Why "Does Torture Work?" Is a Bad Question

The debate about the "effectiveness" of torture is in the news again. The CIA released a trove a documents about its past treatment of detainees. Dick Cheney claims these show that torture (oops, I mean "harsh interrogation techniques") provided the US authorities with valuable counterterrorism information. This Washington Post piece concludes that waterboarding led one big shot Al Qaeda detainee to spill important secrets.

This debate about the effectiveness of torture has been conducted in overly narrow terms. Those that think it is effective define effectiveness as when detainees give up information that help US counterterrorism efforts. Opponents of torture say that this is unlikely to happen, since detainees lie under duress, that other techniques are quicker and more effective, and so on. Opponents also argue that torture is simply bad, and we should never condone its' use.

Both sides in this debate, but especially the former, ignore the much more important second-order effects of torture. The biggest material cost to the US of torture is the impression it creates in foreign and domestic audiences. If the US government tortures, and the US public finds out about it, a lot of us are going to be upset because we don't like torture. And this is going to create problems for the US government. Look at how bogged down the late Bush administration got in tangles over torture, rendition, etc. It had to expend a lot of effort putting out these torture fires, effort that it could not expend on things like, say, counterterrorism. The Obama administration, which literally outlawed torture on the day it took office, is also tangled up in figuring up what to do with those who tortured, or authorized torture, in the past.

Foreign audiences matter too. Foreign governments may be less keen to cooperate with the US if it has been accused of torturing--this helps explain why Obama is having trouble offloading Gitmo detainees on the rest of the world. Torture in places like Afghanistan and Iraq can alienate the locals from the US and its' guilty-by-association Iraqi and Afghan allies, making them less willing to provide intelligence on terrorists and insurgents. And torture by the US is great propoganda for terrorists--they can use it to demolish American claims of moral superiority, and in effect argue that the US is just as much a terrorist as they are.

Thought about in these terms, it's unlikely that we could conclude that torture is effective. Even if the proponents of torture are correct, and it does provide actionable intelligence, we must balance this against the cost to the US reputation. This cost has been very large since 9/11 in terms of tying domestic politics in knots, alienating potential allies, and providing support to enemies. It seems very unlikely that the benefits of the literally handful of known cases in which a detainee *may* have provided some useful leads could outweigh these costs.

Rendition: Can We Make It Kindler and Gentler?

The New York Times reports that

"The Obama administration will continue the Bush administration’s practice of sending terrorism suspects to third countries for detention and interrogation, but pledges to closely monitor their treatment to ensure that they are not tortured, administration officials said Monday."

That's nice. But how do we "ensure that they are not tortured"? That is trickier. Past practice relied on "diplomatic assurances" from other states that they would not torture. States like Syria. So that didn't work very well.

The US also relied on having American officials visit the detainees after they have been moved to a new country. This is promising, since the US could directly monitor the condition of the detainee. But it turns out not to work too well. According to the Times, at least one prisoner was visited many times (by Canadian diplomats) but was to afraid to tell them he had been tortured. And we know from Darius Rejali's work on the topic that torturers have available to them a huge range of cheap and effective means of inflicting suffering while leaving few marks or other physical evidence of abuse.

Now officials are promising that they will not, according to the Times, send detainees to " countries known to conduct abusive interrogations." Like what countries? According to a recent Amnesty International report, a majority of countries in the world practice engage in torture. So that limits the potential recipients to nice democratic countries in Europe, for the most part. But these countries may not want to detainees.

And let's face it, the US does not really want to send its detainees to nice European countries. The basic purpose of extraordinary rendition is the send detainees countries that will mistreat them (see Charli Carpenter's post on this point). There is another, long-established legal process--extradition--for sending prisoners across borders. The modern US practice of rendition was invented (by the Clinton administration in the 1990s) as a way to get around this process. Why bother? Well, judges have to approve extraditions but not renditions. And the Convention Against Torture binds the US from sending detainees to countries that torture. Judges seem more likely to take this commitment seriously, and to refuse to extradite detainees if they might be tortured.

This does suggest a way for the Obama administration to achieve its objective of not sending detainees to be tortured: stop the policy of rendition. Instead, simply extradite the prisoners. No doubt national security officials would object that this would greatly delay the process. If this is the case, another solution might be to require the approval of a US court before a detainee is subject to rendition. This might require new legislation, and perhaps the creation of a special "fast track" procedure so that such cases are decided quickly (but not too quickly; after all the judge would need time to hear evidence about the torture practices of the receiving country and other matters). Such a change would still represent an end run around extradition, but it would at least incorporate the independent assessment of judicial authority that has an obligation to ensure that the detainees is not subject to torture.

Blame Shifting and Torture

The Washington Post writes that the Obama administration has approved a new unit, housed in the FBI but drawing personnel from other national security agencies as well, that will interrogate suspected terrorists.

The article has new details about what the unit, whose creation was floated a few weeks ago in general terms, will and will not do. To my mind one of the most important changes is that the unit will be overseen by the National Security Council (i.e. the White House).

Usually such organizational box shuffling is just that, but in this case it could really matter. One reason detainees were abused in the past was that no one was clearly responsible for setting out how they should be treated; the Justice Department issued rules defining torture, but these were implemented by other agencies such as the CIA. The Bush White House played an indirect but influential role in the crafting of these rules. But since the role was indirect, it allowed the President and principals to blame abuses on overeager or ill-trained subordinates in the field.

By centralizing authority and oversight in the White House, this new arrangement could make it more difficult for the political authorities to shift blame onto others. Knowing they are (now) responsible for making sure nothing bad happens, people in the White House hopefully will exert a lot of effort to ensure that detainees are treated in accordance with domestic and international law.

So you want to go to graduate school....

Consider this comparison from Jorge Cham's PhD Comics first:

I Blackberried My Iphone

Warning: This post is more technical and boring than most on this blog. Really.

I have had an Iphone for about a year, and it's great for all of the reasons you have heard--nice design, apps, integration with music, etc.

I used to have a BlackBerry, which was fine but inferior to the Iphone in many ways. One advantage that he Blackberry had was "push email"--your emails appeared on your device immediately, and a little light went off when you had a new message. The Iphone instead "polls" for email, meaning it check your email server occasionally for new mail. This means you might have wait--gasp--a half hour or so for your messages to show up on your Iphone! Such a delay is of course totally unacceptable, especially if you're used to the CrackBerry experience. Also, the Iphone doesn't even notify you on the main screen when you have new mail.

Recently, though, Apple has been allowing "notification" applications, which will play a sound and display a message on the screen when you receive a new email. I tried three of these email notification services recently. Two were terrible, and one has worked pretty well so far:

Gmail Growl: Not so good. First, it only works if you have a computer that is always on and connected to the Inteernet. Second, it's complicated: you need to install two pieces of software on your computer, the Prowl app on your Iphone, and then configure everything properly. Third, it only works for Gmail; this is fine for me but not for everyone of course. Fourth and worst, I could never get it to work consistently; it stopped sending notifications after a day or so, forcing me to reinstall the software over and over again.

GPush: This is a standalone Iphone app, so there's no need to install software on your computer. It also only works for Gmail, though. And in my case it stopped working after about half a day. I fiddled with settings a bunch of times, which was no fun. Apparently I'm not the only one that was frustrated by this; it has a terrible rating in the App Store.

PushMail: This one works. It's an Iphone app, so that is good. You install it, create and username and password. You then set your email to forward incoming messages to an address that PushMail provides (while making sure you also keep the incoming messages in your normal email account). PushMail uses these to send notifications to your Iphone screen. You can set it to only provide notifications for some messages (i.e. from particular senders), to notify you on the screen, and to play a sound when new mail arrives. More or less just like BlackBerry. So far this has worked flawlessly for me, and I'm very happy with it. One concern is that you are forwarding some or all of your mail to PushMail. They state very clearly that they immediately discard this, but some might worry about it. Not me though, as my email is so boring that no one would gain anything by reading it!

Elections and Terrorism

Afghanistan had a national election today. Many expected (correctly) that this event would be accompanied by a surge in terrorist attacks. This got me to wondering: what does the social science literature have to say about the connection between terrorism and elections?

I didn't find what I expected to find. I expected that someone out there has written a paper using cross-national time series data on elections and terrorism plus some control variables. My (admittedly not exhaustive) search found no such paper. This is surprising. Such a paper would be pretty easy to write; we have good data (or at least data that is widely used) for both phenomena. There is a really nice paper exploring the connections between elections and human rights abuses that could serve as a model. There are of course huge literatures on terrorism and on elections that could be mined for insights. There is also a smaller literature exploring the relationship between political violence, riots, etc. and elections that could certainly be brought to bear on terrorism. It seems like there would be some pretty straightforward hypotheses that could be developed from the existing literature on democracy and terrorism. One might argue that elections prompt terror attacks, as terrorists use them as an opportunity to make a political point at a time when many citizens are paying attention to politics. Or one might argue that elections dampen terrorism by giving dissidents non-violent means to express their political views. More interesting would be to see if any relationship between terrorism and elections is conditional on the history of past elections, the nature of the parties competing for government, the party system, etc.

So why has no one written this paper? Maybe the connection between terrrorism and elections is so obvious that no one (but me) thinks such a paper would be valuable. Or maybe someone has written such a paper, but did not get "interesting" results that could be published. I hope this isn't the case, since some of the ideas I sketched in the past paragraph would suggest that any relationship would be not entirely obvious.

There are a number of papers that look at the effect of terrorism on the outcomes of elections. This is in some ways a more interesting relationship to explore, as terrorists that engage in attacks presumably want to influence the election. This very interesting paper concludes, contrary to what one might initially expect, that terrorist attacks make Israelis more sympathetic to the Palestinian cause; there are a number of other papers that explore additional effects of terrorism on voters. This paper argues that the terrorist attacks in Spain in 2004 reduced support for the government, contrary to the "rally" hypothesis that predicts more support for the government in power after security threats.

These works on effects are really interesting, and as the two cited above suggest, often reach conclusions that contradict our expectations. Most are based, though, on a single country, such as Spain or Israel. This makes sense, as it allows the researchers to obtain high quality and comparable data (such as from polls) and to control in part for the context of the attack and the likely attackers. One obvious and interesting extension would be to see if these insights could apply cross-nationally. This might give us a better idea of the general influence of terrorism on elections as well as highlighting some reasons why these effects vary due to differences in party systems, age of regime, and so on.

New Paper: Understanding Why Terrorist Operations Succeed or Fail

Interesting new paper on this topic by Brian Jackson and David Frelinger available here. The authors start by defining success and failure. They then posit that success is influenced by three factors:

• terrorist group capabilities and resources
• requirements of the operation it attempted or is planning to attempt
• relevance and reliability of security countermeasures.
  

So far, so good. The most interesting contribution of the paper is the argument that it is not useful to look at these factors in isolation. Instead, one has to consider how the implications of various combinations to understand the likelihood of success. Here is the punchline from the paper:

The heart of our argument is that it is not the absolute values of any of the characteristics that fall into these classes that are important for understanding the likelihood a terrorist operation will succeed or fail, but the relationships between them.. . . . .In general, the chances of an attack succeeding increase when (1) the
characteristics of the attackers closely match the characteristics of what they are attempting and (2) when there is a mismatch between those characteristcs and the security or protective measures the attack must overcome.

Most of the paper is devoted to developing this line of thinking in detail. As I read the paper, I wondered him the argument might be tested in a rigorous way. The authors point out that most terrorist datasets lack the level of detail needed to engage in such an evaluation. Case studies seem the way to go, is the implication.

An additional way to evaluate the argument might be with the use of fuzzy set methods. (Charles Ragin wrote the bible on fuzzy set methods in the social sciences, as well as software that can be used to implement them). These might have two advantages. First, they are explicitly designed to evaluate arguments that rely on the combination or interaction of independent variables. This would seem to match up very closely with the type of analysis advanced here. Second, one can use fuzzy set methods with a smallish number of cases (roughly, a dozen or more). It might be feasible to collect information on this number of attacks or campaigns without too much difficulty.

Quadrennial Homeland Security Review

DHS is conducting a quadrennial policy review, and has opened the process up to outsiders for comments and suggestions. Here is what they list as preliminary goals for counterterrorism:

1. Counter terrorists and other malicious actors.
2. Protect critical human, physical, and cyber infrastructure and the services they provide.
3. Prevent illicit access to and movement of dangerous capabilities and material.
4. Prevent loss of life and property from natural and manmade hazards (shared goal with Preparing for, Responding to, and Recovering From Disasters).
5. Foster a prepared, informed, and empowered citizenry and local communities (shared goal with Preparing for, Responding to, and Recovering From Disasters).
6. Build and sustain robust critical systems, networks, and functions.
7. Ensure levels of readiness to rapidly respond and recover (shared goal with Preparing for, Responding to, and Recovering From Disasters).

What about human rights and civil liberties? Not mentioned here. So here is the comment that I left:

"Efforts to achieve these objectives must always respect human rights. Human rights are an integral and valuable part of the American political tradition. Respecting human rights--especially the right not to be tortured, killed, or imprisoned for one's political beliefs--is not only a valuable goal in and of itself, but also reduces grievances that motivate terrorism. Encouraging other countries to better respect human rights should also be an important objective of US counterterrorism policy."

I got this idea here. Pretty clever, yes?

I have never met a homeland security person that disagrees with this basic idea. What's interesting, though, is that they treat it as a background condition. Many assume that *of course* respecting human rights is important, and that since everyone agrees with this sentiment, it does not really need to be discussed or evaluated further. Others assume that *of course* the US has a great human rights record, so we don't really need to work on it.

The US does have a great record of protecting and promoting rights. My worry, though, is that this attitude opens the door to policies that ignore rights or that actually infringe on them. After all, the US record is not perfect--think of Abu Graib, torture, rendition, etc. Assuming that this is not an issue makes it easier to ignore the practical problems that such abuses create for an effective counterterrorism policy.