looking for Intel?…Or Looking for Answers? Reforming Military Intelligence for a Counterinsurgency Environment.” Intelligence and National Security  27: 559– 75. Google Scholar CrossRef Search ADS   Buckley Christopher. 2009. “ Enough with Torture Sanctimony.” The Daily Beast . Accessed April 08, 2016. http://www.thedailybeast.com/articles/2009/04/19/enough-with-the-torture-sanctimony.html#. Bueno de Mesquita Bruce, Alastair Smith, Randolph M. Siverson, James D. Morrow 2003. The Logic of Political Survival . Cambridge: MIT Press. Cardina Miguel. 2013. “ To Talk of Not to Talk: Silence, Torture and Politics in the Portuguese Dictatorship of Estadu Novo.” Oral History Review  40: 251– 70. Google Scholar CrossRef Search ADS   Carpenter Daniel P. 2004 . “The Political Economy of FDA Drug Review: Processing, Politics, and Lessons for Policy.” Health Affairs  23: 52– 63. Google Scholar CrossRef Search ADS PubMed  Central Intelligence Agency. Inspector General (CIA). 2004. “ Counterterrorism Detention and Interrogation Activities (September 2001–October 2003).” The New York Times . Accessed July 18, 2013. http://graphics8.nytimes.com/packages/pdf/politics/20090825-DETAIN/2004CIAIG.pdf. Clarke Richard A. 2009. “ Cheney and Rice Remember 9/11. I Do, Too.” The Washington Post Opinions . Accessed May 2, 2014. http://www.washingtonpost.com/wp-dyn/content/article/2009/05/29/AR2009052901560.html. Conrad Courtney Ryals, Moore Will H.. 2010. “ What Stops Torture.” American Journal of Political Science  54: 459– 76. Google Scholar CrossRef Search ADS   Conrad Courtney Ryals, Moore Will H.. 2011. “ Country-Year Data Users Guide.” The Ill-Treatment & Torture (ITT) Data Collection Project . Accessed February 14, 2014. http://faculty.ucmerced.edu/cconrad2/Academic/Data_files/ITT_CY_UsersGuide19July11.pdf/. Danelius Hans. 2008. “ Convention Against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment.” United National Audiovisual Library of International Law , 1– 4. Accessed April 06, 2016. http://legal.un.org/avl/pdf/ha/catcidtp/catcidtp_e.pdf. Dershowitz Alan M. 2003 . “Torture Warrant: A Response to Professor Strauss.” New York Law School Law Review  48: 275. Dragu Tiberiu. 2011. “ Is There a Trade-Off Between Security and Liberty? Executive Bias, Privacy Protections, and Terrorism Prevention.” American Political Science Review  105: 64– 78. Google Scholar CrossRef Search ADS   European Commission. 1998. “ Regular Report on Turkey’s Progress towards Accession.” Accessed April 06, 2016. http://ec.europa.eu/enlargement/archives/pdf/key_documents/1998/turkey_en.pdf. European Commission. 2002. “ Regular Report on Turkeys Progress Towards Accession.” Accessed April 07, 2016. http://ec.europa.eu/enlargement/archives/pdf/key_documents/2002/tu_en.pdf. European Commission. 2012. “ Turkey Progress Report.” Accessed April 06, 2016. http://ec.europa.eu/enlargement/pdf/key_documents/2012/package/tr_rapport_2012_en.pdf. Fariss Christopher J. 2014. “ Respect for Human Rights Has Improved Over Time: Modeling the Changing Standard of Accountability.” American Political Science Review  108: 297– 318. Google Scholar CrossRef Search ADS   Feinstein Diane. 2014. United States Senate Committee on Intelligence. “Committee Study of the Central Intellgence Agency’s Detenction and Interrogation Program.” Accessed April 06, 2016. https://web.archive.org/web/20141209165504/http://www.intelligence.senate.gov/study2014/sscistudy1.pdf. Fernandes Desmond. 2012. “ Modernity and the linguistic genocide of Kurds in Turkey.” International Journal of the Sociology of Language  217: 75– 98. Finn Peter. 2008. “ Retired Officers Urge Obama to End Harsh Interrogations. Houston Chronicle.” Accessed May 2, 2014. http://www.chron.com/news/nation-world/article/Retired-officers-urge-Obama-to-end-harsh-1775419.php. Foley Conor. 2003. Combating Torture: A Manual for Judges and Prosecutors . Human Rights Centre, University of Essex, Colchester, United Kingdom. Forsythe David, P. 2006. “ United States Policy Towards Enemy Detainees in the War on Terror.” Human Rights Quarterly  28: 465– 91. Google Scholar CrossRef Search ADS   Gartner Scott Sigmund. 1998. “ Opening up the Black Box of War.” The Journal of Conflict Resolution  42. Gartner Scott Sigmund. 2013. “ All Mistakes Are Not Equal: Intelligence Errors & National Security.” Intelligence and National Security  28: 634– 54. Google Scholar CrossRef Search ADS   Gartner Scott Sigmund, Regan Patrick M.. 1996. “ Threat and Repression: The Non-linear Relationship Between Government and Opposition Violence.” The Journal of Peace Research  33: 273– 88. Google Scholar CrossRef Search ADS   Gilligan Michael J., Nesbitt Nathaniel H.. 2009. “ Do Norms Reduce Torture?” The Journal of Legal Studies  38: 445– 70. Google Scholar CrossRef Search ADS   Gronke Paul, Darius Rejali, Dustin Drenguis, James Hicks, Peter Miller, Bryan Nakayama. 2010. “ US Public Opinion on Torture, 2001–2009.” PS: Political Science and Politics  43: 437– 44. Google Scholar CrossRef Search ADS   Hanson R. 2010. “ Torture Kids Instead.” Retrieved July 8, 2013. Overcoming Bias. http://www.overcomingbias.com/2010/01/torture-kids-instead.html. Harnden Toby. 2014. “ Obama’s Mission Impossible in Irak.” The Sunday Times , August 24, 20– 21. Harris Gardiner. 2004. “ Lawmaker Says F.D.A Held Back Drug Data.” New York Times . Accessed January 28, 2009. http://nytimes.com/2004/09/10/politics/10trial.html. Hayden Michael, Mukasey Michael B.. 2009. “ The President Ties His Own Hands on Terror: The Point of Interrogation is Intelligence, Not Confession.” The Wall Street Journal . Accessed May 2, 2014. http://online.wsj.com/news/articles/SB123993446103128041?mg=reno64-wsj&url=http%3A%2F%2Fonline.wsj.com%2Farticle%2FSB123993446103128041.html. Human Rights Watch. 2004. “ Turkey: EU Bid Hinges on Further Rights Reforms” (14 June, 2004). Accessed April 06, 2016. https://www.hrw.org/news/2004/06/14/turkey-eu-bid-hinges-further-rights-reforms. Human Rights Watch. 2006. “ Turkey: First Steps Toward Independent Monitoring of Police Stations and Gendarmeries.” Briefing Paper No. 1. Accessed April 06, 2016. https://www.hrw.org/legacy/backgrounder/eca/turkey0306/turkey0306.pdf. Intriligator Michael D. 1996. “ Drug Evaluations: Type I vs. Type II Errors.” UCLA Research Program in Pharmaceutical Economics and Policy Paper, 96–2. Accessed April 06, 2016. http://escholarship.org/uc/item/5fg9n284. Irvine David R. General. 2005. “ Why Torture Doesn’t Work.” Accessed April 06, 2016. www.alternet.org/story/28585/why_torture_doesn%27t_work. Joint Personnel Recovery Agency. 2002. Operational Issues Pertaining to the Use of Physical/Psychological Coercion in Interrogation, 1–2. Accessed April 06, 2016. http://www.washingtonpost.com/wp-srv/nation/pdf/JPRA-Memo_042409.pdf?sid=ST2009042403231. Johnson David Blake, Ryan John Barry. 2015. “ The Interrogation Game: Using Coercion and Rewards to Elicit Information from Groups.” Journal of Peace Research  52: 822– 37. Google Scholar CrossRef Search ADS   Kahneman Daniel, Tversky Amos. 1979. “ Prospect Theory, an Analysis of Decision Making under Risk.” Econometrica  47: 263– 91. Google Scholar CrossRef Search ADS   Kam Ephraim. 1988. Surprise Attack.  Cambridge, MA: Harvard University Press. Google Scholar CrossRef Search ADS   Keller Bill. 1987. “ Moscow Dismisses More Air Generals.” New York Times . Accessed April 06, 2016. http://www.nytimes.com/1987/06/18/world/moscow-dismisses-more-air-generals.html. Kelly Erin. 2014. “ Senate Report Reopens Battle Over CIA Interrogations.” USA Today . Accessed September 12, 2014. http://www.usatoday.com/story/news/politics/2014/12/09/cia-torture-report-senate-intelligence-feinstein/20134093/. Kelly Christopher E., Miller Jeaneé C., Redlich Allison D., Kleinman Steven M.. 2014. “ Taxonomy of Interrogation Methods.” Psychology, Public Policy and Law  19: 165– 75. Accessed July 2, 2014. http://www.lexisnexis.com.libproxy.nps.edu/lnacui2api/frame.do?tokenKey=rsh-20.749457. Google Scholar CrossRef Search ADS   Kessler Glenn. 2009. “ Rice Defense Enhanced Interrogations.” The Washington Post . Accessed May 2, 2014. http://voices.washingtonpost.com/44/2009/04/30/rice_defends_enhanced_interrog.html. Kocher Matthew. 2007 . “The Decline of PKK and the Viability of a One-State Solution in Turkey.” International Journal on Multicultural Societies  4: 126– 47. Koppl Roger. 2006. “ Epistemic Systems.” Episteme  2: 91– 106. Google Scholar CrossRef Search ADS   Lampel Joseph, Shapira Zur. 2001. “ Judgmental Errors, Interactive Norms, and the Difficulty of Detecting Surprises.” Organizational Science  12: 599– 611. Google Scholar CrossRef Search ADS   Liberman Peter. 2013. “ Retributive Support for International Punishment and Torture.” Journal of Conflict Resolution  57: 285– 306. Google Scholar CrossRef Search ADS   Luban David. 2007. “ Liberalism, Torture, and the Ticking Bomb.” In Intervention, Terrorism, and Torture , Lee Steven P. (ed.), 249– 62. Springer. Malinowski Tom. 2008. “ Restoring Moral Authority: Ending Torture, Secret Detention and the Prison at Guantanamo Bay.” Annals of the Academy of Political and Social Science  618: 148– 52. Google Scholar CrossRef Search ADS   Menafn. 2015. “ Turkey’s Domestic Security Law Enters Into Force.” The Journal of Turkish Weekly . Accessed April 06, 2016. http://www.menafn.com/1094151409/Turkeys-domestic-security-law-enters-into-force. Mercer Jonathan. 2005. “ Rationality and Psychology in International Politics.” International Organization  59: 77– 106. Google Scholar CrossRef Search ADS   Morgan Rod. 2000. “ The Utilitarian Justification for Torture: Denial, Desert, and Disinformation.” Punishment & Society  2: 181– 96. Google Scholar CrossRef Search ADS   Mulrine Anna, 2014. “ Why the US Special Forces Failed to Rescue James Foley.” The Christian Science Monitor . Accessed April 06, 2016. http://www.csmonitor.com/USA/Military/2014/0821/Why-US-special-forces-failed-to-rescue-James-Foley. O’Mara Shane. 2009. “ Torturing the Brain: On the Folk Psychology and Folk Neurobiolgy Motivating ‘Enhanced and Coercive Interrogation Techniques.’” Trends in Cognitive Sciences  13: 497– 500. Google Scholar CrossRef Search ADS PubMed  Persson Mats, Siven Claes-Henric. 2007. “ The Becker Paradox and Type I versus Type II Errors in the Economics of Crime.” International Economic Review  48: 211– 33. Google Scholar CrossRef Search ADS   Phillips Joshua E. S. 2012. None of US Were Like This Before: American Soldiers and Torture . London, UK and Brooklyn, NY, USA: Verso Books. Pollard P., Richardson J. T. E.. 1987. “ On the Probability of Making Type-I Errors.” Psychological Bulletin  102: 159– 63. Google Scholar CrossRef Search ADS   Rejali Darius. 2004. “ Torture’s Dark Allure.” Salon . July 20, 2013. http://www.salon.com/2004/06/18/torture_1. Rejali Darius. 2007. “ 5 Myths about Torture and Truth.” Washington Post Opinions . Accessed April 06, 2016. http://www.salon.com/2004/06/18/torture_1/. Rejali Darius. 2009. Torture and Democracy . Princeton, NJ: Princeton University Press. Ridge Tom. 2009. “ We Were Wrong to Torture.” BBC . Accessed May 2, 2014. http://news.bbc.co.uk/2/hi/americas/7903516.stm. Roberts J. D. 2013. Error Avoidance Theory: Sniper Employment for Military and Civilian Law Enforcement. Doctoral diss., Naval Postgraduate School, Monterey, CA. Rose David. 2008. Tortured Reasoning. Vanity Fair . Accessed July 11, 2013. http://www.vanityfair.com/magazine/2008/12/torture200812?currentPage=2. Schiemann John W. 2012. “ Interrogational Torture: Or How Good Guys Get Bad Information with Ugly Methods.” Political Research Quarterly  3– 19. Schimmelfennig Frank, Engert Stefan, Knobel Heiko. 2003. “ Costs, Commitment and Compliance: The Impact of EU Democratic Conditionality on Latvia, Slovakia and Turkey.” JCMS: Journal of Common Market Studies  41: 495– 518. Google Scholar CrossRef Search ADS   Semel Matthew D. 2013. “ Military Interrogations: Best Practices and Beliefs.” Perspectives on Terrorism  7: 39– 62. Shue Henry. 1978. “ Torture.” Philosophy & Public Affairs  7: 124– 43. Silver James. 2004. “ Why America’s Top Liberal Lawyer Wants to Legalise Torture. The Scotsman.” Accessed August 7, 2013. http://www.scotsman.com/news/world/why-america-s-top-liberal-lawyer-wants-to-legalise-torture-1-532091. Sobel Russell S., Leeson Peter T.. 2006. “ Government’s Response to Hurricane Katrina: A Public Choice Analysis.” Public Choice  127: 55– 73. Google Scholar CrossRef Search ADS   Soufan Ali H. 2009a. “ My Tortured Decision.” New York Times . Accessed July 26, 2013. http://www.nytimes.com/2010/09/26/opinion/terr-soufan.html?gwh=2EB875DFA290006C9D2B5BDD3E7B1589. Soufan Ali H.. 2009b. “ What Torture Never Told US.” New York Times . Accessed July 26, 2013. http://www.nytimes.com/2009/09/06/opinion/06soufan.html. Suskind R. 2006. One Percent Doctrine: Deep Inside America's Pursuit of Its Enemies Since 9/11 . New York: Simon and Schuster. Sussman David. 2005. “ What’s Wrong with Torture?” Philosophy & Public Affairs  33: 1– 33. Google Scholar CrossRef Search ADS   US Army. 1992. “ Intelligence Interrogation.” In US Army Field Manual , 34– 52. Washington, DC: Army Publishing Directorate. PubMed PubMed  US Department of Justice, Office of the Inspector General. 2009. A Review of the FBI’s Involvement in and Observations of Detainee Interrogations in Guantanamo Bay, Afghanistan, and Iraq . Washington, DC: Department of Justice. US Senate. 2008. Inquiry Into the Treatment of Detainees in US Custody . United States Senate. Washington, DC: Committee on Armed Services. United Nations. 2004. “ Manual on the Effective Investigation and Documentation of Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment (‘The Istanbul Protocol’).” Accessed April 06, 2016. http://www.ohchr.org/documents/publications/training8rev1en.pdf. United Nations General Assembly. 1984. “ Convention against Torture and Other Cruel, Inhuman or 15 Degrading Treatment or Punishment.” A/RES/39/46, Meeting No. 93 . Accessed April 08, 2016. https://treaties.un.org/Pages/ViewDetails.aspx?src=IND&mtdsg_no=IV-9&chapter=4&lang=en. Vázquez Carlos Manuel. 2007. “ The Military Commissions Act, the Geneva Conventions, and the Courts: A Critical Guide.” The American Journal of International Law  101: 73– 98. Walzer Michael. 1973. “ The Problem of Dirty Hands.” Philosophy and Public Affairs  2: 160– 80. Wantchekon Leonard, Healy Andrew. 1999. “ The ‘Game’ of Torture.” The Journal of Conflict Resolution  43: 596– 609. Google Scholar CrossRef Search ADS   Weiser Benjamin. 2014. “ Some Captured Terrorists Talk Willingly and Proudly Interrogators Say.” New York Times . Accessed April 06, 2016. http://www.nytimes.com/2014/10/14/nyregion/some-captured-terrorists-talk-willingly-and-proudly-investigators-say.html?_r=0. Wirtz James J. 1989. “ The Intelligence Paradigm.” Intelligence and National Security  4: 829– 37. Google Scholar CrossRef Search ADS   Yingling Raymund T., Ginnane Robert W.. 1952. “ The Geneva Conventions of 1949.” The American Journal of International Law  46: 393– 427. Google Scholar CrossRef Search ADS   Zeydanlıoğlu Welat. 2009. “ Torture and Turkification in the Diyarbakır Military Prison.” In, Rights, Citizenship & Torture: Perspectives on Evil, Law and the State . Zeydanlıoğlu Welat and Parry John T. (eds.), 73– 92. Oxford: Inter-Disciplinary Press. Zhao Bin, Olivera Fernando. 2006. “ Error Reporting in Organizations.” Academy of Management Review  31: 1012– 30. Google Scholar CrossRef Search ADS   © The Author (2016). Published by Oxford University Press on behalf of the International Studies Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Foreign Policy Analysis Oxford University Press

Unbalanced Policy Priorities and the Interrogation of Terror Suspects

Loading next page...
 
/lp/ou_press/unbalanced-policy-priorities-and-the-interrogation-of-terror-suspects-DNazCeLvUN
Publisher
Wiley-Blackwell
Copyright
© The Author (2016). Published by Oxford University Press on behalf of the International Studies Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1743-8586
eISSN
1743-8594
D.O.I.
10.1093/fpa/orw020
Publisher site
See Article on Publisher Site

Abstract

Abstract Why do states shift from opposition to torture to its employment in the face of widespread abhorrence and claims that it generates unreliable information? The expected value of acting upon unreliable information depends on the state’s appraisal of the type of error it might be committing. States may value avoiding Type I errors, which result from acting upon incorrect intelligence, differently than avoiding Type II errors, which result from failing to act on an undetected threat. Interrogations yield a mix of truth and lies; operations employing this information will avoid making Type II errors but will also make Type I errors. A state’s preference over error type changes with circumstances (as we show, occurs in Turkey and the United States), leading states to adopt coercive interrogation techniques despite secular abhorrence of torture. Our results suggest anti-torture advocacy might be more effective targeting preferences over error types than reinforcing anti-torture moral imperatives. A majority of states, many democratic, torture prisoners (Conrad et al. 2011). Why do nations—especially those motivated by liberal norms—practice torture in the face of broad and intense condemnation (Sussman 2005)? Factors such as retribution (Liberman 2013), intimidation, and false confessions might encourage states to employ torture (Rejali 2009). Torture, however, is most often justified as an information tool (O’Mara 2009), with many stating that intelligence efficacy alone provides a liberal justification for its practice. The “only one rationale for torture that might conceivably be acceptable to a liberal: torture as a technique of intelligence gathering from captives who will not talk” (Luban 2007, 1436). In an increasingly networked world, the value of this information has grown: “The global nature of transnational terrorism necessitates that in addition to reacting to directed threats from terrorism, states harness information about the ties of origins’ nationals in attacks worldwide” (Avdan 2013, 6). Thus, a critical answer to the why-do-liberal-states-torture question seems to be the claim that sometimes valuable, timely information necessary for their security can only be obtained using coercive methods. The question, however, is even more complex as states’ policies toward torture vary over time. What conditions shift those opposed to torture to its support? Focusing on policy change requires that we examine political factors that vary (Avdan 2012; Blanken 2012). These political factors can be external or internal to the state (Bayer and Onis 2010). Most literature in this area is structural and thus cannot explain change in torture policy (Conrad and Moore 2010). For example, what altered to cause the notable increase in the global use of coercive practices such as waterboarding following the attacks of September 11, 2001? Some might see this question as trivial—obviously, the deaths of almost 3,000 people on 9/11 generated tremendous feelings of anger, hate, and fear that led many to desire revenge. However, changing counterterrorism policy is difficult (Blanken and Overbaugh 2012), and after 9/11, a majority of the American public did not support the use of torture (Gronke et al. 2010). Thus, it seems unlikely that revenge alone can explain coercive interrogation policies post 9/11 in the United States, let alone in nations that were not directly affected by the attacks (Conrad and Moore 2010). We develop a simple game theoretic model that links policy preferences to the decision to torture and, in so doing, illustrates informational motivations for changing torture policy. As such, our work embraces the viewpoint that it is important to understand the possible rational underpinnings of a state’s decision to torture (Dragu 2011; Schiemann 2012). Our intent is not to justify as rationally desirable an obviously abhorrent practice, but instead to gain a deeper and more nuanced understanding of the informational conditions under which states—even those that abhor torture—will resort to such extreme measures (Wantchekon and Healy 1999, 597). What differentiates us from previous game theoretic treatments of torture is our focus on error types. Errors can be either Type I (mistakes of commission) or Type II (mistakes of omission). For example, a Type I error would include executing someone for a crime they were later found not to have committed. A Type II error, in contrast, would be failing to arrest someone for a crime they had committed. In the first case, the state commits an error of action, and in the second case, the error is that the state fails to act. Although institutions and individuals always try to avoid mistakes, states have preferences over the type of error they want to avoid most, and these preferences fundamentally affect their calculations about the usefulness of information gained through torture. In particular, rational leaders who are strongly biased against Type II errors will still consider “dirtying their hands” with torture to get what they perceive may be vital information necessary to avoid an error of omission (Walzer 1973). Our goal is to use states’ desire to avoid Type II errors as a conceptual tool for developing a better understanding of how unreliable information must be before states decide to forgo torture in the face of potential security threats, avoiding its reputational and ethical costs. We begin by examining the role of torture in intelligence gathering, error types, and how the two interact. We then present a model that analyzes decisions to use torture in view of the quality of the information it likely reveals. We illustrate the game with brief vignettes of torture policy changes in the United States and Turkey and conclude with thoughts on future research and policy. Torture A common definition of torture draws on the United Nations “Convention against Torture and Other Cruel, Inhuman Degrading Treatment or Punishment” (Danelius 2008, 1)1: any act by which severe pain or suffering, whether physical or mental, is intentionally inflicted on a person for such purposes as obtaining from him or a third person information or a confession…when such pain or suffering is inflicted by or at the instigation of or with the consent or acquiescence of a public official or other person acting in an official capacity. (United Nations General Assembly 1984) There are two main reasons provided supporting the prohibition of torture: ethics and efficacy. Ethics States have long viewed torture as inhumane and cruel and banned its practice. The Geneva Conventions (signed in 1949 and ratified by virtually all countries) ban the torture of prisoners of war (Yingling and Ginnane 1952) and promote the “humane treatment and interrogation of prisoners” (Kelly et al. 2013). International prohibitions of torture were expanded by the Torture Convention cited above.2 The most basic normative argument against torture is that it vitiates “the inherent dignity of the human person” by violating peoples’ “inalienable rights” (United Nations General Assembly 1984). Torture does not just reflect an extreme version of other types of punishments but represents a categorical, “special” bad (Sussman 2005). Shue (1978, 124) argues that “no other practice except slavery is so universally and unanimously condemned in law and human convention.” International law places the condemnation of torture among jus cogens or “peremptory norms,” a prohibition that cannot be overturned by treaty or law and is binding regardless of whether or not torture-specific prohibitions are adopted (Foley 2003, 4–5). Torture is seen as irreconcilable with modern democratic values because “[t]orture aims to strip away from its victims all the qualities of human dignity that liberalism prizes” (Luban 2007, 1430). But, democracies do nasty things (covert actions, colonialism, etc.) when they feel like they are unobserved by their constituents (Bueno de Mesquita et al. 2003), suggesting anti-torture normative constraints are part of the exogenous “strategic context” within which rational choice actors maximize utility. Effectiveness Effectively obtaining information represents one of the most commonly identified objectives of torture. “As a method for extracting information, torture currently is quite widely used” (Wantchekon and Healy 1999, 597). While some believe that torture’s effectiveness at procuring information is irrelevant to its prohibition (“a democracy that legalizes the use of torture in its desperation to gain information loses something more important—the trust of its people, the foundation of a democracy” Rejali 2004), others clearly disagree and see torture, even if intrinsically bad, as a vital intelligence tool (Shue 1978; Dershowitz 2003). As Schiemann (2012, 3) states, “determining whether interrogational torture works—defined as providing clear and valuable information to the state—is essential. First, if it does not work, then the main rationale for its use—it is the only method we have left when detainees refuse to talk—disappears, and any further use is pure sadism.” Armshaw (2011, 139) agrees, writing that, “Ethically, torture could only be justified as a means of generating information that could not be found (usefully) in any other fashion.” Even Rejali (2009, 23) states that “The heart of the matter is whether an organization can apply these techniques…scientifically and professionally to generate true and reliable intelligence.” So does torture work? The widespread answer is no. Intelligence agency members and interrogators frequently point to the highly inaccurate nature of the information obtained through torture (Morgan 2000; Finn 2008; Rose 2008; Alexander 2009; O’Mara 2009; Soufan 2009a; Schiemann 2012). For example, interrogation practitioners such as those of the US Army state that “the use of torture and other illegal methods is a poor technique that yields unreliable results, may damage subsequent collection efforts, and can induce the source to say whatever he thinks the interrogator wants to hear” (US Army 1992, 1–8). This anti-torture view is not just policy but reflects actual Army and other organizations’ evaluations (Irvine 2005). For example, recently released Central Intelligence Agency (CIA) documents “fail to show that [enhanced interrogation] techniques stopped even a single imminent threat of terrorism” (Soufan 2009b). physical and/or psychological duress are viewed as an alternative to the more time-consuming conventional interrogation process. The error inherent in this line of thinking is the assumption that, through torture, the interrogator can extract reliable and accurate intelligence. History and a consideration of human behavior would appear to refute this assumption. (US Joint Personnel Recovery Agency 2002, 1) Reflecting these views, US military personnel oppose torture at higher rates than American civilians—especially military personnel who served in Iraq (Gronke et al. 2010, 438, 441). But not everyone agrees. Some believe torture should play a role in intelligence (Hanson 2010). Others maintain that torture is vital for obtaining critical national security information. For example, former CIA Director Hayden and Attorney General Mukasey (2009) argue that “coercive interrogations” generate “intelligence” and add that it is vital to corroborate all intelligence information. Bagaric and Clarke (2005, 588) state, “The main benefit of torture is that it is an excellent means of gathering information.” Republican Senate leader McConnell advocated dropping legal restrictions on interrogating a suspect in the Benghazi attack saying, “The most valuable thing we can get from this terrorist is information” (Baldor and Benac 2014). Some make exceptions to torture prohibitions in the face of potentially deadly and time-dependent threats in what is widely known as the “ticking bomb” scenario (Silver 2004). In its classic form, Walzer states that a leader is asked to authorize the torture of a captured rebel leader who knows or probably knows the location of a number of bombs hidden in apartment buildings around the city, set to go off within the next twenty four hours. He orders the man tortured, convinced that he must do so for the sake of the people who otherwise die in the explosions—even though he believes that torture is wrong, indeed abominable, not just sometimes, but always. (1973, 167) In this scenario, the ability to obtain information that prevents the imminent loss of life creates allowable circumstance for using torture—even if it is recognized as morally wrong (Dershowitz 2003). Conditions can justify “dirty hands” without legitimizing torture (Walzer 1973): “it cannot be denied that there are imaginable cases in which the harm that could be prevented by a rare instance of pure interrogational torture would be so enormous as to outweigh the cruelty of the torture itself” (Shue 1978, 141). Thus, the critical factor is the utility of the information. Types of Error An operation to rescue Americans held captive by ISIS failed when US Special Operations Forces arrived at a base shortly after the prisoners had already been moved (Mulrine 2014). Among the prisoners relocated was journalist James Foley, who was later beheaded in an act that lead to a significant increase in the US military presence in Iraq (Harnden 2014). Wartime intelligence mistakes like these are unavoidable. “Intelligence failures are not only inevitable, they are natural” (Betts 1978, 88). Although errors are inevitable, the type of mistake committed can vary (Kam 1988; Bar-Joseph and Kruglanski 2004; Bar-Joseph and Rose McDermott 2008). A key question is what type of error is more acceptable—Type I (commission) or Type II (omission) (Pollard and Richardson 1987; Sobel and Leeson 2006; Zhao and Olivera 2006)?3 Decision makers have to choose between the “cry wolf” problem and “miscalculated escalation” (Wirtz 1989, 834). By understanding the error type that actors try most to avoid, we can make predictions about their likely behavior (Kam 1987; Babad and O’driscoll 1992; Mercer 2005). This approach has been successful in predicting decisions on foreign policy (Lampel and Shapira 2001), conflict resolution (Avruch 2003), legal systems (Persson and Siven 2007), sniper rules (Roberts 2013), and drug approval (Intriligator 1996; Carpenter 2004; Sobel and Leeson 2006). Gartner (2013) points out that an effort to avoid one type of error may inadvertently promote the other type. In the case of information obtained under torture, it is the reliability of the information that can cause mistakes. Trying to avoid a Type II error by torturing to obtain new information does not promote Type I errors. But the mix of truth and lies that a tortured prisoner may disclose can lead a state that acts upon this information to risk Type I errors as it tries to avoid making Type II mistakes. The Game We focus on rational incentives for the State based on its evaluation of error types and for the Prisoner based on their perception of disclosure as an “extension of their jihad” (Weiser 2014). We develop a simple game model designed to capture the State’s tolerance for faulty intelligence, comparing situations in which avoiding errors of omission take precedence over making errors of commission to situations in which such policy priorities do not hold true. The model allows us to evaluate the link between unbalanced priorities over error types (disproportionately avoiding Type II errors) and the decision to use torture to obtain possibly poor quality intelligence. The State in our model must decide to use torture to obtain noisy information that it must subsequently act upon or not. Interrogated terror suspects, if they talk, are assumed to seek to incite the State to act upon the information, as overreaction facilitates recruitment (Weiser 2014). However, the State cannot ignore the possibility that it is facing a prisoner who will never talk. Cases of resistance under torture have been documented (Cardina 2013). Such prisoners will not inform regardless of whether they have valuable information to disclose or not. If the Prisoner remains silent, it is the result of character and inner resolve in the face of extreme duress. The State, therefore, makes decisions regarding the resort to torture with uncertainty about the type of Prisoner it is facing and has the following initial beliefs: with probability β, the Prisoner will break down under torture and talk; with probability (1 – β), the Prisoner is resilient and will never talk. If under torture the Prisoner talks, he discloses a mix of true and false statements, but the State cannot reliably sort out the truth from the lies. There could be many reasons for the State’s inability to validate perfectly all information. For example, in the classic “ticking bomb” scenario, interrogators have insufficient time and resources to validate all the information they extract: “the less time the authorities have to make use of the information gleaned from coercion, the less incentive they have to investigate its veracity, and the less incentive the captive has to tell the truth” (Armshaw 2011, 138; see also Koppl 2006). But it could also be the case that the information received is by its very nature hard to verify ex ante. As Mulrine (2014) points out “good US military intelligence on…militant groups is in short supply.” For example, a prisoner could disclose a plan to attack a particular site or to change a known headquarters by a particular date. In both cases, the information may be verifiable ex post, after the fact, when it is too late to protect the site or storm the group’s headquarters. The attempt to rescue Foley exemplifies this dynamic as US Special Operations Forces found a “dry hole” where American hostages were expected to be held (Harnden 2014). As a result, the State can anticipate that proportion x∈[0.1] of the Prisoner’s disclosures are false and (1-x) are true. If the tortured Prisoner talks, the State must decide whether to act upon the information received with uncertainty about its reliability. Under what conditions would a State still decide to use torture despite the cost in terms of reputation and international standing? Reputation costs incurred as a result of a state’s use of torture are varied and multi-tiered. Costs are incurred with the public acknowledgment of the state’s practices, such as domestic audience costs that compromise a democratic leader’s chances of re-election. For example, the disclosure of abuse of prisoners at Guantanamo Bay was met with immediate public outrage with attendant damage to the Bush Administration’s credibility and trustworthiness (Malinowski 2008). A state that practices torture also risks losing moral authority on the world stage. Forsythe (2006) likens the risks associated to the Bush Administration’s use of enhanced interrogation techniques to those taken by France during the Algerian war of independence. France, by engaging in such practices, lost “self-respect and reputation in the world” and “generated increased opposition at home and abroad” (Forsythe 2006, 465). Although increased opposition at home in the United States was slow to develop in the aftermath of the 9/11 attacks, “Bush policy towards enemy detainees contributed to a highly negative reaction in the Arab-Islamic world and amongst virtually all European publics” (Forsythe 2006, 475). Thus, states face a combination of domestic and international costs (Gartner and Regan 1996). In the words of Forsythe (2006, 491), “the United States gravely damaged its reputation as a champion of human rights, humanitarian norms and the rule of law.” States that do not have the stature of the United States on the world stage may incur lower reputation costs of this sort, but audience costs at home and the potential for internal dissent and destabilization remain core ingredients of the reputational costs associated to a state’s use of torture. The State’s decision to torture depends on the expected value of the information it hopes to receive. The game model is pictured in Figure 1. The State’s payoffs are in black, whereas the Prisoner’s payoffs are recorded in dark gray. Figure 1. View largeDownload slide The game model Figure 1. View largeDownload slide The game model The State makes an initial decision at nodes S1 and S2 on whether to employ torture with uncertainty about the Prisoner’s type. If the State decides against the use of torture, its payoff is E, which we interpret as the value of the information the State has currently obtained given conventional interrogation techniques. The Prisoner’s initial payoff is normalized to 0. If the State decides to torture and the Prisoner remains silent, nothing is gained and the State incurs a cost c for having resorted to these means. Its payoff is then E-c, while the Prisoner’s payoff of  -y is assumed greater than any payoff he could get by talking and having the State subsequently consider the information. If the Prisoner is not resilient and talks, he makes a decision from node P to disclose information that is part truth part lies with a proportion x of lies. The State must then make a decision at S3 to use the information obtained under torture or not. If the State does use the information received under torture, it must expect that proportion x is misleading and will lead to a Type I error. For example, the State might decide to preemptively guard against an attack that will not happen wasting resources that could have been used elsewhere. Or it could destroy an alleged terror cell safe haven killing innocent parties instead. In that case, the State’s payoff is E1-d-c as the value of the information it does have is diminished by the inaccuracies introduced by the false disclosures of the Prisoner. But proportion (1-x) of disclosures is reliable and use of these new facts increases the State’s effectiveness. In that case, a Type II error is avoided and the State’s payoff is E1+t-c. For example, the information can lead to the arrest of a high-level operative or the destruction of a cache of arms held by a terror organization. In sum, use of the disclosures made by the Prisoner under torture yields an expected payoff to the State of   x[E(1−d)−c]+(1−x)[E(1+t)−c].The Prisoner receives  -z1 if the State uses the disclosures. If the State chooses not to use the information obtained under torture, its payoff is E-c as nothing has been gained by torturing the Prisoner but a reputation cost has been incurred. The Prisoner’s payoff in that case is  -z2. What drives Prisoner and State choices? First, the Prisoner makes a strategic decision only if he does not remain silent. In this case, he must decide on the mix between truth and lies in his disclosures. We assume that the overarching motivation for the Prisoner is to see the State conduct an excessive, imbalanced response (Gartner and Regan 1996). A violent, disproportionate State reaction gives visibility to the terror group and enhances its appeal with potential recruits. We, therefore, assume that  -z1>-z2. This preference for action implies that enough truth be embedded among the lies or that x be small enough. More precisely, the Prisoner will want to choose x so that   x[E(1−d)−c]+(1−x)[E(1+t)−c]>E−c.This means that   x<tt+d or equivalently (1−x)>dt+d. (1) Note that we do not assume that the Prisoner’s payoffs depend on the magnitude of x; instead the Prisoner’s incentive, if he cannot keep silent, is to mobilize the State to act against the terror group as this furthers the cause (Adler 2010). Although there may be a preference for lies over truth, truth telling in and of itself may actually be perceived as more likely to lead to visible action against the terror organization than lies. Moreover, some terror suspects may take pride in boasting about their organization’s plans and achievements (Weiser 2014). By choosing x to satisfy (1) the Prisoner’s priorities are satisfied, however high or low x might be. This modeling choice has an added advantage in that we avoid requiring that a tortured Prisoner be assumed to pick a value for x to precisely maximize an objective function. Rather the choice is to target an ordinal preference. The lower bound on the reliability of the disclosures obtained under torture dt+d depends on the valuation of error types. It is immediately apparent from (1) that a State that values avoiding an error of omission (high t) far more than it does making an error of commission (low d) will also tolerate using information whose reliability is low. From the Prisoner’s point of view, inciting a Type II error-phobic State to action can be achieved while adding a high proportion of lies or misleading information to the mix of disclosures. How unreliable can the Prisoner’s disclosures be before the State decides to forgo the use of torture altogether? The State will only resort to torture if it anticipates that it will make use of the information received by these means. This is because torturing without using the information obtained yields a payoff of E-c<E as reputation cost c>0. The State that tortures must expect a low enough proportion of prisoners to be resilient and a high enough proportion of disclosures by prisoners who talk to be reliable. Torture becomes a rational choice if   (1−β)[E−c]+β[x(E(1−d)−c)+(1−x)[E(1+t)−c)]>E.This leads to the following inequality:   c<Eβ[t−x(t+d)]. (2)As the Prisoner’s priorities lead him to choose x<tt+d relationship, (2) can hold for small enough values of the strictly positive cost c. Stated differently, if reputation costs are small enough, the following behaviors form a subgame-perfect equilibrium (SPE) of the game pictured in Figure 1. The Prisoner who talks, discloses a mix of truth and lies where lies comprise proportion x<tt+d of the information communicated to the State. Anticipating the Prisoner’s disclosure under torture, the State knows that it will use the information obtained by these means for sure. It will, therefore, decide to resort to enhanced interrogation techniques with probability 1.Having established the existence of an SPE we ask: How reliable must information received via torture be to justify its use? And how is reliability linked to States’ valuations of Type I and II errors? In order to examine the consequences of relationship (2), we must first make assumptions about the proportion of prisoners who will not talk under torture. A State considering the use of torture must assume that some prisoners will break down under duress and that a sufficient proportion of valuable information will be revealed by these individuals. One would expect the State to expect that few prisoners will remain silent. However, whether a prisoner will talk or remain silent when tortured is a matter of debate. Rejali (2007) claims that the idea that “everyone talks sooner or later under torture” is a myth. We will, therefore, examine two cases: in one the State assumes that at most 20 percent of prisoners will be able to resist and remain silent, and in the other, the State believes that as many as 80 percent of prisoners will resist and remain silent under torture. Consider the first case. If 20 percent of prisoners remain silent, then β=0.8 and relationship (2) can be rewritten as   c<E * 0.8[t−x(t+d)], (3)where x is the proportion of prisoner disclosures that are unreliable. If the State believes that 80 percent of prisoners will remain silent, then inequality (3) reads   c<E * 0.2[t−x(t+d)]. We explore the relationship between reliability, valuation of error types, and reputation cost both numerically and graphically. In order to visualize the impact of parameter shifts on the decision to use torture, we identify the threshold x for which relationship (3) is binding, assuming first that 80 percent and then 20 percent of prisoners are expected to talk. Table 1 provides threshold values for x for various values of c and t, holding d constant at d=0.1. These are the values obtained when relationship (3) is binding. Parameter E is also held constant at E=2.5. Table 1. Threshold values for proportion x of unreliable information Case 180% of prisoners talk  t=0.2  t=0.3  t=0.7  t=1.5  c=0.3  16.7%  37.5%  68.8%  84.3%  c=1.5  The State will not torture  46.9%    Case 220% of prisoners talk  t=0.2  t=0.3  t=0.7  t=1.5    c=0.3  The State will not torture  12.5  56.3  c=1.5  The State will not torture  Case 180% of prisoners talk  t=0.2  t=0.3  t=0.7  t=1.5  c=0.3  16.7%  37.5%  68.8%  84.3%  c=1.5  The State will not torture  46.9%    Case 220% of prisoners talk  t=0.2  t=0.3  t=0.7  t=1.5    c=0.3  The State will not torture  12.5  56.3  c=1.5  The State will not torture  View Large Consider the two cells shaded in dark gray on the right side of Table 1. When the value of avoiding a Type II error t is large ( t=1.5), torture can be a rational choice even if the prisoner is unlikely to talk. If 80 percent of prisoners talk and cost c=0.3, the use of torture is rational even if as much as 84.3 percent of prisoner disclosures are misleading. If a much lower 20 percent of prisoners are assumed to talk, then torture remains a rational choice if the few prisoners that talk reveal useful information about half of the time. Of course, information reliability must be higher if fewer are expected to talk, but the rationality of the resort to torture when few prisoners are expected to talk at all comes as a surprise. As expected, if it becomes more costly to violate international torture norms, the reliability of information obtained by these means must increase considerably even if the value attributed to avoiding a Type II error is high. When only 20 percent of prisoners are expected to talk, then an increase of costs to c=1.5 rules out the use of torture under any of the parameter values examined for  t. At the other extreme, consider the cell highlighted in dark gray in the top left side of Table 1. For this cell, 80 percent of prisoners are assumed to talk under torture; the cost of violating the torture norm is set at c=0.3, but the benefit associated to avoiding a Type II error has dropped to 0.2. Now, given d=0.1, torture can only be justified if there is at least an 83.3 percent chance that information obtained via torture is valuable (it is unreliable at most 16.7 percent of the time). Clearly, increases in the reputational cost of violating the torture norm and reduced benefits to avoiding Type II errors relative to the cost of making a Type I error favor ruling out torture as a means to gather noisy intelligence information. In Figure 2, we provide a visual representation of the circumstances in which a state might resort to torture when seeking new information. In Figure 2, we fix (1-x) or the proportion of reliable information that can be expected from the disclosures of a tortured prisoner and identify the areas for c and t that satisfy cost constraint (3) assuming that 80 percent of prisoners are expected to talk under torture and given values E=2.5, and d=0.1. The figure limits t to 1.5. Figure 2. View largeDownload slide Information reliability and torture Figure 2. View largeDownload slide Information reliability and torture Figure 2 is constructed as follows: we consider various values for x, the proportion of prisoner disclosures that are false or misleading. For each value of x, we obtain a relationship between c and t for which inequality (3) is binding. These are the upward sloping lines labeled according to the value of (1-x), the proportion of disclosures that are truthful. For example, if 1-x=25 percent, meaning that the State anticipates that 25 percent of the information obtained under torture is reliable, (3) is binding when    c=0.8 * 2.5 * [t−x(t+0.1)] or c=−0.15+0.5t.This relationship between c and t is represented in Figure 2 by the bolded upward sloping line. In the shaded area below this line are all combinations of c and t for which the State finds it in its interest to use torture. For combinations of c and t above the bolded line the state will not resort to torture given the 25 percent anticipated reliability of information. If the anticipated validity of the information obtained through torture decreases to, say, 10 percent, then c must decrease and t must increase for torture to be a rational decision. The combinations of c and t for which the State tortures in this case is shaded in dark gray for values of t that remain below 1.5. Our game theoretic analysis suggests that rational states strongly biased against Type II errors will consider torture to extract information they know to be highly unreliable even if they believe that a high proportion of prisoners may not talk. As Senator Feinstein, Chair of the Senate Select Committee on Intelligence, wrote in the “Committee Study of the Central Intellgence Agency’s Detention and Interrogation Program,” “…prior to the attacks of September 2001, the CIA itself determined from its own experience with coercive interrogatons, that such techniques ‘do not produce intellegence,’ ‘will probabably result in false answers,’ and had historically proven to be ineffective. Yet these conclusions were ignored” (Feinstein 2014, 3). We unpack the calculation whereby torture is simultaneously recognized as generating unreliable information and yet still employed. Our results are not US specific. For example, we think this calculus may help to explain the hardening of interrogation practices by other democratic states following the trauma of the 9/11 attacks. We elaborate on these dynamics in the vignettes that follow looking at both the United States and Turkey. Vignettes We provide two vignettes to show how the model helps to explain actual situations. These brief descriptions are not full case studies but rather suggestive illustrations of the link between policy decisions and priorities over error types. Vignettes are especially useful with game theoretic models as they provide illustrations of different types of choice calculations. Lacking data on individual torture cases, the vignettes focus on the policy implications of the model and not its specific dynamics. The goal is not to match each aspect of the model to a historical event, but rather to capture the general flavor of the policy trade-offs we examine. United States Since 9/11 The 9/11 attacks altered US leaders’ preferences over error types and torture. 9/11—Bush 9/11 forced the United States to confront “a new kind of threat” (former Homeland Security Secretary Tom Ridge 2009) that fundamentally altered leaders’ and organizations’ views. As a result of 9/11, priorities changed. As Vice President Cheney stated, “I’ll freely admit that watching a coordinated, devastating attack on our country from an underground bunker at the White House can affect how you view your responsibilities” (Clarke 2009). The attacks altered attitudes toward types of error across American leadership and intelligence organizations. For example, in a “dramatic shift in focus of the Department of Justice (DOJ),” the FBI shifted from a emphasis on criminal activities such as catching bank robbers to a new Type II avoidant anti-terrorism perspective: “Since September 11, the Attorney General and the FBI Director have elevated counterterrorism and the prevention of future terrorist attacks against United States interests as the top priority of the DOJ and FBI” (US Department of Justice, Office of the Inspector General 2009, 11—emphasis added). This change represented “a huge paradigm shift within the DOJ from prosecution to prevention,” which we would argue maps directly to a shift from avoiding Type I to avoiding Type II errors (US Department of Justice, Office of the Inspector General 2009, 11). As former CIA Director Hayden stated, “I was in government for ten years after 9/11, and let me tell ya, a phrase I never heard from anybody in a position of authority: ‘Whatever you guys do about this terrorism threat, please, please don’t overreact’” (Kelly 2014). In other words, have no fear about committing Type I errors. Other organizations and leaders similarly shifted their perspective on intelligence errors. US National Security Advisor (and later Secretary of State) Rice stated, “Unless you were there, in a position of responsibility after September 11th, you cannot possibly imagine the dilemmas that you faced in trying to protect Americans…you were determined to do anything that you could that was legal to prevent that from happening again” (Kessler 2009). The National Intelligence Director on 9/11, Dennis Blair, echoed this perspective: “Those methods, read on a bright, sunny, safe day in April 2009, appear graphic and disturbing. But we will absolutely defend those who relied on these memos [supporting torture]” (Buckley 2009). Although the Bush Administration’s support for employing torture in interrogations following 9/11 has been widely described (CIA 2004; Department of Justice 2008; US Senate 2008), we think that an important aspect of the story is missing. Specifically, following 9/11, the Bush Administration became extremely wary of committing a Type II error—failing to take actions that might prevent additional attacks on the United States and American citizens—and thus became more willing to commit Type I errors (Suskind 2006; Gartner 2013). As President Bush’s initial appointee to lead the 9/11 inspired Homeland Security Department stated, “The criminal justice system is about prosecution and counter-terrorism is about prevention” (Ridge 2009). The goal of prevention is to avoid Type II errors—avoid failing to take actions that would have prevented an attack. Prevention requires acting before a crime is committed, thus opening up the door for the commission of Type I errors and torture: “On February 7, 2002, President Bush signed a memorandum stating that the Third Geneva Convention did not apply to the conflict with al Qaeda and concluding that Taliban detainees were not entitled to prisoner of war status or the legal protections afforded by the Third Geneva Convention” (US Senate xiii). The use of torture was national policy, implemented not by “a few bad apples” as was often repeated after Abu Ghraib, but by leaders fearing making Type II mistakes. “Interrogation policies endorsed by senior military and civilian officials authorizing the use of harsh interrogation techniques were a major cause of the abuse of detainees in US custody” (US Senate xxv). Despite recognizing the low quality of the information gained through its use, American officials, profoundly fearful of committing another Type II error, systematically endorsed the practice of using torture to obtain counterinsurgency information. Finally, it is important to recognize that, despite the national hurt inflicted by 9/11, the American public did not support torture. “Not once during the eight years of the Bush administration was there an American majority in favor of the use of torture” (Gronke et al. 2010, 441). Rather than public opinion, the motivation was an informational logic, driven by fear of committing a Type II error, which made the data error rate from torture acceptable to US leaders and the information it provided appear useful. 9/11—Obama Focusing on error type leads to a different way of considering the change in torture policy implemented by the Obama Administration. In 2009, the President issued an Executive Order requiring personnel to follow the US Army Manual on interrogations, which explicitly prohibits torture (Semel 2013). Although Obama Administration levels of coercive interrogation and actions such as the extraordinary rendition program still exceed that of pre-9/11 (especially at “black site” prisons), they clearly represent a change in US actions from the Bush Administration. The standard explanation for the change in policy is that the political costs of violating the norm became too large, so the administration changed policy. “Torture was only ended (if indeed it has ended) by the public outcry after the revelation of torture by U.S. soldiers, as well as the leaking of the Red Cross reports to the media, made the costs to elites too high to bear” (Armshaw 2011, 137). Our approach confirms that a shift in the perceived cost of violating torture norms will inhibit its use, but this inhibition will only be reinforced if the administration’s assessment of errors becomes less biased toward avoidance of Type II errors. As a result, the threshold for information reliability beyond which a state’s calculus favors torture dramatically increases. Figure 3 illustrates a possible Bush–Obama shift in t and c and the implications for thresholds on information reliability. We assume for both administrations that d=0.1, E=2.5, and they anticipate that 80 percent of prisoners will talk. Figure 3. View largeDownload slide Possible shifts in costs and benefits from Bush to Obama Figure 3. View largeDownload slide Possible shifts in costs and benefits from Bush to Obama Using (3), we can then identify the minimum level of reliability that each administration requires given its evaluation of the avoidance of a Type II error as well as the costs of torture. For example, given the assumed priorities of the Bush Administration, minimum information reliability must satisfy cBush=2.5*0.8*[tBush-xtBush+0.1] or 1-x=cBush+0.22tBush+0.2. Clearly, the constraint on information reliability increases under the Obama Administration, and this favors foregoing the use of torture in the interrogation of terror suspects. With reference to Figure 3, given anticipated reliability of information, the combination of costs and valuation of Type II error avoidance for the Bush Administration falls in the shaded area in which the state will torture, whereas those same costs and valuations for the Obama Administration fall outside of the torture zone. Assessments of the ineffectiveness of torture to produce timely and important results made it easier for the administration to oppose its use as far more reliable information would have been needed to justify the practice. In terms of the model, minimum required reliability CObama+0.22tObama+0.2 exceeds anticipated reliability of the information received under torture. As a result, one of the four “central themes” in the US Senate’s recent report on the CIA’s enhanced interrogation techniques following 9/11 is titled, “The CIA’s ‘enhanced entorrogation techniques’ were not effective and did not prodcue any ‘ticking time bomb’ information crucial to saving lives” (Kelly 2014). The War on Terror builds on the fear of committing an error of omission. Our analysis contributes to explaining the widespread alleged use of torture since 9/11. Turkey The 1980 military coup led to a rise in the use of torture by the Turkish government. The magnitude and severity of torture at the hands of the military government in the early 1980s has been referred to as “the period of barbarity” (Zeydanlıoğlu 2009, 7). Zeydanlıoğlu in his characterization is referring to the experiences of prisoners in Diyarbakır Military Prison No. 5 in Turkey’s Kurdish region in the early to mid-1980s. The return to a civilian government at the end of 1983 and the restoration of a multi-party political system did not end the dispute with the Turkish Kurdish population, a dispute that has long cultural, historical, and even linguistic roots (Fernandes 2012). Martial law remained in place in provinces in the South-East, where the Turkish government continued its use of torture in its efforts to suppress Kurdish secession attempts. The high incidence of reported torture cases continued into the first half of the 1990s. A 1998 European Union (EU) progress report noted that “persistent cases of torture, disappearances and extra-judicial executions are regularly recorded despite repeated official statements of the government’s commitment to ending such practices” (European Commission 1998). The use of torture began to decrease in 2002. As shown in the ITT database, Turkey’s use of torture decreased from the highest level “Systematic/level 5” from 1995 to 2002 to “Routinely/level 4” in 2003 and “Widespread/level 3” in 2004 (Conrad and Moore 2011). In 2001, the government began to make progress in protecting the population’s human rights and decreasing the practice of torture as observed by both the European Commission (2002) and Amnesty International (2001). A 2004 Human Rights Watch notes that “Turkish legal protections for detainees are better than in many EU member states.” In 2005, the Turkish government amended provisions in the constitution to improve the legal framework regulating torture, which criminalized torture and reduced the maximum length of detention from 90 days in 1980 to 24 hours (Human Rights Watch 2006). In 2012, the EU reported that “overall, there has been a downward trend in torture and ill-treatment in places of detention” (European Commission 2012; see also UN Manual on the Effective Investigation and Documentation of Torture—“The Istanbul Protocol”). What motivated this change? One factor that did not change is membership in anti-torture conventions and legal prohibitions against torture. Turkey is a signatory to the United Nations and European Conventions against Torture and Other Cruel, Inhuman, or Degrading Treatment, signed and ratified in 1988. The Turkish constitution explicitly forbids the use of torture (Article 17, 1982). In addition, torture occurred under both civilian and military rule (although not at equal levels). The changes that did occur were a reduction in the threat posed by the Kurdish terrorist group, the PKK (and the cost of Type II errors), and an increase in the government’s assessment of the costs of Type I errors (and their potential impact on the likelihood of EU admission). In 1998, the PKK threat was “escalating” (Bruinessen 1998, cited in Kocher 2007, 1), supporting fear of Type II errors. The security situation changed quickly and dramatically. “As of 2002, the Kurdish Southeast of Turkey is largely pacified. PKK guerrillas have withdrawn across the Iraqi border” (Kocher 2007, 2). The result is that, “for all intents and purposes, PKK is no longer a military presence in Turkey” (Kocher 2007, 2). The dramatic reduction in threat posed by the PKK made Type II errors more acceptable. At the same time, the EU put specific conditions on human rights protections in Turkey as a condition for admission (Schimmelfennig et al. 2003). “The EU expected Turkey to…abolish the death penalty and torture” (Bayer and Onis 2010, 187). This high level of scrutiny put a premium on avoiding Type I errors. Thus, the decreasing nature of the security threat posed by the PKK and the desire for Turkey to enter the EU changed the incentives for error types and altered the decision to torture prisoners. As a result of changes such as adopting a revised penal code that criminalized torture, in 2004 “the European Council agreed to open accession negotiations with Turkey on the grounds it had sufficiently fulfilled the Copenhagen criteria” (Bayer and Onis 2010, 189). In terms of our model, parameter t decreased, whereas parameter d increased as the threat represented by the PKK was diminished. At the same time, the costs of violating torture norms increased. Graphically, the situation is similar to that of Figure 3: Information obtained under torture must now be far more reliable to justify the practice. More recently, however, the threat of Kurdish revolt has increased the heavy handedness of Turkish security forces, whereas the economic crisis in the EU has made membership less attractive. The result is that Turkey is increasingly accepting of Type I and fearful of Type II security errors as evidenced by a powerful new domestic security law that entered force in April of 2015 (MENAFN 2015). The two vignettes demonstrate that variation in the incentives for error type can alter preferences for unreliable information. Both vignettes also show that torture policy varies—and thus requires explanations that also vary. Conclusion Why might those who agree on the ethical imperative of anti-torture policies disagree on the utility of torture? We think the two sides talk past each other; each focuses on minimizing a different type of error. Those motivated by the ticking bomb scenario want to do everything possible to avoid a horrific Type II error—failing to stop an event that will have deadly consequences. Those who see torture as ineffective point out the Type I errors committed by acting on the bad data resulting from torturing captives. The debate then is really about the utility of Type I versus Type II errors and not about the morality of torture. By not addressing the error type issue, the two sides fail to engage the key issue driving their disagreement. Prisoners must also make a choice. Interrogated prisoners, who want the state to respond with an imbalanced “over the top” reaction, need to provide enough true information in the mix with lies as incentive for the state to act. This creates a strategic question: How little truth can the states anticipate from prisoners and still employ torture? This question links information reliability, error types, and the choice to torture despite the cost. Because our approach is not regime-type specific, we can help explain variation in the use of torture between and within regime types, which is critical given the increasing practice among democracies. Although the number of democratic states increased in the last quarter of the twentieth century, the practice of torture also increased. “During the final quarter of the twentieth century, freedom from torture was the most widely violated human right to the physical integrity of the person” (Conrad and Moore 2010, 459; although there may have also been changes in monitoring standards, Fariss 2014). Our approach represents a general process that is not unique to the calculus of torture; the treatment of information in many situations depends on the balance between fears of making errors of commission and omission. For example, our approach helps to explain why the media, politicians, firms, and individuals react to non-random surveys, such as Internet polls that are recognized as unrepresentative and highly erroneous. If these surveys address an unbalanced situation where the cost of a Type II error dramatically outweighs a Type I error, our model can identify the conditions in which this information—even if recognized as inaccurate—leads to action. More generally, the model can be adapted to fit a wide variety of topics in which the key dynamic is an excessive fear of making a Type II error. Does this study provide a rational justification for torture? Absolutely not! We provide an examination of the decision-making structure underpinning choices over torture. Our objective is to make opaque arguments more transparent so they can be more clearly viewed and more effectively evaluated. Just like studying war does not make one a hawk (Gartner 1998), systematically studying torture does not make us torture champions (Wantchekon and Healy 1999; Schiemann 2012). Only by understanding the rational dynamics of torture can we appreciate why states value it. In fact, understanding these underpinnings might help torture opponents to have greater influence. Opponents of torture diligently advocate anti-torture norms and treaties. But these actions have seemingly failed to stymie the practice: the “spreading of an international norm against torture has, according to our findings, not limited the practice of torture to any extent” (Gilligan and Nesbitt 2009, 467–95). Our argument suggests that these appearances might be deceptive; a state can have a high cost for Type I errors, a relatively low benefit for avoiding Type II errors, and still employ torture if the cost of violating the norm is low enough. Steps taken to enhance the norm might thus help to reduce torture, even if this cannot be observed directly. At the same time, if efforts to promote anti-torture policies occur simultaneously with dramatic increases in the preference for avoiding Type II intelligence errors, states might practice torture despite embracing anti-torture norms. For example, one reason anti-torture norms appear ineffective may be that after 9/11 states across the globe raised their estimates on the costs of Type II errors, thus leading to an increase in the practice of torture and a greater willingness to accept more unreliable information. Given the vivid fears generated by terrorism, those attempting to decrease the practice of torture might be more effective trying to lower the costs of Type II and raise the costs of Type I intelligence errors, rather than continuing to champion already widely accepted anti-torture norms. Footnotes 1 Adopted by the General Assembly December 1984 and entered force in June 1987. Signed (1988) and ratified (1944) by the United States. US reservations included: “in order to constitute torture, an act must be specifically intended to inflict severe physical or mental pain or suffering” (United Nations 1984). 2 Although the US Supreme Court found in Hamdan v. Rumsfeld that Common Article III applied to the US War on Terror, the Bush Administration developed ways to work around this prohibition (Vázquez 2007). 3 Error type is different from prospect theory, which argues that how an error is framed influences risk propensity (Kahneman and Tversky 1979). For a critique of applying prospect theory to torture, see Johnson and Ryan (2015). References Adler Emanuel. 2010. “ Damned if You Do, Damned if You Don’t: Performative Power and the Strategy of Conventional and Nuclear Defusing.” Security Studies  19: 199– 229. Google Scholar CrossRef Search ADS   Alexander Matthew. 2009. “ Torture Doesn’t Work. The Daily Beast.” Accessed May 2, 2014. http://www.thedailybeast.com/articles/2009/04/20/torture-doesnt-work.html#. Amnesty International. 2001. “ Turkey: An End to Torture and Impunity is Overdue.” Accessed April 06, 2016. https://www.amnesty.org/en/documents/eur44/026/2002/en/. Armshaw Patrick. 2011. The Torturer's Dilemma: Analyzing the Logic of Torture for Information. Diss., Tallahassee, Florida: Florida State University. Avdan Nazli. 2012. “ Human Trafficking and Migration Control Policy: Vicious or Virtuous Cycle?” Journal of Public Policy  32: 171– 205. Google Scholar CrossRef Search ADS   Avdan Nazli. 2013. “ Controlling Access to Territory Economic Interdependence, Transnational Terrorism, and Visa Policies.” Journal of Conflict Resolution  58: 592– 624. Google Scholar CrossRef Search ADS   Avruch Kevin. 2003. “ Type I and Type II Errors in Culturally Sensitive Conflict Resolution Practice.” Conflict Resolution Quarterly  20: 351– 71. Google Scholar CrossRef Search ADS   Babad Elisha Michael Hills, O’driscoll Michael. 1992. “ Factors Influencing Wishful Thinking and Predictions of Election Outcomes.” Basic and Applied Social Psychology  13: 471. Google Scholar CrossRef Search ADS   Bagaric Mirko, Clarke Julie. 2005. “ Not Enough (Official) Torture in the World? The Circumstances in Which Torture Is Morally Justifiable.” University of San Francisco Law Review  581– 616. Baldor Lolita C., Benac Nancy. 2014. “ U.S. Prepares for Trial of Captured Benghazi Attack Suspect.” PBS . Accessed August 13, 2014. http://www.pbs.org/newshour/rundown/u-s-prepares-trial-captured-benghazi-attack-suspect/. Bar-Joseph Uri, Kruglanski Arie W.. 2004. “ Intelligence Failure and Need for Cognitive Closure: On the Psychology of the Yom Kippur Surprise.” Political Psychology  24: 75– 99. Google Scholar CrossRef Search ADS   Bar-Joseph Uri, McDermott Rose. 2008. “ Change the Analyst and Not the System: A Different Approach to Intelligence Reform.” Foreign Policy Analysis  4: 127– 45. Google Scholar CrossRef Search ADS   Bayer Resat, Onis Ziya. 2010. “ Turkish Big Business in the Age of Democratic Consolidation: The Nature of Limits of Its Influence.” South European Society and Politics  15: 181– 201. Google Scholar CrossRef Search ADS   Betts Richard K. 1978 . “Why Intelligence Failures Are Inevitable.” World Politics  31: 61– 89. Google Scholar CrossRef Search ADS   Blanken Leo J. 2012. Rational Empires: Institutional Incentives and Imperial Expansion . Chicago, IL: University of Chicago Press. Google Scholar CrossRef Search ADS   Blanken Leo J., Overbaugh Justin. 2012. “ Looking for Intel?…Or Looking for Answers? Reforming Military Intelligence for a Counterinsurgency Environment.” Intelligence and National Security  27: 559– 75. Google Scholar CrossRef Search ADS   Buckley Christopher. 2009. “ Enough with Torture Sanctimony.” The Daily Beast . Accessed April 08, 2016. http://www.thedailybeast.com/articles/2009/04/19/enough-with-the-torture-sanctimony.html#. Bueno de Mesquita Bruce, Alastair Smith, Randolph M. Siverson, James D. Morrow 2003. The Logic of Political Survival . Cambridge: MIT Press. Cardina Miguel. 2013. “ To Talk of Not to Talk: Silence, Torture and Politics in the Portuguese Dictatorship of Estadu Novo.” Oral History Review  40: 251– 70. Google Scholar CrossRef Search ADS   Carpenter Daniel P. 2004 . “The Political Economy of FDA Drug Review: Processing, Politics, and Lessons for Policy.” Health Affairs  23: 52– 63. Google Scholar CrossRef Search ADS PubMed  Central Intelligence Agency. Inspector General (CIA). 2004. “ Counterterrorism Detention and Interrogation Activities (September 2001–October 2003).” The New York Times . Accessed July 18, 2013. http://graphics8.nytimes.com/packages/pdf/politics/20090825-DETAIN/2004CIAIG.pdf. Clarke Richard A. 2009. “ Cheney and Rice Remember 9/11. I Do, Too.” The Washington Post Opinions . Accessed May 2, 2014. http://www.washingtonpost.com/wp-dyn/content/article/2009/05/29/AR2009052901560.html. Conrad Courtney Ryals, Moore Will H.. 2010. “ What Stops Torture.” American Journal of Political Science  54: 459– 76. Google Scholar CrossRef Search ADS   Conrad Courtney Ryals, Moore Will H.. 2011. “ Country-Year Data Users Guide.” The Ill-Treatment & Torture (ITT) Data Collection Project . Accessed February 14, 2014. http://faculty.ucmerced.edu/cconrad2/Academic/Data_files/ITT_CY_UsersGuide19July11.pdf/. Danelius Hans. 2008. “ Convention Against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment.” United National Audiovisual Library of International Law , 1– 4. Accessed April 06, 2016. http://legal.un.org/avl/pdf/ha/catcidtp/catcidtp_e.pdf. Dershowitz Alan M. 2003 . “Torture Warrant: A Response to Professor Strauss.” New York Law School Law Review  48: 275. Dragu Tiberiu. 2011. “ Is There a Trade-Off Between Security and Liberty? Executive Bias, Privacy Protections, and Terrorism Prevention.” American Political Science Review  105: 64– 78. Google Scholar CrossRef Search ADS   European Commission. 1998. “ Regular Report on Turkey’s Progress towards Accession.” Accessed April 06, 2016. http://ec.europa.eu/enlargement/archives/pdf/key_documents/1998/turkey_en.pdf. European Commission. 2002. “ Regular Report on Turkeys Progress Towards Accession.” Accessed April 07, 2016. http://ec.europa.eu/enlargement/archives/pdf/key_documents/2002/tu_en.pdf. European Commission. 2012. “ Turkey Progress Report.” Accessed April 06, 2016. http://ec.europa.eu/enlargement/pdf/key_documents/2012/package/tr_rapport_2012_en.pdf. Fariss Christopher J. 2014. “ Respect for Human Rights Has Improved Over Time: Modeling the Changing Standard of Accountability.” American Political Science Review  108: 297– 318. Google Scholar CrossRef Search ADS   Feinstein Diane. 2014. United States Senate Committee on Intelligence. “Committee Study of the Central Intellgence Agency’s Detenction and Interrogation Program.” Accessed April 06, 2016. https://web.archive.org/web/20141209165504/http://www.intelligence.senate.gov/study2014/sscistudy1.pdf. Fernandes Desmond. 2012. “ Modernity and the linguistic genocide of Kurds in Turkey.” International Journal of the Sociology of Language  217: 75– 98. Finn Peter. 2008. “ Retired Officers Urge Obama to End Harsh Interrogations. Houston Chronicle.” Accessed May 2, 2014. http://www.chron.com/news/nation-world/article/Retired-officers-urge-Obama-to-end-harsh-1775419.php. Foley Conor. 2003. Combating Torture: A Manual for Judges and Prosecutors . Human Rights Centre, University of Essex, Colchester, United Kingdom. Forsythe David, P. 2006. “ United States Policy Towards Enemy Detainees in the War on Terror.” Human Rights Quarterly  28: 465– 91. Google Scholar CrossRef Search ADS   Gartner Scott Sigmund. 1998. “ Opening up the Black Box of War.” The Journal of Conflict Resolution  42. Gartner Scott Sigmund. 2013. “ All Mistakes Are Not Equal: Intelligence Errors & National Security.” Intelligence and National Security  28: 634– 54. Google Scholar CrossRef Search ADS   Gartner Scott Sigmund, Regan Patrick M.. 1996. “ Threat and Repression: The Non-linear Relationship Between Government and Opposition Violence.” The Journal of Peace Research  33: 273– 88. Google Scholar CrossRef Search ADS   Gilligan Michael J., Nesbitt Nathaniel H.. 2009. “ Do Norms Reduce Torture?” The Journal of Legal Studies  38: 445– 70. Google Scholar CrossRef Search ADS   Gronke Paul, Darius Rejali, Dustin Drenguis, James Hicks, Peter Miller, Bryan Nakayama. 2010. “ US Public Opinion on Torture, 2001–2009.” PS: Political Science and Politics  43: 437– 44. Google Scholar CrossRef Search ADS   Hanson R. 2010. “ Torture Kids Instead.” Retrieved July 8, 2013. Overcoming Bias. http://www.overcomingbias.com/2010/01/torture-kids-instead.html. Harnden Toby. 2014. “ Obama’s Mission Impossible in Irak.” The Sunday Times , August 24, 20– 21. Harris Gardiner. 2004. “ Lawmaker Says F.D.A Held Back Drug Data.” New York Times . Accessed January 28, 2009. http://nytimes.com/2004/09/10/politics/10trial.html. Hayden Michael, Mukasey Michael B.. 2009. “ The President Ties His Own Hands on Terror: The Point of Interrogation is Intelligence, Not Confession.” The Wall Street Journal . Accessed May 2, 2014. http://online.wsj.com/news/articles/SB123993446103128041?mg=reno64-wsj&url=http%3A%2F%2Fonline.wsj.com%2Farticle%2FSB123993446103128041.html. Human Rights Watch. 2004. “ Turkey: EU Bid Hinges on Further Rights Reforms” (14 June, 2004). Accessed April 06, 2016. https://www.hrw.org/news/2004/06/14/turkey-eu-bid-hinges-further-rights-reforms. Human Rights Watch. 2006. “ Turkey: First Steps Toward Independent Monitoring of Police Stations and Gendarmeries.” Briefing Paper No. 1. Accessed April 06, 2016. https://www.hrw.org/legacy/backgrounder/eca/turkey0306/turkey0306.pdf. Intriligator Michael D. 1996. “ Drug Evaluations: Type I vs. Type II Errors.” UCLA Research Program in Pharmaceutical Economics and Policy Paper, 96–2. Accessed April 06, 2016. http://escholarship.org/uc/item/5fg9n284. Irvine David R. General. 2005. “ Why Torture Doesn’t Work.” Accessed April 06, 2016. www.alternet.org/story/28585/why_torture_doesn%27t_work. Joint Personnel Recovery Agency. 2002. Operational Issues Pertaining to the Use of Physical/Psychological Coercion in Interrogation, 1–2. Accessed April 06, 2016. http://www.washingtonpost.com/wp-srv/nation/pdf/JPRA-Memo_042409.pdf?sid=ST2009042403231. Johnson David Blake, Ryan John Barry. 2015. “ The Interrogation Game: Using Coercion and Rewards to Elicit Information from Groups.” Journal of Peace Research  52: 822– 37. Google Scholar CrossRef Search ADS   Kahneman Daniel, Tversky Amos. 1979. “ Prospect Theory, an Analysis of Decision Making under Risk.” Econometrica  47: 263– 91. Google Scholar CrossRef Search ADS   Kam Ephraim. 1988. Surprise Attack.  Cambridge, MA: Harvard University Press. Google Scholar CrossRef Search ADS   Keller Bill. 1987. “ Moscow Dismisses More Air Generals.” New York Times . Accessed April 06, 2016. http://www.nytimes.com/1987/06/18/world/moscow-dismisses-more-air-generals.html. Kelly Erin. 2014. “ Senate Report Reopens Battle Over CIA Interrogations.” USA Today . Accessed September 12, 2014. http://www.usatoday.com/story/news/politics/2014/12/09/cia-torture-report-senate-intelligence-feinstein/20134093/. Kelly Christopher E., Miller Jeaneé C., Redlich Allison D., Kleinman Steven M.. 2014. “ Taxonomy of Interrogation Methods.” Psychology, Public Policy and Law  19: 165– 75. Accessed July 2, 2014. http://www.lexisnexis.com.libproxy.nps.edu/lnacui2api/frame.do?tokenKey=rsh-20.749457. Google Scholar CrossRef Search ADS   Kessler Glenn. 2009. “ Rice Defense Enhanced Interrogations.” The Washington Post . Accessed May 2, 2014. http://voices.washingtonpost.com/44/2009/04/30/rice_defends_enhanced_interrog.html. Kocher Matthew. 2007 . “The Decline of PKK and the Viability of a One-State Solution in Turkey.” International Journal on Multicultural Societies  4: 126– 47. Koppl Roger. 2006. “ Epistemic Systems.” Episteme  2: 91– 106. Google Scholar CrossRef Search ADS   Lampel Joseph, Shapira Zur. 2001. “ Judgmental Errors, Interactive Norms, and the Difficulty of Detecting Surprises.” Organizational Science  12: 599– 611. Google Scholar CrossRef Search ADS   Liberman Peter. 2013. “ Retributive Support for International Punishment and Torture.” Journal of Conflict Resolution  57: 285– 306. Google Scholar CrossRef Search ADS   Luban David. 2007. “ Liberalism, Torture, and the Ticking Bomb.” In Intervention, Terrorism, and Torture , Lee Steven P. (ed.), 249– 62. Springer. Malinowski Tom. 2008. “ Restoring Moral Authority: Ending Torture, Secret Detention and the Prison at Guantanamo Bay.” Annals of the Academy of Political and Social Science  618: 148– 52. Google Scholar CrossRef Search ADS   Menafn. 2015. “ Turkey’s Domestic Security Law Enters Into Force.” The Journal of Turkish Weekly . Accessed April 06, 2016. http://www.menafn.com/1094151409/Turkeys-domestic-security-law-enters-into-force. Mercer Jonathan. 2005. “ Rationality and Psychology in International Politics.” International Organization  59: 77– 106. Google Scholar CrossRef Search ADS   Morgan Rod. 2000. “ The Utilitarian Justification for Torture: Denial, Desert, and Disinformation.” Punishment & Society  2: 181– 96. Google Scholar CrossRef Search ADS   Mulrine Anna, 2014. “ Why the US Special Forces Failed to Rescue James Foley.” The Christian Science Monitor . Accessed April 06, 2016. http://www.csmonitor.com/USA/Military/2014/0821/Why-US-special-forces-failed-to-rescue-James-Foley. O’Mara Shane. 2009. “ Torturing the Brain: On the Folk Psychology and Folk Neurobiolgy Motivating ‘Enhanced and Coercive Interrogation Techniques.’” Trends in Cognitive Sciences  13: 497– 500. Google Scholar CrossRef Search ADS PubMed  Persson Mats, Siven Claes-Henric. 2007. “ The Becker Paradox and Type I versus Type II Errors in the Economics of Crime.” International Economic Review  48: 211– 33. Google Scholar CrossRef Search ADS   Phillips Joshua E. S. 2012. None of US Were Like This Before: American Soldiers and Torture . London, UK and Brooklyn, NY, USA: Verso Books. Pollard P., Richardson J. T. E.. 1987. “ On the Probability of Making Type-I Errors.” Psychological Bulletin  102: 159– 63. Google Scholar CrossRef Search ADS   Rejali Darius. 2004. “ Torture’s Dark Allure.” Salon . July 20, 2013. http://www.salon.com/2004/06/18/torture_1. Rejali Darius. 2007. “ 5 Myths about Torture and Truth.” Washington Post Opinions . Accessed April 06, 2016. http://www.salon.com/2004/06/18/torture_1/. Rejali Darius. 2009. Torture and Democracy . Princeton, NJ: Princeton University Press. Ridge Tom. 2009. “ We Were Wrong to Torture.” BBC . Accessed May 2, 2014. http://news.bbc.co.uk/2/hi/americas/7903516.stm. Roberts J. D. 2013. Error Avoidance Theory: Sniper Employment for Military and Civilian Law Enforcement. Doctoral diss., Naval Postgraduate School, Monterey, CA. Rose David. 2008. Tortured Reasoning. Vanity Fair . Accessed July 11, 2013. http://www.vanityfair.com/magazine/2008/12/torture200812?currentPage=2. Schiemann John W. 2012. “ Interrogational Torture: Or How Good Guys Get Bad Information with Ugly Methods.” Political Research Quarterly  3– 19. Schimmelfennig Frank, Engert Stefan, Knobel Heiko. 2003. “ Costs, Commitment and Compliance: The Impact of EU Democratic Conditionality on Latvia, Slovakia and Turkey.” JCMS: Journal of Common Market Studies  41: 495– 518. Google Scholar CrossRef Search ADS   Semel Matthew D. 2013. “ Military Interrogations: Best Practices and Beliefs.” Perspectives on Terrorism  7: 39– 62. Shue Henry. 1978. “ Torture.” Philosophy & Public Affairs  7: 124– 43. Silver James. 2004. “ Why America’s Top Liberal Lawyer Wants to Legalise Torture. The Scotsman.” Accessed August 7, 2013. http://www.scotsman.com/news/world/why-america-s-top-liberal-lawyer-wants-to-legalise-torture-1-532091. Sobel Russell S., Leeson Peter T.. 2006. “ Government’s Response to Hurricane Katrina: A Public Choice Analysis.” Public Choice  127: 55– 73. Google Scholar CrossRef Search ADS   Soufan Ali H. 2009a. “ My Tortured Decision.” New York Times . Accessed July 26, 2013. http://www.nytimes.com/2010/09/26/opinion/terr-soufan.html?gwh=2EB875DFA290006C9D2B5BDD3E7B1589. Soufan Ali H.. 2009b. “ What Torture Never Told US.” New York Times . Accessed July 26, 2013. http://www.nytimes.com/2009/09/06/opinion/06soufan.html. Suskind R. 2006. One Percent Doctrine: Deep Inside America's Pursuit of Its Enemies Since 9/11 . New York: Simon and Schuster. Sussman David. 2005. “ What’s Wrong with Torture?” Philosophy & Public Affairs  33: 1– 33. Google Scholar CrossRef Search ADS   US Army. 1992. “ Intelligence Interrogation.” In US Army Field Manual , 34– 52. Washington, DC: Army Publishing Directorate. PubMed PubMed  US Department of Justice, Office of the Inspector General. 2009. A Review of the FBI’s Involvement in and Observations of Detainee Interrogations in Guantanamo Bay, Afghanistan, and Iraq . Washington, DC: Department of Justice. US Senate. 2008. Inquiry Into the Treatment of Detainees in US Custody . United States Senate. Washington, DC: Committee on Armed Services. United Nations. 2004. “ Manual on the Effective Investigation and Documentation of Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment (‘The Istanbul Protocol’).” Accessed April 06, 2016. http://www.ohchr.org/documents/publications/training8rev1en.pdf. United Nations General Assembly. 1984. “ Convention against Torture and Other Cruel, Inhuman or 15 Degrading Treatment or Punishment.” A/RES/39/46, Meeting No. 93 . Accessed April 08, 2016. https://treaties.un.org/Pages/ViewDetails.aspx?src=IND&mtdsg_no=IV-9&chapter=4&lang=en. Vázquez Carlos Manuel. 2007. “ The Military Commissions Act, the Geneva Conventions, and the Courts: A Critical Guide.” The American Journal of International Law  101: 73– 98. Walzer Michael. 1973. “ The Problem of Dirty Hands.” Philosophy and Public Affairs  2: 160– 80. Wantchekon Leonard, Healy Andrew. 1999. “ The ‘Game’ of Torture.” The Journal of Conflict Resolution  43: 596– 609. Google Scholar CrossRef Search ADS   Weiser Benjamin. 2014. “ Some Captured Terrorists Talk Willingly and Proudly Interrogators Say.” New York Times . Accessed April 06, 2016. http://www.nytimes.com/2014/10/14/nyregion/some-captured-terrorists-talk-willingly-and-proudly-investigators-say.html?_r=0. Wirtz James J. 1989. “ The Intelligence Paradigm.” Intelligence and National Security  4: 829– 37. Google Scholar CrossRef Search ADS   Yingling Raymund T., Ginnane Robert W.. 1952. “ The Geneva Conventions of 1949.” The American Journal of International Law  46: 393– 427. Google Scholar CrossRef Search ADS   Zeydanlıoğlu Welat. 2009. “ Torture and Turkification in the Diyarbakır Military Prison.” In, Rights, Citizenship & Torture: Perspectives on Evil, Law and the State . Zeydanlıoğlu Welat and Parry John T. (eds.), 73– 92. Oxford: Inter-Disciplinary Press. Zhao Bin, Olivera Fernando. 2006. “ Error Reporting in Organizations.” Academy of Management Review  31: 1012– 30. Google Scholar CrossRef Search ADS   © The Author (2016). Published by Oxford University Press on behalf of the International Studies Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Journal

Foreign Policy AnalysisOxford University Press

Published: Jan 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

Monthly Plan

  • Read unlimited articles
  • Personalized recommendations
  • No expiration
  • Print 20 pages per month
  • 20% off on PDF purchases
  • Organize your research
  • Get updates on your journals and topic searches

$49/month

Start Free Trial

14-day Free Trial

Best Deal — 39% off

Annual Plan

  • All the features of the Professional Plan, but for 39% off!
  • Billed annually
  • No expiration
  • For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588

$360/year

billed annually
Start Free Trial

14-day Free Trial