Update (3/26/2015): Amanda Knox and Raffele Sollecito were acquitted today by the Italian Court of Cassation. So ends their legal nightmare. I wish them both the best going forward, and I’m glad the Italian justice system finally did the right thing. I wish they’d done it before trying almost every possible wrong thing, but I’m still pleased.

Fifteen years ago, Errol Morris released Mr. Death: The Rise and Fall of Fred Leuchter, Jr., which tells the story of execution equipment designer and Holocaust denier Fred Leuchter. Morris interviews a number of experts, including those who conducted the tests that Leuchter cited in his report, to debunk all of Leuchter’s claims that the gas chambers at Auschwitz and Majdanek never existed. I’ve since wondered if Leuchter ever saw the whole movie and if so, whether it gave him any pause. But, in a way, a quote from the movie answers my question:

Errol Morris: “Have you ever thought that you might be wrong, or do you think that you could make a mistake?”

Fred Leuchter: “No, I’m past that. When I attempted to turn those facilities into gas execution facilities and was unable to, I made a decision at that point that I wasn’t wrong. And perhaps that’s why I did it. At least it cleared my mind, so I know that I left no stone unturned. I did everything possible to substantiate and prove the existence of the gas chambers, and I was unable to.”

We’re about to see another verdict in the endless trials of Amanda Knox. Whatever happens, it won’t be the last. I imagine that for many the most puzzling aspect of this case is how educated adults can continue to insist, despite the overwhelming evidence that drug dealer and petty thief Rudy Guede killed Meredith Kercher alone, that Amanda Knox and Raffele Sollecito were involved.

I certainly wouldn’t maintain that the police, prosecutors and judges in this case are ordinary fools. Neither would I think that they’d deliberately railroad innocent people for amusement. Certainly the Italian court system, like Italian society in general, has flaws that Knox and Sollecito’s prosecutors were able to exploit to gain a conviction. But the unwillingness of Italian prosecutors, the Italian Court of Cassation, and their supporters on the Internet to admit they might have been wrong can be traced to more universal flaws in reasoning, which can afflict even able and gifted people.

Reluctance to Seek, or Listen To, Independent Confirmation of Evidence

In the Italian courts this problem appears to be systemic:

Il Giornale, a conservative newspaper, for instance, recently published an interview with Marco Morin, a Venice-based firearms expert who declared he no longer wanted to work in Italian courts. “In the United States, federal judges must study a 637-page manual in order to be able to evaluate [forensic] evidence,” he told the newspaper. “Here, they accept everything without questioning, as long as it comes from the institutional laboratory.”

This is a reckless practice. A necessary condition for good science is verification of observational data and experimental results by outside experts. It’s an essential part of the self-correcting mechanism that makes science work. Any work produced by a single lab that hasn’t been subjected to rigorous scrutiny by independent scientists in the field is automatically suspect. That judges would treat it any other way indicates a basic ignorance of how the scientific method works; and since faulty scientific evidence is a leading cause of wrongful convictions in cases worldwide–it got at least one man in Texas wrongfully executed–it’s a matter of (literally) grave concern.

In Knox’s first trial, defense experts were allowed to critique the findings of the scientific police, but weren’t permitted to perform tests or conduct a full, independent analysis of their methods and results. The court sided with the scientific police, using their results as the basis of their convicting both defendants.

Two years later, the appellate judge ordered independent tests of the supposed murder weapon and a bra clasp found at the scene. The independent experts found that the scientific police made egregious errors in evidence collection, lab procedure, and data interpretation, and in doing so debunked the core of the prosecution’s case against the defendants.

In vacating Knox’s acquittal, the Court of Cassation moved the goalposts on some evidence (making the impractical demand that independent tests not only identify lab errors and other possible or probable sources of DNA contamination, but also find the exact vectors that led to each piece of evidence being contaminated) and demanded additional tests be done on the knife.

The new tests concluded on the knife in question showed no trace of the victim’s DNA or blood anywhere on it, confirming the results of the independent experts’ report and further debunking the prosecution’s scientific case.

In the Knox case, scientific evidence of her guilt weakens each time it’s examined. (And it was tissue-thin to start with.) Their alleged murder weapon and the bra clasp, the only evidence tying Sollecito to the crime scene, have been discredited. Prosecutors still have Knox’s DNA in the apartment she shared with Kercher and two others, but since there’s no way to determine when that DNA was deposited–people constantly shed DNA in the form of skin cells, skin oils, hair, and sweat–it doesn’t tie her to the scene of the murder at the time of the murder.

Science is supposed to work this way, forever evaluating and reevaluating evidence and coming to increasingly compelling conclusions. This can be a great help in criminal investigations, but only if judges take the scientific method seriously.

Failure to Follow the Strong Signal or See the Convergence of Evidence

From Skeptic magazine (Thanks, NIzkor):

In 1840 the English philosopher of science, William Whewell, published his classic work on The Philosophy of the Inductive Sciences, in which he talked at length about inductions, or generalizations drawn from specific facts. But to prove a theory one must have more than just one generalization. And these multiple inductions must all point to a definite conclusion, building upon one another independently but in conjunction. Whewell said of these inductions that they “jumped together” to establish the veracity of a theory. A fond coiner of words (e.g., “scientist”), Whewell called this method of thinking a consilience of inductions. “Accordingly the cases in which inductions from classes of facts altogether different have thus jumped together, belong only to the best established theories which the history of science contains. And, as I shall have occasion to refer to this particular feature in their evidence, I will take the liberty of describing it by a particular phrase; and will term it the Consilience of Inductions” (p. 230).

A less cumbersome phrase might be a convergence of evidence. Evolution, for example, is proved by the convergence of evidence from geology, paleontology, botany, zoology, herpetology, entomology, biogeography, anatomy, physiology, and comparative anatomy. No one piece of evidence from these diverse fields says “evolution” on it. A fossil is a snapshot. But a fossil in a geological bed, with other fossils of the same and different species, compared to species in lower strata and upper strata, contrasted to modern similar organisms, juxtaposed with species in other parts of the world, past and present, and so on, turn that snapshot into a motion picture. Each set of inductions from each field jumps together to a grand conclusion–evolution. This is how narrative history becomes scientific history.

How do we apply “conciliance of inductions” to the Knox case? A single piece of evidence, considered in isolation, won’t help us, because we need more than one generalization to achieve a conciliance. Meredith Kercher’s death involved several events which had to combine to result in her murder in her bedroom. So it is sensible to start with the place where her body was found and work outward, following the strongest evidence, and see how it converges. It would look something like komponisto did in his Bayesian analysis of the case back in 2009:

We live in a lawful universe. Every event that happens in this world — including human actions and thoughts — is ultimately governed by the laws of physics, which are exceptionless.

Murder may be highly illegal, but from the standpoint of physics, it’s as lawful as everything else. Every physical interaction, including a homicide, leaves traces — changes in the environment that constitute information about what took place.

Such information, however, is — crucially — local. The further away in space and time you move from the event, the less entanglement there is between your environment and that of the event, and thus the more difficult it is to make legitimate inferences about the event. The signal-to-noise ratio decreases dramatically as you move away in causal distance from the event. After all, the hypothesis space of possible causal chains of length n leading to the event increases exponentially in n.

By far the most important evidence in a murder investigation will therefore be the evidence that is the closest to the crime itself — evidence on and around the victim, as well as details stored in the brains of people who were present during the act. Less important will be evidence obtained from persons and objects a short distance away from the crime scene; and the importance decays rapidly from there as you move further out.

It follows that you cannot possibly expect to reliably arrive at the correct answer by starting a few steps removed in the causal chain, say with a person you find “suspicious” for some reason, and working forward to come up with a plausible scenario for how the crime was committed. That would be privileging the hypothesis. Instead, you have to start from the actual crime scene, or as close to it as you can get, and work backward, letting yourself be blown by the winds of evidence toward one or more possible suspects.

In the Meredith Kercher case, the winds of evidence blow with something like hurricane force in the direction of Rudy Guédé. After the murder, Kercher’s bedroom was filled with evidence of Guédé’s presence; his DNA was found not only on top of but actually inside her body. That’s about as close to the crime as it gets. At the same time, no remotely similarly incriminating genetic material was found from anyone else — in particular, there were no traces of the presence of either Amanda Knox or Raffaele Sollecito in the room.

In addition, Guede’s bloody footprints start from inside the murder room, heading in two directions, toward the bathroom and toward the exit door, indicating that after his bathroom visit (presumably to wash Kercher’s blood off as much as he could), he went back into the murder room, stepped in blood again, and tracked it out again during his escape. His feces is in the toilet of one of the bathrooms. The evidence that Rudy Guede was in Kercher’s apartment at the time of the murder comes from multiple directions–footprints, palm prints, fingerprints, DNA, all closely connected to Meredith Kercher’s body or her blood. There is simply no way of explaining the state of the murder scene that doesn’t demand Rudy Guede’s presence and involvement.

Further evidence: When he was arrested, Rudy Guede still had cuts on his hand consistent with having used a knife. Also, Guede was arrested in Germany trying to illegally board a train, having fled from Italy in the days following the murder. This indicates consciousness of guilt.

All that remains are questions of how Guede got into the murder scene and how he left, and whether either ingress or egress required help from a third party.

The prosecution has maintained that entry through the bedroom window of Knox and Kercher’s Italian roommates was impossible, that the wall could not have been scaled. This turns out to be wrong, and easily debunked:

[youtube=://www.youtube.com/watch?v=8JL6nIkaYLs&w=560&h=315]
Does this mean that Guede didn’t have help getting in? No. It simply means that a third party isn’t necessary to explain how he got in. As for the question of getting out, I’m pretty sure Guede, a sometime burglar, had mastered the art of grabbing a door knob, turning it, and pulling. No third party needed there.

What does this add up to? As I see it, the only person other than Meredith Kercher whose presence is required to explain Kercher’s death and the state of the crime scene is Rudy Guede. Amanda Knox and Raffele Sollecito are completely superfluous. Any attempt to add them to the explanation of this crime adds complications requiring heavy doses of post hoc rationalization to excuse the lack of physical evidence against the couple. These come mostly in the the form of cleanup theories that don’t explain how Knox and Sollecito could have found, much less cleaned up, all of their own DNA, hair, and fiber traces while leaving all of Rudy Guede’s physical traces intact and undisturbed. (Or indeed, why they should have wished to leave Rudy Guede’s traces intact and undisturbed.)

If police, prosecutors, the original trial judges, and the Court of Cassation had simply followed the best evidence they had, the physical evidence on, adjacent to, or leading away from the body, they could have saved everyone a lot of time and grief. But they went a different way, and since they started down that path, they’ve refused to correct course–with the notable exception of Judge Hellmann. Why?

Ideological Immunity

From Why People Believe Weird Things:

In day-to-day life, as in science, we all resist fundamental paradigm change. Social scientist Jay Stuart Snelson calls this resistance an ideological immune system, “educated, intelligent, and successful adults rarely change their most fundamental presuppositions” (1993, p. 54). According to Snelson, the more knowledge individuals have accumulated, and the more well-founded their theories have become (and remember, we all tend to look for and remember confirmatory evidence, not counterevidence), the greater the confidence in their ideologies. The consequence of this , however, is that we build up an “immunity” against new ideas that do not corroborate previous ones.

This TED talk by Kathryn Schulz, the author of Being Wrong, goes into greater detail explaining the why this is a problem:

http://embed.ted.com/talks/kathryn_schulz_on_being_wrong.html

Ideology takes many forms in the Knox case, and leads to several different categories of bias. I’ve discussed Italian-style misogyny elsewhere on this blog, and lead prosecutor Mignini’s conspiracy biases have been well documented. In this space, I’d like to get to something more universal: our confidence in our ability to read social cues and use them to spot bad people.

What struck me about the way that police and prosecutors and some members of the public reacted to the Kercher case was their claimed ability to see Amanda Knox’s guilt based on her appearance or her “odd behavior”. One of the investigators actually said that he knew that she and Sollecito were guilty just from observing them in their daily lives (including, according to Douglas Preston, their meal choices):

[youtube=://www.youtube.com/watch?v=sWkZPWRS3N0&w=420&h=315]

Many police officers are like this one and believe that they have a special gift, from their training and experience, for detecting liars and criminals. It should be said that this has a certain intuitive appeal. Applying the scientific method to criminal investigation is a practice that’s only slightly more than a century old. For the vast bulk of the history of human civilization, social cues were often our only tools for unmasking the bad guys. The question is, how reliable are those tools?

A recent study in the journal Human Communication Research offers some insight. Basically, human beings are pretty bad lie detectors, particularly when dealing with strangers. When they take in all social cues, they can discriminate liars from truth tellers 54% of the time, which means their decisions are only slightly better than those of the coin flipping Batman villain Two-Face.

Surprisingly, the wider the range of social cues observers have to examine, the worse they get, probably because each additional data point has a chance to trigger a bias in the observer. The best results are obtained from limiting the observation to audio only (59%), but even then, let’s face it, people kind of stink at this.

Evaluating people by social cues can provide us evidence, but because the various biases of the observer have a severe impact on the reliability of that evidence, it has little inferential power on its own, leading too often to false positives (truth tellers accused of deception) and false negatives (liars believed).

Don’t misunderstand me. Indentifying criminals through social cues can pan out. After all, Rudy Guede behaved in a guilty fashion after the murder–he fled the country. But his trip considered in isolation could have meant any of a number of things–family problems, immigration worries, a hankering for a good German brew. That this sudden yearning for Teutonic scenery struck Guede shortly after he deposited DNA, hair, fingerprints, a palm print, and his own fecal matter at a murder scene puts his actions into a criminal context.

Still, since following a social cue to a murderer can sometimes work, police do what most people often do when a method succeeds: they forget or rationalize the times it fails. From Why People Believe Weird Things:

After the ESP experiment, one woman followed me out of the room and said, “You’re one of those skeptics, aren’t you?”

“I am indeed,” I responded.

“Well, then,” she retorted, “how do you explain coincidences like when I go to the phone to call my friend and she calls me? Isn’t that an example of psychic communication?”

“No, it is not,” I told her. “It is an example of statistical coincidence. Let me ask you this: How many times did you go to the phone to call your friend and she did not call? Or how many times did your friend call you but you did not call first?”

She said she would have to think about it and get back to me. Later she found me and said she had it figured out: “I only remember the times that these things happen, and I forget all those others you suggested.”

“Bingo!” I exclaimed, thinking I had a convert. “You got it. It is just selective perception.”

But I was too optimistic. “No,” she concluded, “this just proves the psychic power works sometimes but not others.”

As James Randi says, believers in the paranormal are like “unsinkable rubber ducks.”

If a policeman catches a criminal by social cues, and later evidence backs him up, he’s a great success. Imagine further that this policeman is also good at getting suspects to confess. The confession then becomes the justification for the initial hunch. Confessions usually lead to conviction (whether the confession was genuine or not), meaning more success , and a greater enhancement of the policeman’s confidence. Of course, there would most likely failures in there as well, maybe as often as 46% of the time, but the policeman will rationalize them or forget them entirely. His faith in his “nose for a criminal”, like a faith in ESP or in John Edward’s skill in talking to the dead, is born .

The police’s investigation of Knox, and their aggressive interrogation, was predicated on their belief–based on their hunches about her “odd” behavior and mistranslation of a text message Knox sent the night of the murder–that Knox conspired with Sollecito and her boss, Patrick Lumumba, to kill her roommate. And so they interrogated Knox for hours, breaking her down until she was willing to say that she could imagine a scenario in which she’d been in the apartment when Lumumba killed Kercher.

Lumumba’s alibi and the test results on the physical evidence found at the scene made this scenario untenable. The physical evidence pointed instead to Rudy Guede, who was nowhere among Knox’s or Sollecito’s contact lists and whom Knox had met, briefly, only once.

Still, like “unsinkable rubber ducks”, the police and prosecutor refused to let new evidence contradict their belief in their “nose for the criminal”. They instead chose to defend that faith, concocting a scenario that included Knox and Sollecito in a murder plot with this stranger, and then sending their  investigators back to the scene to hunt for something, anything, to back up their new story.

Their narrative of the crime has since gone through several iterations, changing every time a piece of physical and testimonial evidence against the couple is revealed as a chimera. Why bother? Why not admit the mistake and move on? Likely because it’s easier, emotionally and professionally, to rationalize debunked evidence than it is to admit that the police methods they trusted were faulty, that their finely honed senses aren’t quite so fine, in short, to admit to themselves and the world that they blew it.

I can understand, and even to a limited degree sympathize with those feelings. It’s fun to feel right and to feel righteous, and what can be more righteous than seeking justice for a murdered young woman? It’s also hard to admit an error after one has put a great deal of time and effort into building and justifying an entire belief system based on it.

For prosecutors and police to admit error now would mean admitting that their blinkered inability to see the facts in front of them did incalculable damage to two young lives while helping the real killer get out of prison far earlier than should have been possible. (Guede will be getting work release this year, most likely.) It would mean disgrace in front of the whole world.

So what can be done about them?

The people in the Italian justice system responsible for Knox and Sollecito’s predicament  shouldn’t, as a group, be dismissed as simply corrupt, stupid, or evil. I don’t think they’re conspiring against Knox and Sollecito for kicks.  They’re intelligent people in the grip of faulty ideas, which means that when they paper over the Grand Canyon sized holes in their case with post-hoc rationalization, speculation, and dubious inference, their intelligence allows them to explain their views–both to others and to themselves–in ways that can sound reasonable and convincing.

Again, from Why People Believe Weird Things:

Psychologist David Perkins conducted an interesting correlational study in which he found a strong positive correlation between intelligence (measured by a standard IQ test) and the ability to give reasons for taking a point of view and defending that position; he also found a strong negative correlation between intelligence and the ability to consider alternatives. That is, the higher the IQ, the greater the potential for ideological immunity.

This describes very well the police, prosecutors, and several of the judges involved in the various stages of the Knox case, who have refused to recognize their own biases and nurtured a brazen overconfidence in their methods of evaluating evidence and reconstructing reality. Their flawed approach to reasoning has made it impossible for them to see the case evidence in its proper perspective and use it come to the most sensible conclusion.

Six years on, it’s not likely we can expect much change from them. As my Dad would say, they have too much time in. The prosecutors, the original trial judge, and the judges of the Italian Supreme court are like Fred Leuchter: they’re past the point where they can conceive of having been wrong. Knox’s best hope now is that the judges in the current trial are, like Judge Hellmann, able to see the evidence anew, recognize their predecessors’ mistakes, and find the courage to try to correct them.

UPDATE: 1/30. Well, so much for that best hope.