 |
 |
 |
 |
 |
 |
Last week was not the first time the United States was transfixed by an act of terror. In 1964, three civil rights workers in Philadelphia, Mississippi were (quoting Wikipedia) “threatened, intimidated, beaten, shot, and buried by members of the Mississippi White Knights of the Ku Klux Klan, the Neshoba County Sheriff’s Office and the Philadelphia Police Department.” It took 44 days and an FBI-initiated act of torture to locate their bodies.
The FBI, in a nod to the theory of comparative advantage, subcontracted the torture to the Mafia, more specifically to the Colombo family associate Gregory Scarpa. Here’s the story as relayed by Selwyn Raab, the New York Times investigative reporter who covered the Mafia for 25 years:
[Scarpa] went down to Mississippi for the FBI and kidnapped a KKK guy agents were sure was involved in disposing of the bodies. The guy had an appliance store. Scarpa bought a TV and came back to the store to pick it up just as he was closing. The guy helps him carry the TV to his car parked in the back of the store. Scarpa knocks him out with a bop to the head, takes him off to the woods, beats him up, sticks a gun down his throat and says “I’m going to blow your head off”. The KKK guy realized he was Mafia and wasn’t kidding and told him where to look for the bodies.
(Source: Raab’s book Five Families, which is fascinating throughout. Raab says the story has been verified by “former law enforcement officials who asked for anonymity and lawyers who are aware of the circumstances”.)
The moral of the story is that torture sometimes works. Other times it doesn’t, eliciting either no information, or false information, or whatever “information” the victim believes the inquisitor wants to hear. I am almost 100% ignorant, and hence virtually 100% agnostic, about the relative frequency of these outcomes in those cases where the torturer is both skilled in his art and genuinely interested in eliciting the truth. I will be very glad if any educated reader can shed light on this question. I doubt that we’re likely to learn of any controlled experiments, but I’ll settle for sketchy data or even well-chosen anecdotes. Failing that, I’ll settle for plausibility arguments.
Tales of the Inquisition (which always seem to come up in these discussions) are not terribly relevant here, for at least two reasons. First, not every agent of the Inquisition was an objective truthseeker. Second, even when they did seek truth, they were constrained by the technology of their era. A lot of things didn’t work very well in the 16th century. Medicine, for example. The techniques of medicine have improved a lot in the past several hundred years, and plausibly so have the techniques of torture.
It should of course be obvious that the positive question “How well does torture work?” is very different from the normative question “Should we empower government officials (or, as in the Mississippi case, government subcontractors) to employ torture?”. Powers that can be employed competently for good can also be employed incompetently or for evil. But surely the positive question is relevant to the normative one, as well as being interesting in its own right.
So: What can you tell me about the efficacy of torture? And if you wanted to be better informed about this question, what sort of plausibly available evidence would you find most useful?
I refer you to The Body in Pain by Elaine Scarry.
On the anecdotal side, and not exactly on the nose for what you’re looking for, but the selection of interviews on this page entitled “Interrogation Techniques” are taken from retired professional interrogators, frankly discussing some of their borderline methods.
Objective question first. I am sure that torture sometimes “works” in that it elicits correct information, but this is not enough. To really work, it must elicit true information that could not be obtained without torture. It seems overwhelmingly likely that it sometimes does this as well. But even this is not enough, it must NOT elicit wrong information that would not be obtained without torture. It seems overewhelmingly likely that it sometimes does this as well. No amount of examples where correct information has been obtained can prove that that torture is beneficial unless the false information is recorded and weighed as well. For obvious reasons, there is likely to be a serious under-reporting of negative results (mis-information).
Normative question- which is not the one Steve poses, but I can’t ignore it. Hypothetically, IF we had a situation where to torture one person would prevent the torturing of 10 people and the torturer is rational and informed, so he knows that his inflicting of torture has saved even more torture, therefore he suffers no guilt, THEN I don’t see how we could rationally say the single torture should not occur. Therefore torture in principle can be the moral thing to do. It is possible to have a system of morality where this torture should still not occur, but it is not mine.
In the world, we always fail to know the things that are required to be certain that torture is justified, so generally I am very much in favour of banning it. Given our tendency towards (for example)confirmation bias, I think it very likely that torture will lead to net negative results. I certainly don’t think the known costs are worth the chances of positive results.
One could possibly have something like a public interest defence. The torturer will always face charges, but if sufficient public interest could be shown, then there could be an aquital. The torturer may have to show that beneficial results actually occured. Thus the torturer would always face the possibility of conviction if their torture failed, even if they had good intentions.
Steve, whenever I hear that a torture victim will just tell you whatever you want to hear, I have a very simple response: couldn’t you threaten them with continued torture if the information they reveal turns out to be wrong, and stop it if it is verified? In that case won’t “what you want to hear” be the truth?
Harold,
Objective question first. I am sure that torture sometimes “works” in that it elicits correct information, but this is not enough. To really work, it must elicit true information that could not be obtained without torture.
This has absolutely nothing to do with the first objective question and everything to do with the second question. It matters not a wit if the information gained under torture can be gained by other means under the first question. The only thing that matters under the first question is whether or not the information gained under torture is true.
What you’ve said above is like saying the directions I give you to the local 7-11 “works” only if no one else can give you another set of directions to that same 7-11.
Presumably a large part of the reason a torturer gets false information from torture is that the tortured person doesn’t know the answers to the questions. They must either remain truthfully silent in the face of the torture, or say something to make it stop – even if they know the torturer will be back after learning the information was false. Who knows – perhaps the torturer will lose interest, or learn what they want to know from some other source in the meantime. And, a person under torture probably has a very high discount rate anyway.
Beyond that, the tortured person knows that any questions they answer truthfully will simply lead to more questions, until the torturer arrives at questions to which the tortured person does not know the answer. Truthful answers will not stop the torture. No doubt this becomes clear pretty quickly.
Finally, to the extent that the tortured person is able to think straight at all, he will realize that any truthful answers will simply invite further torment by assuring the torturer that there are truthful answers to be had. In contrast, if all the answers are lies, the torturer may eventually conclude that the tortured person simply doesn’t know anything.
What it comes down to is whether the tortured person’s cognitive or instinctive processes react this way, or are so scrambled by pain that they arrive at the irrational answer of only answering questions truthfully.
According to Plutarch, Zeno of Elea, while being tortured to speak, bit off his tongue and spat it out at his inquistor.
Ken #5 and Harold #3,
I disagree with both of you.
Ken, the relative effectiveness and efficiency IS relevant to discussion of the efficacy of torture. If, at least hypothetically, it were clear that other methods are equally available and provide superior results in every way with lower “costs” of every type, one would be remiss to simply remark that torture works sometimes to some degree or even to provide some magnitude in isolation (some absolute measure) rather than compared to the other method(s).
That said: Harold, I disagree with your argument that “To really work, it must elicit true information that could not be obtained without torture.” No, it could (at least hypothetically) elicit such information at lower “cost”, or elicit the information faster.
(It could also “work” in a broader sense of deterrence insofar as some criminal activity may be reduced out of fear of torture. And it could also “work” if it is known to be happening or potentially happening and it is perceived to be more efficacious — if accomplices think the detainee will give them up and as a result change their behavior in some way conducive to their capture, such as coming out of hiding to try to flee.)
If we assume that the interrogators are rational, then giving them more tools should give better results.
I would expect that givers of false info would be tortured, as they are willfully fighting back. Those who cooperate truthfully would get better treatment.
@ Roger #9,
Re: If we assume that the interrogators are rational, then giving them more tools should give better results.
First of all, they may be operating under invalid assumptions re: the effects and efficacy of torture (in itself, and relative to other measures).
But even if not, “results” is too broad. Yes, a rational, well-informed actor would optimize results relative to his objective(s) and priorities. But the actor(s) (whoever the decision-maker[s] is/are, at whatever point[s] up and down the line of those involved in approval, implementation, etc.) may have objectives other than getting good information. For example, one may wish to appear to be doing all one can by torturing, if people think it would/may be effective. One may also desire to inflict pain as revenge for his own emotional benefit and/or for the favorable reaction of others who get an emotional benefit.
Gordon,
Ken, the relative effectiveness and efficiency IS relevant to discussion of the efficacy of torture.
Effectiveness and efficiency are two different things. Saying something is effective means how well it works. Efficiency is effectiveness given a cost. It’s true that using a shotgun to kill a fly is effective, but that’s far less efficient that using a fly swatter, given the cost of a shotgun and shells vs the cost of a fly swatter, then considering the collateral damage of using a shotgun vs a fly swatter.
Suppose you’ve got somebody in front of you who *might* be a terrorist possessing the exact piece of crucial information that you need. Or might be entirely innocent, or might be guilty but not in possession of that info. You aren’t sure. The odds against any one person – even a “suspicious” one from a relevant ethnic/geographic/political group – either being a terrorist OR (if they are a terrorist) knowing the info you want are pretty big – the prior probability assessment of them knowing what you want has to be a fraction of 1%.
If you have 100 people in custody and you are *nice* to them and one guy helpfully voluntarily *tells* you the info you need while under no compulsion to tell you anything, the odds seem pretty good that what he tells you is the truth. So tracking down the resulting lead is NOT a waste of precious time and resources.
Whereas if you torture the prisoners, the odds of them ALL giving you SOME answer are now nearly 100%, but the odds of them KNOWING the answer haven’t changed. If you estimate the prior probabilities and apply Bayes Theorem, you should get the result that confessions under torture are more than 99% likely to be false and designed to appeal to your preexisting prejudices. In switching from “be nice” to “torture” you’ve gone from one good lead (or the true knowledge that you have no good leads) to 100 false leads.
In short, confessions under torture are generally not worth following up on.
It’s possible to construct scenarios that make it more plausible, but most of them are movie-plot stories. None resemble the situation in Guantanamo.
Ken #11,
Steve referred to the question, ““How well does torture work?” and asked “What can you tell me about the efficacy of torture?”
Although my grammar in #8 was incorrect (“IS”), I was, and am, saying that both the relative effectiveness and efficiency ARE relevant to discussion of the efficacy of torture, because it would be at best not very meaningful/useful to simply conclude that torture has a greater than zero probability of having greater than zero effectiveness, without any discussion or implication or hypothesis, etc,, re: relative effectiveness or relative efficiency.
So when you say that “It matters not a wit [to the above question] if the information gained under torture can be gained by other means”, I disagree, at least if we are to address the question in a meaningful or useful way.
Some historical background:
The Romans used torture and they were a very practical bunch. It is hard to believe they used something that was completely worthless.
Also, if a slave was to testify against a free citizen, the Roman legal system required that the slave be tortured before the testimony was admissible. Hard to believe they would have done this if torture made the testimony less reliable.
The example you give does not in any way sound like an illustration of torture working. It sounds like an example showing that sometimes *threats* work. Credible threats, as you point out.
Cos: I’d have called a blow leading to unconsciousness, followed by a beating, followed by a gun down the throat an instance of torture. Your mileage may vary.
I see two arguments against torture: that is does not work, and that it works too well (creating pressure to use it more).
a panel discussion a few months ago (involving people who would know) suggested that the waterboarding after 9/11 involved asking questions the interrogators already knew the answers to. it was done not to elicit new truthful information but to break the resolve of the prisoner, after which the torture would end and the interrogators could then ask questions they did NOT know the answer to and get lots of truthful answers…
http://www.slate.com/articles/health_and_science/human_nature/2013/01/the_case_for_torture_ex_cia_officials_explain_enhanced_interrogations.single.html
SD: That’s just the sort of controlled experiment I was looking for (but pessimistic about finding). Thanks!
Torture is one area where my views are deontological. I feel so strongly about this that to the extent you can find utilitarian arguments rationalizing torture, I find that to be an indictment against utilitarianism, not a defense of torture.
That said, all the wavering and if’ing and posturing makes me suspect that we have no clue what the efficacy of torture is.
@Roger #11 “If we assume that the interrogators are rational, then giving them more tools should give better results”… for the interrogator. Not necessarily for anyone else with an interest in the investigation.
“Sir, the suspect has divulged the names of his accomplices!”
“Congratulations, good work! I’ll be glad to report this excellent news to our Precious Ruler!”
Ken & Gordon/Brooks. Ken is correct – whether torture is effective does not depend at all on whether the information could be gathered elsewhere. I had mixed up my assumptions about efficiency with the argument about effectiveness.
The effectiveness of torture depends only on whether you get correct information and you do not get disinformation you would not get by other means. As to whether we should torture (the normative question) – I think it is clear that we should not torture if we can get just as good information by other means. So in practical terms, whether torture is effective becomes moot if we can get the information elsewhere.
@19, 20 -Steve, I do not see a control in the link – I do not even see an experiment. I see anecdote. Have I missed something?
Harold:
I do not even see an experiment. I see anecdote. Have I missed something?
What I meant to say is that this shows me how one *could* do an informative experiment —- if the questioner, unbeknownst to the prisoners, already has a key piece of information, he can experiment to see which techniques elicit the information accurately.
#24 – Ah, I see. In trials of this sort the gold standard is “double blind”. Now the victim is not the test subject, so there is no need to blind him (at least in the experimental sense – exactly how one does the torture is up to the torturer). It does not matter if the victim knows whether or not he is tortured. The outcome being tested is the quality of information.
It *is* crucial that the person assessing the outcome is blind – i.e. does not know if the information results from torture or other methods. This prevents the interrogator simply ignoring false information, or interpreting whatever the victim says as what they know is true. This would be absolutely essential for any realistic assessment of the information quality.
It would require that victims were tortured to aquire information that was already known by someone other than the interrogator. This would be useless for the method described in the article, where the interrogator must know the information in order to appear omniscient.
Whilst it may be informative, but I think it falls well short of a controlled trial.
“Are you people insane?”
http://youtu.be/TMCKe_EpJJk
If we’re defining ‘torture’ as pretty much any threat of some “unpleasant effect” (chiefly, severe physical or moral violence) towards an otherwise recalcitrant victim unless he cooperates, then, yes, of course it works: Just ask, e.g., the countless muggers who demand a victim’s PIN at the point of a gun (with or without beating him first). Or, for that matter, Allen West. And how about the times a prosecutor agrees to charge a suspect as an accessory after the fact, rather than with a capital charge of felony-murder, in exchange for the name of his unknown partner? They all seem like examples.
Harold #23,
Re: Ken is correct – whether torture is effective does not depend at all on whether the information could be gathered elsewhere. I had mixed up my assumptions about efficiency with the argument about effectiveness.
You are completely misrepresenting my point. Of course the question of whether or not torture is effective (as opposed to completely ineffective) does not depend at all on whether or not the information could be gained via other methods. Indeed, I said so. But as I explained, “it would be at best not very meaningful/useful to simply conclude that torture has a greater than zero probability of having greater than zero effectiveness, without any discussion or implication or hypothesis, etc,, re: relative effectiveness or relative efficiency”, and thus, relative effectiveness is indeed relevant to a meaningful or useful discussion of the efficacy of torture. Re-read my #8 and #13 if you don’t get it.
“Torture doesn’t work” is an argument that I wish would go away.
I know it is not true by mere introspection. (It would work on me). False information can be checked by outside sources.
We (and anybody else) should not torture because it is immoral.
When do you stop? Where do you draw the line? What information is too insignificant to justify it? Who makes those decisions? In the future – it will make things worse by turning truly good people against you.
Gordon /Brooks #28 and Geoffry #29
We can split into two parts – efficacy and efficiency. My initial point was intended to be about false positives and false negatives. If we have a test for a medical condition that detects 100% of positive cases can we say the test “works”? Not if it falsely says 100% of negative cases have the condition. Geoffrey says torture would work on him – but he does not know how much false information would be obtained from him, and therefore does not really know that torture would work on him. In cases where the veracity can be readily checked (where is the ticking bomb) then this is not an issue. Where the information cannot be checked (who are your conspirators?) it is very important.
This is a separate point from the efficiency argument. I agree with Gordon / Brooks – that a consideration of the efficiency (can we get the information elsewhere) is essential to a *useful* discussion about torture, but not about the technical grammatical point.
Harold #30,
Not to belabor the point (further), but I’m not just talking about efficiency and relative efficiency. I’m also saying that we need some consideration of relative effectiveness to have a useful discussion of the effectiveness of torture (“How well does torture work?”), rather than than just finding that torture works sometimes (something Steve already observed in his post) or even arriving at some measure of effectiveness in absolute (not relative) terms.
An experiment can show a drug is 10% more effective vs. placebo (or vs. nothing). But if there are other treatments out there, then this measure of effectiveness is not useful to us, even before we consider costs. For that “10% effectiveness” to be useful as a measure of effectiveness, we need to know the % effectiveness of the other treatments for comparison.
The only exception would be if, for some reason, it were either torture or nothing (no other means of trying to get info from the person), or if all other methods have been maxed out somehow (not sure if that could be said though, since more of the same can always be applied) and torture is a complement and has been shown to be X% effective in that context.
Gordon / Brooks. I think we are broadly in agreement. Just differing over definition of effectiveness. For any such comparison we need some sort of cost/benefit analysis. How do we measure the benefit of torture? Is mis-infomation a cost or a reduction in benefit?
Looking at it from a person who has basically no consequence for his actions (FBI agent in the 50s to 60s).
If my odds are .00001% of finding a body without torturing someone and I believe there is a 50% chance that someone knows where the body is and a 25% chance that if the Mob tortures him that I will find the body, then torture seems to increase my odds.
My reward is prestige and probable promotion, my (FBI agent in 50s and 60s) potential cost seems relatively low.
Geoffrey – Agree on “torture doesn’t work”, but the blanket statement “torture is immoral” is another ‘argument’ I find difficult to take seriously. It seems to me the type of thing we say with the luxury of being sufficiently far removed from the possibility and consequences of having to make such decisions. E.g. is it ever immoral to NOT try to get information from someone you think is highly likely to have it in order to save any number of innocent people? Your objections seem to be practical not moral. Of course there are lots of judgment calls to make and lines to draw, but that doesn’t establish that they’re never worth drawing.
iceman, of course there may be situations where torture is warranted, but that is true for almost anything. There are cases where manslaughter is warranted. Most people, even most libertarians, would generally agree we should still have rules against it. Incidentally, this distinction explains why we have rule-utilitarians.
You should instead consider whether you are comfortable institutionalizing it. I, for one, am not. Even if I was in the abstract, given governments’ poor track record with abusing and overstepping their powers, I don’t believe there is a case to be made that this is a good idea. Even a utilitarian one.
The efficacy of torture as a method of information gathering depends not only on the willingness of the subject to answer truthfully, but also on the interrogator asking the right questions. Here’s how I’d design a study, if you assume no ethical or monetary restraints.
1. Select a large pool of individuals from various backgrounds for subjects.
2. Divide them into three evenly distributed groups: one to be tortured, one to be traditionally interrogated, and one to undergo a combination of the two.
3. Give them an information packet prior to the experiment for them to memorize that includes 10 facts.
4. Weight the facts by importance of information. The subjects will receive $10 for every fact withheld from their questioners, with multiplier values based on fact weight. Furthermore, they receive an additional $10 for every lie they successfully convince their interrogators of.
5. No two subjects have the exact same set of facts, but every fact on a subject’s list is repeated at least once on another subject’s list in their group. The subjects are aware of this.
6. If at the end of the experiment, the interrogators have learned all of the facts on your list from other subjects, you receive no money.
7. If you reveal all the information on your list and convince your interrogator to believe it, but everyone else in your group withheld information, you receive $10 for every fact on your list.
8. The interrogator does not know what 10 facts a subject holds.
9. Interrogation length is random and the subjects have no knowledge as to how long they need to hold out.
10. After questioning is over, each interrogator group looks at the totality of data and makes judgment calls on what is true and what is not true. From this, you can learn the actual truth value of each method, the perceived truth value, how much false information each method generates, and how much of that information is believed by the interrogators. Because the facts are weighted you can also assess the methods for quality of information versus quantity.
You can repeat the experiment as needed, for instance seeing the difference between when the interrogator has little to go on versus a lot to go on. Or you can repeat giving the interrogators false assumptions to see how that affects the data.
Henri Hein – another good practical point (“powers we grant for good can be used for evil”), but slippery slope arguments are rarely very satisfying particularly in times like this, and I would say it seems like a difficult case to make on grounds of pure utilitarianism. E.g. how many people wrongly interrogated would it take to offset the suffering of all the victims of an event like Boston or the Twin Towers? Rather to me that’s more an indication that one has serious reservations about utilitarianism, and must in fact be giving more weight to the risk of taking “positive” action that wrongly harms an innocent vs. failing to protect other innocents. This is why some think the “rules” in rule-utilitarianism are really attempts to reverse-engineer side constraints on our actions, out of an instinctive recognition that people do have certain inherent rights not to be sacrificed even for the ‘greater good’.
I would add it also seems like a difficult case for the deontologist – we recognize that you lose some rights when you harm or seek to harm someone, and so if clearly planning to harm someone…but if we’re only 99% certain our hands are tied? 80%? 51%? That’s why I don’t see this as a bright-line moral “issue”, rather each situation presents its own unique moral conundrum (based on probabilities of violations vs. preventing violations – so yes this is likely a gray area for most libertarians as well). I may ultimately come down on your side, my point is the moral pedestal is out of place on this stage. I do also take some comfort in the idea that as long as we have free elections and a free media to expose abuses, this is perhaps a rare area where elected officials of almost any stripe will find a similar interest; in at least trying to do what is minimally necessary to protect their constituents. (I now have to digest the fact that I just said that.) The devil is literally in the details.
As far as the controlled experiment, since ethical constraints almost surely rule out something like #36 (unless you could somehow find enough extremely patriotic volunteers), #18 seems like about the best we could come up with.
The problem I have with #18 as an experiment is that it simplifies the question of torture’s efficacy to whether or not the subject says something that is true. It fails to evaluate the interrogator’s ability to sift through the information to determine fact from fiction. The very idea of knowing the answer you’re looking for and then waterboarding it out of the subject seems dangerously close to forcing false confessions.
Are we able to distinguish between a subject who answers the questions truthfully as a sign he’s broken down and a subject who doesn’t know the answers to the questions but guesses correctly what the interrogator is pushing him toward?
#37 I don’t quite get your point about 9/11. How many people would you have needed to torture to have prevented 9/11 or Boston? Are we suggesting a random rounding up of people and torturing them in case they are planning something? In which case the suffering would outwiegh that of th evictims. Or are you suggesting a hypothetical “ticking bomb” situation where we know with a very high degree of certainty nearly everything form other sources, and require only one more piece of information that is easily verifiable?
As I said before, #18 is useless as an experiment without some verification. We are in the same situation as 19th century doctors and their leeches. It is very, very easy for educated, intelligent people to arrive at the wrong conclusion if based on people deciding if their own methods have been succesful.
Jeff: I would change your study a bit. Each subject would be given a number of facts form 0 to 10 weighted towards 0.