Interview with Alan Dershowitz
Interview conducted in 1995.
"The question on the other side is, can we ever do better than 90 percent? By the way, I don't think 1 out of every 10 people in jail is innocent. I don't believe that's the case."
Alan Dershowitz is well known for his legal work on the defense teams in such famous cases as Claus Von Bulow, Mike Tyson, Patty Hearst, Michael Milken, and O.J. Simpson. He is currently Felix Frankfurter Professor of Law at Harvard Law School and received his JD from Yale where he graduated first in his class and was editor-in-chief of the Yale Law Journal. His newest book, Contrary to Public Opinion, was published by Pharos Books in 1992.
Aczel: As you know …. I would like to ask you about probability and the law.
Dershowitz: Well, I'm very interested. I wish I had thought to invite you to my class, because in the last week I've been talking about probability. I'll tell you the three contexts in which I was dealing with it. First, I was asking the class whether or not proof beyond a reasonable doubt ought to be given a number, whether jurors ought to be told proof beyond a reasonable doubt means 90 percent or 75 percent or 80 percent. Should jurors have different assessments of what proof beyond a reasonable doubt is and should judges explain to jurors "what does it mean that this person is 90 percent likely? Either he did or he didn't do it." And I explain that we have a fingerprint test or a DNA test or a ballistics test - in these cases all we are ever doing is saying that this fingerprint falls into a category, that if we were to convict all people who had matches like this, 90 percent would be guilty and 10 percent would be innocent. The guy is either guilty or innocent, but I explain to them very simply what probability means I introduce them to Kahnemann and Tversky's arguments about the difficulties of probabilistic thinking. So that was one session in class, whether or not we should allow probabilities to get to the jury. Then, another section dealt with just the problem you're talking about now: unanimous juries, size of juries, peremptory challenges to juries. What's the difference if you have the requirement of 8-4 for conviction, or if you have 12-0. What is the difference if you have 6-2 and you have only 8 people on the jury? Does the defendant or the prosecution benefit from smaller size, from different ratios? Finally I'm going to be spending a lot of time on the issue of statistics in rape and whether it is underreporting or overreporting or both in the context of rape. So my class spends a lot of time on trying to understand simple statistics, simple probability thinking. I think the law is allergic to numbers. The law is terrified of numbers, and you'd rather have jurors being told that proof beyond a reasonable doubt means proof to a moral certainty along with the other instruction that I hate, the one that says you should make this decision in a criminal case as if it was a very important decision in the other important aspects of your life. Well, most of us operate on 51 percent. If you can get 51 percent, why take 49 percent? But in the criminal justice system, if you have 51 percent versus 49 percent, you're supposed to go with the 49 percent. Even if you have 60 percent versus 40 percent, you're supposed to go with the 40 percent. It's very counterintuitive.
You know in the end what I argue - I give them the classic example of the blue buses and the white buses and there is an accident in which a blue bus ran somebody down at a very high speed. We know that whoever was driving the bus is guilty, and whatever company is driving the bus is guilty, and we know nothing more about the bus except that 90 percent of the blue buses in town are owned by company A and 10 percent are owned by company B. So it is clear that it is 90 percent likely - that is, if we were to convict under all these, 90 percent of the time we'd be right to convict and 10 percent of the time we'd be wrong. I ask the students how many of you would convict under those circumstances and a lot would convict but a lot wouldn't. I ask the students who wouldn't convict because 90 percent is not enough: What would you do with the following case? Let's assume somebody says, "That's the person. I saw him. I am certain of identification. He's the killer." And then we introduce Elizabeth Laftes and she comes into class and testifies "That's a pretty good identification. It's an 80 percent identification. In 80 percent of the cases that level of identification with that level of knowledge is going to be accurate and in 20 percent of the cases it's going to be wrong." Would you permit conviction there? They all say yes because that's a clinical ID, that's clinical even though you get a statistician coming in and saying, "Sure, we know it's a clinical ID, but the clinician is right only 80 percent of the time." We don't want to hear that.
Aczel: Like "Mr. 100%", Dr. Bradley [Reversal of Fortune] …. Actually that's an argument the other way - the problem of overconfidence.
Dershowitz: Right. There's a lot of good writing on that, on the overconfidence phenomenon, and going back even to some of the early work by the guy who wrote a book called Statistical versus Clinical Prediction in which he assessed the comparative accuracy of clinicians and statisticians in making valid predictions. In the end he came to the conclusion that statisticians are likely to be more accurate only because they are more sensitive to issues of accuracy.
Aczel: Because they're used to thinking of numbers….
Dershowitz: Clinicians are saying "sounds good." They never understand that you can make the same mistake 20 times.
The whole notion of probabilistic thinking is key, it seems to me, to the legal system. Whether in the end you want to give juries explicitly probabilistic roles, that's a different question. But surely the law has to be attuned to the fact that it is in a probabilistic business. We never deal with certainty. We know that rules of evidence are designed to create a certain probability and that it's a complex probability because we are very sensitive to type 1 and type 2 errors. I have a section in my class where we deal with predicting the future. I ask basically how many false positives are we willing to tolerate in order to avoid how many false negatives. Or bail decisions. We're going to let somebody out on bail. Is he going to kill, or isn't he going to kill? We're going to reduce somebody's sentence or we're going to increase somebody's sentence. These are all probabilistic determinations. And yet the law is allergic to putting a number on it. We have no jurisprudence of prediction. (I've written fairly extensively about that.)
Aczel: You're unusual in that thinking, aren't you?
Dershowitz: I think so. I want to at least expose the law to the fact that they are doing this. I'm not sure in the end that everything ought to be numbers - I don't want to substitute computers for juries - but I do want the law to know what it is not doing and what it is doing and want it to face up to the fact, why they're so afraid of numbers.
One reason obviously is that we want a democratic jury and because a democratic jury will have lots of different levels of aptitude and ability and if you gave jurors explicitly probabilistic tasks, the elite members of the jury would play a greater role in the deliberation. You don't want that to happen. You can't have everything in a jury system. A jury system requires lots of compromise with truth.
Aczel: It's my understanding that probability reasoning was used until 1968 with the reversal of People v. Collins in the California Supreme Court.
Dershowitz: Well, it's the case of People v. Collins that I teach in my criminal law class. And, no, I don't think that's an accurate assessment. I think that there was no prohibition on using probabilities, but I don't think it's been widely used.
Aczel: I see. It was a bad case.
Dershowitz: I think it's a bad case for two reasons. [Professor Larry Trive was the law clerk on that case.] It confuses what was wrong in that case. Obviously, the math instructor didn't know what he was doing in terms of the independence of the variables and so on. It's confused in the inherent problem. Now what I do is throw the students a real curve ball because I reverse the facts of Collins. I say to the students: Now, what would you do if the following were the situation? The prosecutor in Collins says: "Ladies and gentlemen of the jury, use your common sense. Just use your common sense. Black guy, white woman, pony tail, yellow Cadillac. Come on. What's the likelihood that there would be another couple like that? Just use your common sense. This can't be coincidence." The defense attorney then introduces a statistician who says, "Hey, wait a minute. That's not such an uncommon combination of characteristics. Let me tell you how many black men and white women drive yellow cars and have pony tails. If you think of that statistically, there's probably a one in six chance that there could be another couple like this." Would you allow the defense to use it, to undercut a common sense misconception? Larry [Tribe] won't answer that question. I've tried.
Aczel: Why?
Dershowitz: Because he made a clever argument for the prosecution not being able to use it. When it comes to the defense being denied an opportunity to use math, when I want to show a reasonable doubt, I would like to use statistics because I think jurors are more likely to accept coincidences as absolute certainty than they are really warranted in accepting. There are a lot of coincidences in life. When your two fingerprints with four rings or five rings look the same. I would like to introduce a statistician to say "Wait a minute. First of all, we're not so good at replicating rings. Maybe it's not exactly the same ring. Second, if each one has a 1 in 10 chance and there's only 10 times 10 times 10 times 10, that's not so improbable when you have an extremely large population. After all, how was he spotted? He wasn't ID'd."…. I think it's sometimes useful to introduce mathematics - simply, though I wouldn't call it highlevel mathematics.
Aczel: So why don't people do it?
Dershowitz: I think, first of all, lawyers aren't trained; they get very nervous about numbers.
Aczel: So it's a matter of education, doing it in the classroom?
Dershowitz: It's a matter of education and doing it in the classroom. I give my students an article (I did a few years ago) on a probabilistic approach. It has some charts, but it's accessible. It's very easy, I wrote it about a subject that the students cared deeply about. The American Bar Association was trying to predict which law students would get in trouble later on so maybe they shouldn't be admitted to law school. And what I did was write a simple article about how complex this notion of analysis, discriminating in this way is, and how much overprediction you need in order to get any significant number. What I tried to argue was that law shouldn't borrow mathematics. Law should use mathematics. We should say: Here's our problem. Now let's think about how math can help us solve that problem.
Aczel: The lawyers have just shied away from it?
Dershowitz: They've traditionally shied away from it. I think after People v. Collins, Tribe wrote an article called "Trial by Mathematics" in which there was an exchange of a lot of people arguing…
Aczel: Mosteller, here at Harvard, wrote about it…
Dershowitz: Yes. I, in fact, took Mosteller's course when I first came to Harvard because I wanted to become more sensitive to mathematical models and how I could use math in my own research. For example, when they developed the concept of the xyy chromosome, everybody was saying, "Oh gee, the xyy chromosome is going to be able to predict violence." I did a little article saying, "Wait a minute, let's look at the statistics on this for a minute. Let's see what it does and what it doesn't do."
Aczel: I see.
Dershowitz: The heuristic value of math is powerful.
Aczel: What then is reasonable doubt to you, if you had to put a number on it?
Dershowitz: Well, if I had to put a number on it, I would say proof by a preponderance of evidence is really 51 percent and proof by clear and convincing evidence, I would put at around 75 percent, and I would put proof beyond a reasonable doubt at about 90 percent. But is that enough, though? Does that mean we're willing to have 1 out of every 10 people in jail being innocent? The question on the other side is, can we ever do better than 90 percent? By the way, I don't think 1 out of every 10 people in jail is innocent. I don't believe that's the case.
Aczel: You think it's much less?
Dershowitz: I would put it at two or three percent. Well, it depends on how you define innocence or guilt. If you are asking me, is it the wrong person who did it, two or three percent. Are you asking me whether, for example, in a date rape case - it's a close case that falls on one side or the other…
Aczel: It's a fuzzy-set type problem.
Dershowitz: It's much more there. Are we talking about degrees of guilt? Is it first degree murder or second degree murder? We're getting up there. But on the pure issue: "Is this the wrong person?" two or three percent.
Aczel: So, we're doing quite well.
Dershowitz: We're doing quite well. But it still means for every 10 thousand people in jail, there are 300 who are innocent. That's a lot of people.
Aczel: But it's still better than 10 percent - that's scary.
Dershowitz: If you figure the numbers, there are almost one million people confined in the United States today. That means 30,000 people would be innocent in American prisons. Now, of these people who are innocent, a lot of them did it before and will do it again.
Aczel: So that's where it's a matter of degree?
Dershowitz: No, it's not a matter of degree. They are really innocent. But they are more likely to be picked up. They are more likely to be convicted. They are more likely to be sentenced to jail because they are bad guys. Now, to me, the paradigm case is a case that I'm currently involved in which is Jeffrey McDonald. Jeffrey McDonald is the man who was convicted of killing his whole family - the Marine, the Green Beret. He is either the worst, most horrible criminal imaginable, or the most unbelievable vilified innocent person. There is no gray area. In most cases there is gray. Not in this one.
Aczel: What's the most important thing you want your students to take out of the classroom?
Dershowitz: Passion is a very important thing for me. They're very smart and I want them to approach their roles with a real feeling. In terms of intelligence, there's not much I can teach them. The idea of coming to their task with a real passion is very important. The idea of thinking creatively, of not being bound by existing structures, being able to really think in a way that breaks boundaries. I tell them, "Don't think like a lawyer. Think like a brilliant scientist. Think like a discoverer, like a creator. Think like an explorer, not like a lawyer."
